What is legibility, and why does it matter?

I came across the term “legible company” a few months ago while helping a group of founders adapt their operations to a new generation of AI tools. I kept hitting the same wall. Not a technical wall. An organizational one. The tools were capable. The companies were not readable.

Legibility, in this context, means something precise: the degree to which a company’s knowledge, workflows, data, and decisions are structured in a way that AI agents can actually read, parse, and act on. Not just store. Act on.

In practice, this means knowledge written down rather than held in people’s heads. Data in structured formats rather than screenshots of spreadsheets or PDFs of PDFs. Context attached to decisions, not assumed. Consistent labels, clean ownership, explicit operating logic.

The reason I kept thinking about it is that its absence is usually the real reason early automations fail. Not the model. Not the tool. The input. Garbage in, mediocre out.

What a legible company actually looks like

The difference between a legible and a non-legible company is not about having more software. It’s about whether the software knows what it’s doing.

In a traditional, non-legible company: customer feedback lives in five different inboxes, meeting notes are in someone’s personal Notion, pricing decisions happened in a call that nobody documented, and the person who understood the product roadmap left eight months ago. An agent dropped into this environment is essentially blind. It will produce output, but that output will be wrong in ways that are hard to detect.

In a legible company: there is a system of record for every domain. Decisions carry context. Workflows have defined handoffs. An agent can query what happened, understand why, and take the next step without a human translating everything first.

The productivity delta between these two states is not incremental. It’s structural.

Levels of AI-native: not all “AI-forward” is the same

Ann Miura-Ko, co-founding partner at Floodgate, recently mapped AI organizational maturity as a spectrum — from L0, where humans coordinate with humans and occasionally use AI to summarize or draft, to L5, where the organization essentially runs itself and humans are optional. The framework is useful precisely because it separates capability from aspiration. Four questions decide where you actually sit: Can the system see the work? Can it do the work? Can it extend itself? Can it change itself?

Most companies are somewhere between L1 and L2 and describing themselves as L4.

Legibility is the underlying condition that determines how far up that ladder you can actually go. You cannot give an agent meaningful autonomy over a system it cannot read. The autonomy levels are downstream of legibility levels.

This is where companies consistently get into trouble. They implement agents on top of messy, unstructured environments, get mediocre outputs, and blame the model. The model is fine. The foundation is broken. Early automations built on non-legible infrastructure are not prototypes of the future. They are expensive proof that the foundation needs fixing first.

The risk nobody is talking about: being too legible

Here is the part of the legibility conversation that is almost entirely absent from the current discourse.

Brian Halligan raised it recently: as founders race to make their companies legible to AI, they risk commoditizing their own moat. The mechanism is worth understanding slowly.

When a company makes its operating logic legible enough for an agent to execute, it also makes it legible enough for the vendors running those agents to pattern-match across customers. They see the playbook. They abstract it. They ship it as a best practice to the rest of the market. You don’t have to be worried about training data clauses to see this risk. You just have to notice what happens to knowledge once it moves from the ambiguity of human judgment into structured, queryable form: it becomes shareable, reproducible, and averaged.

I saw a version of this firsthand. At Meep, we had built backend connectivity logic that was cleaner and less convoluted than most of what we encountered during API integrations with third parties. That clarity was an advantage. It was also exactly what became visible the moment we opened an API. The integration required exposure. The exposure was the cost. The thing that made our integrations work well became a reference point for anyone who touched it.

Miura-Ko, responding in the same thread, points toward a more surgical approach: keep sensitive knowledge behind a firewall, use open source models where possible, and send context out only when strictly necessary. The principle underneath is right: legibility is not binary. It has a perimeter, and you get to decide where that perimeter sits.

Where to draw the line

Before you pipe something into a system, two questions are worth sitting with.

The first is about exposure. If the vendor running this agent could see this logic — not the output, the logic — would they learn something about how you operate that you’d rather they didn’t? Most founders answer this too quickly. The real test is not whether the information feels sensitive. It’s whether, in the hands of someone who serves your competitors too, it becomes a reference point.

The second is about the container. Legibility built on rigid infrastructure is almost worse than no legibility at all. A structure that can’t absorb change locks in the assumptions of the moment you built it. Before you make something legible, the prior question is whether the architecture you’re putting it into has enough elasticity to evolve as the company does. If it doesn’t, you’re not building a foundation. You’re building a constraint.

A third question, slower and less obvious: if someone read the legible version, could they reconstruct not just the process but the judgment underneath it? Process is often copyable. The judgment that produces the process usually isn’t — until you write it down.

Some things lose most of their value the moment they become readable. A few that I’ve seen made legible too quickly, usually under pressure to move fast:

The map of how complementary partnerships actually connect — not the partnerships themselves, but the sequencing logic, the joint GTM threads, the informal handshakes that make pieces fit together. The moment that’s in a document, it’s analyzable. And analyzable strategy is strategy anyone can counter.

Key workflows deployed for speed. The point of a fast workflow is that competitors don’t know it exists yet. Documentation is the beginning of its expiration.

Culture traits that make a team flow. These are the hardest to keep illegible because they feel like they should be written down — for onboarding, for hiring, for alignment. The trap is that fully articulated culture is a description of the past. The living version is always slightly ahead of what you can write about it.

And the pricing and negotiation layer — not just price floors, but the read you do on a customer before you name a number, the concession patterns, the judgment calls that happen in the room. Once that’s structured data, it’s modelable. And modelable negotiation is negotiation with no edge.

Two different problems: rebuilding vs. building from scratch

The approach to legibility looks completely different depending on which problem you have.

If you are rebuilding a non-AI-native company, the first step is not automation. It is archaeology. You have to map what actually exists before you can make it readable. That means identifying the real systems of record (often not the official ones), surfacing the tribal knowledge before the people carrying it leave, and accepting that the first round of legibility work will feel slow and produce almost nothing useful for agents. That’s correct. That’s the foundation.

If you are building from scratch, you have a different and rarer advantage: you can design for legibility from day one. Every workflow documented as it’s created. Every decision carrying its context. Every handoff explicit. AI-native companies built this way do not need to retrofit legibility. It is the product. The risk here is different: moving so fast that you over-document, making everything legible including the things that should stay embodied.

Both paths require discipline. The rebuilder’s discipline is patience. The builder’s discipline is restraint.

How to use this when evaluating a company

Everything is AI-native now. Every deck says so. Every founder says so.

A more useful question — whether you’re evaluating a company as a vendor, a partner, or a potential employer — is: what level of legibility does this company actually operate at? Can their systems see the work? Can agents do the work, or is a human still translating at every step? And the question almost nobody asks: what have they deliberately kept illegible, and why?

The last question is the most diagnostic. A company with no answer to it either hasn’t thought about it, or has made everything queryable and is running on borrowed time. A company with a clear answer has probably thought harder about competitive advantage than most.

AI-native is not a label. It is an architecture. And like most architectures, what matters is not what it can do at the surface. It is what it was built to protect underneath.

Originally published on LinkedIn in January 2026

The ground is shifting beneath transport operators, and most haven’t noticed.

On one front, AI agents are already accessing your services through channels you don’t control. Every major AI platform is pulling transport information through web scraping, public API queries, and third-party aggregators. Your schedules, routes, and real-time data are being represented to millions of users by systems you’ve never integrated with, can’t monitor, and have no way to correct when they get it wrong. This isn’t a future threat—it’s today’s reality.

On another front, hundreds of new companies are being born with agentic commerce as their founding premise. They’re not building better apps or prettier interfaces. They’re building AI-native platforms designed from day one for a world where agents intermediate every transaction, where AGI capabilities reshape what software can do, and where AI tools become embedded in every digital service. These startups understand something incumbents often miss: in an agentic economy, whoever controls the intelligence layer captures the value.

These two forces are converging into a single strategic reality for transport operators.

The company that orchestrates the journey owns the customer relationship. The operator that merely provides the vehicle becomes a commodity—interchangeable, price-compressed, invisible. Think about what happened to electricity. Essential infrastructure, but when was the last time you chose your utility provider based on anything other than price? That’s the future waiting for transport operators who cede the intelligence layer to others. You become the power grid: necessary, regulated, commoditized. Someone else builds the applications, owns the customer, captures the margin.

The agentic startups entering your market don’t need to own buses or trains. They just need to intermediate more efficiently than you can serve customers directly. And with AI agents, they will—unless you act first.

The question isn’t whether your operations will be accessible through AI agents. It’s whether you’ll control how that happens, and whether you’ll capture the value it creates.

The Gen AI Paradox Hits Transport

Here’s an uncomfortable truth from recent McKinsey research: 78% of companies are now using generative AI. Yet more than 80% report no material impact on earnings.

Transport operators face the same paradox. You’ve deployed chatbots, piloted copilots, experimented with AI-assisted scheduling. The results? Modest gains, hard to measure, difficult to scale. And most operators are asking the wrong question—focused on how to “ChatGPT” their existing direct-to-consumer channels, adding conversational interfaces to apps and websites they already own.

But while you’re optimizing your app, your customers are abandoning apps altogether. They’re asking AI assistants to plan their journeys, expecting agents to handle the complexity. The DTC channel you’re busy enhancing may not be where passengers interact with you at all in three years.

This isn’t a technology problem. It’s a deployment problem.

The companies stuck in the paradox made a common mistake: they deployed AI to automate existing tasks rather than reinvent underlying processes. They added intelligence to broken workflows instead of building new ones around what intelligence makes possible.

Consider how this plays out in passenger service during disruptions. McKinsey’s research on agentic AI identifies three levels of transformation—originally illustrated through call center operations, but directly applicable to transport. The pattern translates clearly:

Level 1 - AI-Assisted: Human agents use AI to find alternatives faster. Result: 5-10% improvement in resolution time. The process is identical; tools are better.

Level 2 - AI-Optimized: AI autonomously handles routine rebooking within defined parameters, escalating edge cases to humans. Result: 20-40% improvement. The process is streamlined; humans handle exceptions.

Level 3 - AI-Reinvented: AI proactively detects disruptions, initiates rebooking before passengers know there’s a problem, communicates alternatives, and coordinates across operators automatically. Result: 60-90% improvement in resolution time. 80% of incidents are resolved without human intervention.

Most operators are stuck at Level 1. Some are pushing toward Level 2. Almost none have reached Level 3—yet that’s where the transformative value lives.

The difference isn’t the AI. It’s the willingness to redesign the process around what AI agents can actually do.

Three Futures for Transport Operators

Every transport operator will end up in one of three scenarios. The decisions you make in the next 18-24 months will determine which one.

Scenario A: The Disintermediated Operator

In this future, AI agents access your services through whatever means available—scraping websites, abusing public APIs, pulling from third-party aggregators. You have no control over how your services are represented. When an agent tells a passenger the 8:15 bus is on time and it isn’t, you bear the reputational damage. When agents route passengers away from your services because they lack accurate real-time data, you lose ridership without knowing why.

Revenue flows to platforms and aggregators. Your operational data—the patterns, the preferences, the demand signals—feeds someone else’s optimization engine. You become infrastructure: necessary but invisible, squeezed on margins, unable to differentiate.

This isn’t hypothetical. It’s the default trajectory for operators who wait.

Scenario B: The Disoriented Modern Operator

Here, you’ve recognized that AI matters—but you’ve misread the game.

You’ve established basic API connections to the agentic ecosystem. AI agents can access your schedules and book tickets through authorized channels. You’ve checked the box on “AI integration.” But you’ve also convinced yourself that you need to become a tech company. You’re building your own chatbot—essentially trying to replicate ChatGPT or Claude, but closed, proprietary, and focused solely on selling your tickets. You’re hiring AI teams to create consumer experiences that compete with platforms backed by billions in R&D.

The market has already moved past what you’re building. While you spend 18 months developing a conversational interface, OpenAI and Anthropic and Google ship three generations of improvements. Your chatbot launches feeling dated. Passengers try it once, find it inferior to asking their general-purpose AI assistant, and never return.

Meanwhile, your underlying operations remain unchanged. Your dispatch still runs on fixed schedules with human oversight. Route changes still require supervisor approval. Disruption response still takes 10-15 minutes to coordinate. You’ve built a shiny front door to a house that hasn’t been renovated.

The result: you’ve spent your innovation budget competing where you can’t win (consumer AI interfaces) while underinvesting where you actually could (operational transformation and ecosystem integration). You’ve connected to the future without transforming for it, and built products the market didn’t need.

Many operators will land here and call it digital transformation. They’ll point to their AI initiatives, their chatbot launch, their “innovation lab.” But they’ll still be operating the same processes, serving the same commoditized role, wondering why the investment didn’t pay off.

Scenario C: The Future Operator

This is what strategic clarity looks like.

You’ve understood a fundamental truth: you’re not going to out-build OpenAI on conversational AI. And you don’t need to. Instead, you’ve focused on what only you can provide—operational excellence made accessible to the agentic ecosystem.

Your processes have been redesigned around agent autonomy. Disruptions trigger automatic rerouting and passenger rebooking before anyone reaches a delayed stop. Predictive maintenance schedules repairs during low-impact periods. Capacity optimization runs continuously in real-time.

You’ve made yourself indispensable to the ecosystem rather than trying to replace it. When a passenger asks their AI assistant to plan a journey, your services surface first—not because you built a competing chatbot, but because your data is accurate, your APIs are robust, and your MCP integration provides the rich context agents need.

The result: same infrastructure, same vehicles—but 30% higher capacity utilization, 45% reduction in service disruptions, new revenue streams from ecosystem participation. You didn’t waste capital building consumer AI products; you invested in becoming the operator that every AI product wants to work with.

The gap between Scenario B and Scenario C isn’t budget. It’s strategic focus: whether you’re trying to become something you’re not, or becoming the best version of what you actually are.

The Economics of Early Action

Let’s be specific about what’s at stake.

Why Horizontal AI Fails and Vertical AI Wins

According to McKinsey’s research on the agentic AI advantage, the gen AI paradox persists because most deployments are “horizontal”—general-purpose copilots and chatbots that assist across many functions without transforming any of them. These tools scale quickly but deliver diffuse gains that are hard to measure and harder to compound.

Vertical AI is different. It targets specific processes within specific domains, embedding intelligence into workflows where it can be measured, optimized, and scaled. Research shows the travel and transport sector is already seeing this play out:

• Housekeeping optimization in hotels: 10-30% reduction in labor hours through intelligent task routing

• Dynamic bundling for airlines: 20-30% revenue uplift by personalizing ancillary offers in real-time

• Load factor optimization: 3-4% improvement through AI-driven demand forecasting vs. 1-2% with traditional analytics

• Loyalty personalization: 15-25% revenue uplift through individualized reward offers

These aren’t pilot results. They’re operational outcomes from organizations that moved past experimentation into industrial deployment.

Transport operators have equivalent opportunities sitting untapped:

• Predictive maintenance: 20-30% reduction in vehicle out-of-service time by catching issues before they cause breakdowns

• Real-time capacity management: Dynamic service adjustments that match supply to actual demand, not historical averages

• Disruption recovery: Coordinated response across modes and operators that turns 45-minute recovery windows into 5-minute ones

• Accessibility optimization: Real-time routing that accounts for elevator status, accessible vehicle availability, and companion coordination

• Personalized journey offerings: Tailored service bundles, fare options, and route suggestions based on individual travel patterns, preferences, and loyalty status—turning anonymous riders into known customers with higher lifetime value

The common thread: these aren’t general AI applications. They’re process-specific implementations that require domain expertise, operational data, and integration with existing systems. They require domain expertise, operational data, and deep integration with existing systems, exactly what generic AI deployments miss.

The Compounding Advantage

Early movers don’t just capture first-mover benefits. They establish compounding advantages that late entrants can’t easily replicate.

Consider how this works: You integrate your booking and fare systems with an MCP-enabled platform. AI agents start purchasing tickets and recommending your services. You gain data on how agents evaluate options, which fare combinations they prefer, what personalized bundles convert best. You optimize your offers based on actual agent behavior and passenger responses. Your services become the preferred choice for agent-mediated journeys. Agents recommend you more often. More transactions, better personalization, stronger position.

Meanwhile, the operator who waited is still figuring out their API strategy while you’re on your third iteration of agent-optimized operations.

This is the network effect in action. The value of integration compounds over time—not just for you, but across the ecosystem of operators, agents, and passengers who interact through the platform. Getting in early means participating in that value creation. Getting in late means paying to access value others created.

The Cost of Delay

Every month of delay has a measurable cost:

Data accumulation: AI systems improve through learning. Agents interacting with your services generate insights about passenger behavior, demand patterns, and service quality. That data has value—value that accrues to whoever captures it. Delay means that learning happens without you, through scraped data you don’t control and can’t monetize.

Integration complexity: The agentic ecosystem is evolving rapidly. Standards are being established. Patterns are being set. Operators who engage now help shape how transport integrates with AI agents. Operators who wait inherit decisions made without their input.

Competitive positioning: In fragmented markets, the operators who establish agent-trusted status first will be harder to displace. Once a passenger’s AI assistant learns that your services are reliable, well-integrated, and accurate, switching costs emerge. The inverse is also true: if agents learn your data is unreliable, that reputation persists.

Talent and capability: Organizations that start now build internal expertise—staff who understand how AI agents work, what they need, how to optimize for them. That institutional knowledge doesn’t appear instantly when you decide to catch up.

The executives waiting for the technology to “mature” are actually waiting until their competitors have insurmountable leads.

The Evolution Timeline

The transformation from today’s fragmented systems to tomorrow’s intelligent networks follows a clear progression—one that’s already underway.

Short-term (2026-2030): The Foundation and Operational Phase

This is where the ground rules get established and the first movers pull ahead.

AI agents become the primary interface for journey planning. Passengers stop comparing apps and start asking assistants. The operators with clean data feeds, robust APIs, and MCP integration become the default recommendations. Those without become invisible—not because agents deliberately exclude them, but because they simply can’t be discovered or trusted.

By the end of this window, agents move beyond recommending journeys to orchestrating them—directly interfacing with booking systems, dynamically adjusting capacity, coordinating across operators during disruptions. Demand-responsive transport shifts from exception to norm. Fixed schedules give way to dynamic optimization.

During this phase, the winners focus on four priorities: ensuring their real-time data is accurate and accessible, establishing controlled integration channels that give them visibility into how agents use their services, building the internal capabilities to learn from agent interactions, and—critically—organizing the proprietary data that isn’t publicly available. Your operational patterns, capacity utilization, passenger flow dynamics, pricing elasticity, maintenance histories—this is the data AI platforms can’t scrape. It’s the data that will power personalization, predictive services, and premium offerings in the agentic economy. Operators who structure and prepare this data now will have assets to monetize later. Those who don’t will watch competitors build intelligence layers on information they never thought to capture.

The operators who treat this as a “wait and see” period will find themselves playing catch-up for years. The foundation you build now determines your options later.

Medium-term (2030+): The Integration Phase

The boundaries between transport modes dissolve.

A passenger’s AI assistant doesn’t distinguish between your bus, a competitor’s metro, an autonomous shuttle, or a shared bike. It assembles the optimal journey from whatever components serve the passenger best. Operators who’ve established themselves as reliable, well-integrated options get included. Those who haven’t get bypassed.

Physical infrastructure begins adapting to AI recommendations—dynamic bus lanes, reconfigurable stations, demand-responsive routing that changes by the hour. The intelligence layer and the physical layer become inseparable.

This future rewards operators who started building relationships with the agentic ecosystem years earlier. The trust, the data flows, the operational patterns—these compound over time. Late entrants face an ecosystem that’s already optimized around the early movers.

A Day in the Life: 2030

María is getting ready in her apartment in northern Madrid when her home sound system plays a notification: “Your route to the coworking space has changed. Line 10 maintenance means I’ve switched you to express bus C1 and adjusted your first meeting by 10 minutes. You’ll arrive 5 minutes early and save €2 with your monthly pass.”

She never checked a screen. Never opened an app. Never compared options.

During her commute, her smartring vibrates gently at transfer points while her audio glasses provide quiet guidance—”Bus arriving in 90 seconds, platform 3”—keeping her oriented without ever pulling out a phone. When unexpected traffic hits Paseo de la Castellana, her assistant coordinates with the Metro system to hold her connecting train at Nuevos Ministerios for two minutes. She doesn’t even know there was a problem.

At midday, her 2pm meeting gets rescheduled to a location near Retiro Park. Her assistant spots a 40-minute window and books her a meditation class at a studio two blocks away—something she’d been meaning to try. It adjusts her afternoon schedule and coordinates with her colleague’s assistant to confirm the new meeting time.

Here’s what María doesn’t see:

Behind that seamless experience, an operator made choices. They connected their systems to an MCP-enabled platform. They established data governance policies. They invested in real-time feeds and in quality structured proprietary data. They chose controlled integration over uncontrolled scraping.

That operator is now the one María’s assistant trusts—not the one it works around.

The technology enabling María’s morning isn’t remarkable. What’s remarkable is how unremarkable it feels to her. The complexity has disappeared into ambient experience—no screens, no decisions, no friction. That only happens when infrastructure anticipates needs rather than reacting to requests.

The operators who built that infrastructure didn’t just survive the agentic transition. They defined it.

The CEO Mandate

McKinsey’s research on the agentic AI advantage concludes with a pointed observation: “The time has come to bring the gen AI experimentation phase to an end—a pivot only the CEO can make.”

This isn’t a technology decision that can be delegated to IT. It’s not an innovation initiative that belongs in a lab. It’s a strategic transformation that requires executive ownership.

Three Actions for Transport Leaders

1. End experimentation. Start executing.

Audit your current AI initiatives. How many are pilots that never scaled? How many delivered measurable business impact? How many are still “exploring potential”?

The honest answer for most organizations: too many experiments, too few results.

Retire the initiatives that won’t scale. Consolidate resources around the ones that will. Stop treating AI as a series of disconnected projects and start treating it as a transformation program with clear objectives, timelines, and accountability.

2. Reframe the investment question.

The wrong question: “How much should we spend on AI?”

The right question: “What’s the cost of being excluded from the agentic ecosystem?”

Every month without controlled integration, AI agents access your services through channels you don’t control, represent you in ways you can’t correct, and generate data you can’t capture. The investment isn’t optional; the only choice is whether you make it proactively or reactively.

Build the business case around risk mitigation, not just efficiency gains. The operators who get this right will capture disproportionate value. The ones who don’t will watch that value flow to competitors and intermediaries.

3. Launch a lighthouse project while building the foundation.

You don’t need to transform everything at once. Start with a bounded initiative that demonstrates value and builds organizational capability.

Information services are the natural starting point—ensuring AI agents can access accurate, real-time data about your operations through controlled channels. This can be operational in weeks, not years. It establishes the integration infrastructure, builds internal expertise, and generates data about how agents interact with your services.

From that foundation, expand into ticketing and booking integration, then operational coordination, then full ecosystem participation. Each phase builds on the last. Each generates learning that informs the next.

The lighthouse project isn’t the transformation. It’s the proof point that makes the transformation possible.

The Choice

The agentic economy isn’t arriving. It’s here.

AI agents are already accessing transport services—through channels operators don’t control, generating value operators don’t capture, building patterns operators can’t influence. Every day of delay widens the gap between the operators shaping this transition and those being shaped by it.

The parallel to what happened in other industries is exact. Hotels that dismissed online travel agencies as a niche channel watched those agencies capture the customer relationship. Retailers that viewed e-commerce as separate from “real” retail found themselves competing with Amazon. Media companies that treated digital as a side project discovered it was the main event.

Transport operators face the same inflection point. The companies that control the intelligence layer will capture the value. The companies that merely provide the underlying service will be commoditized.

But there’s a crucial difference from those earlier disruptions: this time, operators can see it coming. The technology is visible. The trajectory is clear. The playbook is emerging.

The operators who act now—who establish controlled integration, who redesign processes around agent capabilities, who build the organizational muscle to operate in an agentic ecosystem—will define how AI agents interact with transport. They’ll capture the data, the relationships, and the margins that come from being essential to the ecosystem rather than interchangeable within it.

The operators who wait will inherit an ecosystem designed without their input, optimized for someone else’s benefit, operating on someone else’s terms.

The question isn’t whether transportation will integrate with AI agents. It’s whether you’ll control how that integration happens. And whether you’ll capture the value it creates. Or whether someone else will.

Originally published on LinkedIn in December 2025.

The transport industry stands at an unprecedented technological inflection point. While artificial intelligence transforms every sector from healthcare to finance, urban mobility operators confront a unique challenge: how to evolve infrastructure that has served cities reliably for decades without compromising operational continuity or financial stability.

Unlike industries that can deploy AI solutions on greenfield digital platforms, transport operators must integrate intelligent systems with legacy infrastructure that often predates the internet itself. Bus dispatch systems from the 1990s, ticketing platforms built for paper-based processes, and scheduling software designed for static timetables must somehow accommodate AI agents that expect real-time data exchange and dynamic operational control.

This integration imperative cannot be delayed. AI agents are already accessing transport information through uncontrolled methods: scraping websites, exploiting public APIs, and creating fragmented passenger experiences that operators cannot influence or monetize. Meanwhile, passenger expectations, shaped by seamless AI integration in other aspects of their digital lives, increasingly demand the intelligent, proactive mobility services that only properly integrated systems can deliver.

The solution lies not in wholesale infrastructure replacement but in strategic integration that adds an intelligence layer to existing systems. This requires understanding which current systems can support AI integration, what foundational capabilities must be established, and how to implement changes that enhance rather than disrupt daily operations.

Understanding the Technical Foundation

Most operators already possess the core systems needed for AI agent integration, though they may not realize it. The challenge is creating the connectivity layer that allows intelligent agents to interact with your current operational systems through standardized protocols.

Computer-Aided Dispatch (CAD) and Automatic Vehicle Location (AVL) Systems form the backbone of modern transport operations. Whether you’re running traditional platforms like Trapeze and Init, or newer solutions like Swiftly, these systems already track vehicle positions in real-time. The challenge isn’t the data itself, it’s making that operational data accessible through standardized APIs and structured formats that AI agents can programmatically consume and act upon.

Backend Scheduling and Operations systems contain route information, timetables, and operational parameters that agents need for intelligent decision-making. Legacy mainframes and modern cloud systems alike can be connected to AI agents without requiring replacement, the key is creating appropriate interface layers that translate between your existing infrastructure and standardized agent protocols.

Ticketing and Payment Systems represent the most complex integration challenge, as agents must validate fares, process payments, and handle dynamic rebooking. Major platforms like Scheidt & Bachmann, Cubic, and Masabi each require specific approaches, but established integration patterns exist for connecting these systems to intelligent agents.

Passenger-Facing Applications and Information Systems include mobile apps, web platforms, real-time passenger information displays, and digital signage. These systems must evolve from static information displays to dynamic interfaces that respond to AI agent recommendations. Agents need to push real-time updates, alternative routes, and rebooking confirmations directly to passenger touchpoints, while collecting preference data and feedback that guides future optimization decisions.

The Three-Pillar Operational Foundation

Before any technical integration, operators must establish three foundational capabilities that determine whether AI implementation succeeds or creates operational chaos:

Data Ownership and Control Architecture

Operators need granular policies defining what information AI agents can access and under what conditions. Real-time vehicle locations might flow freely to verified agents, while passenger capacity data requires explicit partner authorization. This control architecture must be defined upfront, not retrofitted after agents already access your systems.

Real-Time Data Management Capability

AI agents become useless with stale information. Your systems must provide data current within seconds, not minutes or hours. If your current infrastructure updates every five minutes, that’s your starting point but AI integration requires moving toward true real-time feeds that benefit both agent functionality and overall operational efficiency.

Industry Standards Compliance

Data must be accessible through industry standards: GTFS, GTFS-RT, SIRI, GBFS, MDS. AI agents communicate through protocols that transform standardized data into intelligent decisions. Proprietary, locked formats make your services invisible to the emerging agent ecosystem.

Overcoming Common Implementation Barriers

Technical Infrastructure Barriers: Legacy systems don’t prevent AI integration. Meep wraps existing technology, transforming data into AI-compatible formats without requiring infrastructure modernization. Even decades-old mainframes can connect to AI agents through appropriate middleware layers that handle translation complexity.

Data Quality and Standardization: Transport data exists in multiple formats with inconsistent quality across operators. AI agents can clean, standardize, and enrich data in real-time, creating coherent streams from fragmented sources. MCP’s contextual awareness helps agents work with imperfect data while identifying quality issues for continuous improvement.

Regulatory and Policy Barriers: Transport systems require exceptional reliability and safety standards. The solution is gradual automation progression, start with AI agents in advisory roles, implement robust fail-safe mechanisms, and maintain human oversight capabilities. Agents can actually improve regulatory compliance by automatically flagging potential violations and maintaining comprehensive audit trails.

Organizational Change Barriers: AI agents will transform jobs rather than eliminate them. Traditional roles evolve while new positions requiring different skills emerge. Success comes from focusing on human-AI collaboration, emphasizing roles requiring human judgment, creativity, and interpersonal skills. Comprehensive retraining programs help existing staff transition to AI-augmented operations, making their operational knowledge more valuable when combined with AI capabilities.

The Business Case for Early Action

Operators who integrate AI agents now gain multiple competitive advantages. They maintain control over how their data is accessed and used, rather than having AI systems scrape information without control. They participate in new revenue streams from agent-mediated bookings instead of being bypassed by intermediaries.

Early integrators also gather operational intelligence about passenger behavior and network optimization that improves service delivery. This data advantage compounds over time—operators with six months of AI agent analytics understand their systems better than competitors still relying on traditional reporting.

Beyond operational insights, AI agent integration creates multiple strategic advantages. Early adopters often discover new revenue streams through agent-mediated services—premium routing options, dynamic pricing optimization, or partnerships with AI platforms seeking transport integration. The market recognizes innovation leadership, generating positive publicity that attracts both passengers and potential partners who want to work with forward-thinking operators.

Perhaps most valuably, operators frequently uncover unexpected use cases during implementation. AI agents might reveal demand patterns that justify new routes, identify maintenance optimization opportunities that reduce costs, or suggest service modifications that increase ridership. These discoveries often emerge from the rich data interactions that only become visible once intelligent agents begin analyzing operations comprehensively. Early implementers gain exclusive access to these insights while competitors remain limited to traditional analysis methods.

Risk of Delayed Implementation

Every day of delay increases the risk of losing control over the integration process. AI agents are already accessing transport data through website scraping and unauthorized API usage. This uncontrolled access creates poor passenger experiences, potential liability issues, and gives operators no insight into how their services are being represented.

More critically, passengers are beginning to expect agent-mediated mobility. Once they experience AI-orchestrated journeys in one city, they demand similar capabilities everywhere. Operators who cannot provide integrated, intelligent mobility face the same trajectory as hotels that couldn’t integrate with booking platforms, gradual marginalization as passengers choose operators who meet modern expectations.

Consider this scenario: Your morning commute requires no manual planning or app-checking. An AI mobility assistant analyzes your calendar, weather conditions, and real-time transport data across the entire city network. It automatically secures your optimal journey, proactively adjusts for potential disruptions, and coordinates seamlessly with every transport operator required—all before you wake.

This represents the Agentic Future of mobility—the direction urban transportation is rapidly heading.

What Makes This Different from Today’s Apps?

We’ve all experienced the frustration: checking three different apps to plan one journey, missing connections because systems don’t talk to each other, standing at bus stops with no real information. Traditional transport apps are fundamentally reactive, they show you schedules and let you book trips, but they can’t think ahead or adapt dynamically.

AI agents represent a complete paradigm shift from reactive to proactive mobility management. Instead of multiple app interfaces, you configure your preferences once: preferred transport modes, accessibility needs, budget constraints, sustainability priorities, timing flexibility. The user interface transforms from today’s trip-by-trip booking systems to preference dashboards and passive journey monitoring.

The technical architecture changes dramatically too. Rather than connecting to individual operator apps, AI agents interface directly with transport services through standardized protocols, enabling real-time coordination across entire networks. This direct integration resolves persistent issues that plague current systems: the lack of accuracy in arrival predictions, absence of true multimodal planning, inability to adapt when plans change mid-journey, and delayed or missing real-time data.

Once configured, users assume a largely passive role. The agent continuously monitors your calendar, transport conditions, and service availability, making autonomous decisions and only surfacing information when intervention is needed or better options emerge. Think of them as digital assistants that don’t wait for commands but proactively solve mobility challenges before you encounter them.

The Paradigm Shift: From Fragmented APIs to Intelligent Integration

Today’s transport ecosystem operates through fragmented APIs and isolated applications. Each operator maintains separate systems with different capabilities—some provide real-time data, others don’t; some allow booking, others are information-only; payment systems rarely connect across operators. This fragmentation forces passengers to navigate multiple interfaces and creates operational silos that prevent true coordination.

The emergence of Large Language Models and AI agents is driving a fundamental shift toward the Model Context Protocol (MCP). Unlike traditional APIs that require rigid, predetermined interactions, MCP enables AI agents to dynamically discover services, understand their capabilities, and interact intelligently without requiring pre-programmed instructions for every scenario.

Instead of receiving basic data like “here’s the bus schedule,” agents gain comprehensive intelligence: schedule information plus real-time delays, accessibility features, current capacity, alternative routes, payment options, and service policies. This contextual richness allows agents to make sophisticated decisions considering the full complexity of urban mobility.

This transformation demands new operational infrastructure. Transport systems need enhanced data quality and standardization to support intelligent decision-making. Ticketing systems must evolve to allow agents to purchase and modify reservations autonomously, not just display options for manual booking. Flexible fare structures become essential to enable dynamic rebooking and route changes as conditions evolve.

Most critically, connectivity between previously isolated systems becomes mandatory. Agents must coordinate between ride-hailing services like Uber and public transit, orchestrate multimodal journeys across different operators, and provide sophisticated notification systems that keep both passengers and operators informed of changes in real-time. The intelligence layer only works when systems can communicate seamlessly across the entire transport ecosystem.

The Strategic Imperative: Building on Existing Infrastructure

The most compelling aspect of this transformation is that it doesn’t require replacing existing infrastructure. The buses, trains, and metros serving cities for decades don’t become obsolete—they become exponentially more capable when intelligence flows through their networks. This represents evolution, not revolution.

At Meep, we’ve spent years building the integration infrastructure that makes this transition possible. Our platform already connects 175+ transport operators across 28 cities, creating the connectivity foundation AI agents need to orchestrate seamless multimodal journeys. This existing network demonstrates that intelligent integration can happen without wholesale infrastructure replacement.

However, we’re at a critical inflection point that demands immediate action. AI agents are already accessing transport information—often through scraping websites and unauthorized API usage. This uncontrolled access creates poor user experiences, liability issues for operators, and gives transport providers no control over how their data is used or monetized by AI platforms.

The operators who proactively integrate AI agents through controlled platforms will define the future of urban mobility. They’ll maintain control over their data, participate in new revenue streams, and shape how AI serves their passengers. Meep serves as the strategic bridge enabling this controlled integration, allowing operators to enter the agentic era on their terms rather than being subject to external disruption.

Those who delay face a binary outcome: either their competitors establish decisive advantages while passengers adopt agent-mediated mobility through other channels, or they risk being replaced entirely by tech players who build direct relationships with passengers, reducing traditional operators to mere infrastructure commodities. The choice isn’t whether to integrate with AI agents—it’s whether to control that integration or be controlled by it.

The Choice Ahead

The agentic revolution isn’t coming—it’s here. Every day operators delay, more AI systems access their data uncontrolled, more passengers experience disconnected transport in an increasingly integrated world, and more revenue flows to intermediaries rather than the operators providing actual transportation.

The transformation of other industries provides clear warning signals for transport. AI agents have already reshaped online advertising, fundamentally altered how consumers search and shop, and created new intermediary layers that capture value from traditional providers. These early implementations demonstrate the pattern: industries that fail to integrate intelligently find themselves disintermediated, while those that adapt proactively maintain control and capture new value streams.

This is about more than technology—it’s about ensuring that efficiency gains benefit passengers, optimization considers equity, and human judgment guides automation rather than being displaced by it.

The future of mobility will be intelligent, integrated, and adaptive. The only question is whether traditional transport operators will shape this evolution or find themselves relegated to infrastructure providers in someone else’s vision.

Durante mi reciente visita al evento de UATP en Hamburgo, mi socio y yo decidimos probar MOIA , el servicio de transporte a demanda de Volkswagen. Dos trayectos después —uno urbano, otro al aeropuerto— y con una experiencia excelente en el bolsillo, nos sumergimos en una conversación que no puedo dejar de compartir: ¿hacia dónde va realmente el DRT en Europa?

La pregunta no es trivial. Estamos en un momento clave donde las decisiones que tomemos hoy sobre el transporte a demanda definirán si se convierte en una herramienta de sostenibilidad urbana o en otro lujo tecnológico más.

MOIA: Cuando la abundancia se vuelve premium

MOIA no es un ride-pooling asequible como Uber Share, aunque así se autodenominen. Es un servicio DRT urbano premium, operado con una flota de unos 500 vehículos eléctricos en Hamburgo. Los números hablan por sí solos:

• 450,000 usuarios registrados

• 650,000 viajes por trimestre

• Vehículos con Wi-Fi, USB y paradas virtuales a 150-200 metros del usuario

• Precio similar a Uber, pero con la promesa de ser “compartido”

Aquí está el primer dilema: hablando con startups que desarrollan software para DRT, nos contaron algo revelador. En una ocasión, tres usuarios pidieron el mismo trayecto a la misma hora, y el sistema de MOIA asignó tres vehículos distintos habiendo espacio para ir en una sola furgoneta.

¿Por qué? Porque priorizan la experiencia individual sobre la eficiencia colectiva. Es decir, prefieren que te sientas como en un Uber antes que optimizar realmente el recurso compartido.

Y aquí es donde el concepto de “ride-pooling” hace aguas. Si un sistema que se autodenomina “compartido” asigna vehículos individuales cuando podría agrupar usuarios, ¿realmente estamos hablando de transporte colectivo? O simplemente hemos encontrado una forma elegante de llamar “sostenible” a lo que sigue siendo consumo individual disfrazado.

La Teoría de la Abundancia: ¿más es siempre mejor?

Esta situación conecta con un debate más amplio que se está gestando en los círculos de políticas públicas: el de la “transport abundance” o abundancia en el transporte. Pensadores como Ezra Klein y Derek Thompson defienden que deberíamos reducir las barreras al acceso a la movilidad, aumentando masivamente la oferta:

• Más infraestructura

• Más modos de transporte

• Más operadores compitiendo

• Menos regulaciones restrictivas

La premisa es simple: si hay más opciones disponibles, las personas se moverán mejor y más eficientemente.

Y en Hamburgo, esta teoría parece funcionar. La ciudad cuenta con un ecosistema de movilidad envidiable: transporte público robusto (bus y metro), taxis tradicionales, Uber, micromovilidad en todas sus variantes, y ahora MOIA completando el puzzle. La movilidad global es, efectivamente, excelente.

Pero aquí hay una trampa.

La crítica de Fernando Fernández-Monge: Calidad sobre Cantidad

Fernando Fernandez-Monge introduce un matiz crucial en su blog: no necesitamos más movimiento, sino mejor movimiento. Su argumento es poderoso: la abundancia sin dirección puede generar más problemas que soluciones. No se trata solo de subvencionar más autobuses o infraestructuras, sino de incentivar los comportamientos correctos.

Ejemplos como el Mobility Impact Market, que paga a operadores según su impacto en CO₂ y salud pública, o el sistema de Certificados de Ahorro Energético (CAE) en España para fomentar el carpooling urbano, muestran cómo alinear intereses públicos y privados hacia una movilidad realmente eficiente.

España y el DRT: la escasez mal dirigida

Mientras Hamburgo experimenta con abundancia premium, España mantiene una visión restrictiva del DRT. Según datos del Observatorio de Transporte y Logística, aquí se concibe como:

• Servicio para zonas rurales exclusivamente

• Sustituto de líneas regulares deficitarias

• Transporte social con tarifas subsidiadas

• Sin posibilidad de escalar en entornos urbanos

Esta aproximación asistencialista desaprovecha una oportunidad enorme: el DRT como modo intermedio entre el autobús y el taxi, capaz de descongestionar líneas saturadas, cubrir eficientemente trayectos de media demanda, e incentivar usos del transporte más eficientes como describe Fdez-Monge. Es como si hubiéramos decidido que el DRT es un “transporte de segunda” cuanto podría aportar una pieza clave del puzzle de movilidad urbana.

La Paradoja de Tres Modelos

Aquí tenemos tres aproximaciones al DRT que ilustran perfectamente la tensión entre abundancia y dirección:

Modelo Hamburgo (MOIA): Abundancia Ciega

• Filosofía: Más opciones para todos, experiencia premium

• Realidad: 500 vehículos priorizando comodidad individual

• Resultado: Éxito comercial, pero ¿impacto sostenible cuestionable?

Modelo España: Escasez Mal Dirigida

• Filosofía: DRT solo para casos específicos y marginales

• Realidad: Impacto limitado, sin escalabilidad urbana

• Resultado: Oportunidad desaprovechada de crear modos intermedios

Modelo Ideal: Abundancia Inteligente

• Filosofía: Más opciones, pero orientadas hacia comportamientos sostenibles

• Herramientas: Incentivos que premien la eficiencia real

• Objetivo: El mejor de ambos mundos

¿Es Posible la Abundancia Inteligente?

La pregunta del millón es: ¿puede el DRT ser abundante Y dirigido a la vez? Creo que sí, pero requiere un cambio de mentalidad en tres niveles:

🚌 Para Operadores

No competir solo en comodidad, sino en impacto sostenible real. Desarrollar algoritmos que optimicen la ocupación genuina, no la percepción de servicio premium. Integrar métricas de CO₂ y salud pública en los KPIs principales.

🏛️ Para Administraciones

Abandonar la visión del DRT como “transporte de segunda” y crear marcos regulatorios que incentiven el agrupamiento eficiente. Implementar sistemas de pago por impacto, como ya se está haciendo con los CAE.

💻 Para Startups Tecnológicas

Enfocar la innovación en optimización colectiva, no individual. El reto no es hacer que cada usuario se sienta especial, sino maximizar el beneficio del sistema completo.

Y aquí hay una oportunidad clave: las startups que provean herramientas de datos holísticas pueden dar a todos los actores las palancas necesarias para que el sistema se mueva de manera sincronizada. Si un operador de DRT no tiene información de lo que pasa en otros modos —saturación del metro, picos de demanda de buses, disponibilidad de micromovilidad— muy difícilmente podrá asignar sus recursos de manera dirigida hacia donde realmente se necesita.

El Futuro que Necesitamos

Hamburgo nos mostró que el DRT puede escalar, puede ser atractivo y puede integrarse sin competir destructivamente con el transporte público. Pero también nos enseñó los riesgos de una abundancia sin dirección.

España, por su parte, nos recuerda que la escasez dirigida tampoco es la respuesta. Necesitamos más opciones, pero mejores opciones.

El DRT del futuro debería ser:

• Un nuevo modo de transporte que cubre necesidades diferentes y se sitúa entre el transporte colectivo de alta frecuencia y el ride-hailing o taxi

• Una solución eficaz para zonas de media demanda urbana

• Una herramienta para descongestionar líneas saturadas

• Un espacio de innovación para startups que aportan tecnología puntera

Mi vehículo llegando al punto de recogida

La Pregunta Final

Si queremos evitar que la abundancia en el transporte se traduzca en más emisiones, más coches y más congestión, debemos apostar por políticas que incentiven la demanda correcta. No se trata de elegir entre más opciones o mejores opciones. Se trata de crear más opciones mejores.

¿Nos atrevemos a imaginar un DRT que sea abundante, accesible Y sostenible? La tecnología ya existe. La experiencia de Hamburgo demuestra que la demanda está ahí. Solo falta la voluntad de diseñar sistemas que optimicen para el bien común, no solo para la satisfacción individual.

El futuro del DRT y de la movilidad urbana en general depende de que encontremos este equilibrio. Y el momento de empezar es ahora.

In our current reality of rapid technological disruptions, it’s critical to understand how and where we can act effectively. The technologies we handle daily are becoming more immediate, their effects more radical, and their consequences more unpredictable. This makes it essential to consider certain aspects when making strategic decisions to avoid falling into the gray zone between pioneers and the backward-looking or naive. The starting point is to use tools to face changes, hypothesize scenarios, and anticipate consequences.

Internationally, technological bipolarity between the United States and China has generated an overabundance of companies with direct influence on all levels and aspects of our lives. While the US maintains its global economic leadership, China is advancing rapidly with developments like Deepseek and other proprietary AI solutions. This competition has allowed new technological innovations, such as AI or quantum computing as a service, to emerge. These innovations are less than a decade old but have generated a seismic effect with chain reactions transforming industries and relativizing the importance of global economies.

Looking at last year’s major technological disruption, generative artificial intelligence has infiltrated all our daily tasks. OpenAI with the GPT-2 model from 2019 held the record as the first digital platform to reach 100 million users in two months and currently has 300 million weekly active users. Similarly, in China, Deepseek has experienced exponential growth, positioning itself as the main Asian competitor and attracting millions of users overnight. In perspective, this represents an adoption rate 2700% faster than Facebook in reaching 100 million users, or 887.5% faster than Instagram in reaching 10 million users per day. Since these percentages are unfathomable to our minds, we can compare the speeds of walking versus driving a Formula 1 car. And this adoption speed is coupled with funding initiatives like the Stargate project, where the United States will invest $500 billion over four years.

This massive usage has generated multiple changes in five aspects of our society. The first, automation and productivity where 75% of global workers claim to use these solutions in their daily tasks, according to LinkedIn and Microsoft’s 2024 Labor Trends Index. The second, job creation and transformation; this innovation redefines roles, supplants some and pressures the demand for others for certain new needs that have emerged in companies. The third, the sectoral impact of tools where certain industries have been more affected than others. The fourth aspect concerns changes in skills and competencies of employees who have to adapt to the new needs of the labor market. And the fifth, the perception and adoption of these tools where there exists great uncertainty among workers because their jobs may eventually run the risk of disappearing.

The Collingridge Dilemma

This is the game on which we play as professionals, companies, and countries. And this is where it’s imperative to stop and reflect to chart the best strategy for personal, professional, and institutional success. Not all of us have the same tools or resources to compete on equal terms, but David Collingridge published in 1980 the book “The Social Control of Technology,” a fundamental concept for evaluating the impact of technology. The Collingridge dilemma is essential to understand when regulating technology and innovation. It poses a timing problem: On one hand, if we regulate a technology too early, when it’s still developing and hasn’t reached mass adoption, we cannot foresee all the negative consequences of such regulation. This could lead to ineffective or even unnecessary regulation. On the other hand, if we wait too long to regulate it, the technology will already be widespread and difficult to modify without significant costs, political conflicts, or industry resistance.

This dilemma makes particular sense today with regulatory proposals such as the EU’s Artificial Intelligence Act, which aims to balance the supposed challenges of AI systems according to their risk level: unacceptable, high, limited, and minimal. Here, the EU seeks to establish safeguards for the reliable and responsible use of AI to ensure transparency, but there are three major insurmountable obstacles to this seemingly laudable mission:

1. Leading companies in these technologies are private and foreign.

2. The strategic advantage of the United States and China can establish local regulations for their companies to advance without restrictions to the detriment of European ones, as demonstrated by Chinese government support for Deepseek.

3. The power of Big Tech lies in offering different services by regions, relegating Europe to a disadvantaged position.

We can state with certainty that these technologies are very incipient, but their great penetration makes them more difficult to limit through regulation. Even thinking that the winning way is proprietary technological development, such as the tool recently launched by the Spanish Government, ALIA. This open-source linguistic tool represents an effort to develop AI in Spanish and promote digital sovereignty, but its practical relevance and capabilities lag far behind those of other market solutions.

In this scenario, where geopolitics has also gained significant weight, EU companies and professionals start with a clear disadvantage. It’s important to define a winning strategy as soon as possible, a winning strategy to promote the use of the most relevant, consolidated, and functional technologies. Only then can innovations originating in Europe exist to compete with other regions.

Debido a la realidad que vivimos donde existen rápidas disrupciones tecnológicas es necesario saber hasta dónde podemos actuar y cómo. Las tecnologías que manejamos cada día son: más inmediatas, sus efectos más radicales y sus consecuencias más imprevistas. Por esto es importante tener en cuenta ciertos aspectos a la hora de tomar decisiones estratégicas para no caer en esa zona gris que está a medio camino entre los pioneros y los retrógrados o ilusos. Usar las herramientas para afrontar los cambios, hacer hipótesis de escenarios y poder adelantarse a las consecuencias de estos es el punto de partida.

A nivel internacional vemos claramente que la bipolaridad tecnológica entre Estados Unidos y China ha generado una sobreabundancia de empresas con influencia directa en todos los niveles y aspectos de nuestras vidas. Mientras EE UU mantiene su liderazgo económico mundial, China avanza rápidamente con desarrollos como Deepseek y otras soluciones de IA propias. Esta competencia ha permitido que surjan nuevas innovaciones tecnológicas como la IA o la computación cuántica como servicio. Estas innovaciones tienen menos de una década pero han generado un efecto sísmico con reacciones en cadena transformando industrias y relativizando la importancia de las economías mundiales.

Si observamos la gran disrupción tecnológica del año pasado, la inteligencia artificial generativa se ha colado en todas nuestras tareas diarias. OpenAI con el modelo GPT-2 del año 2019 ostentaba el récord de ser la primera plataforma digital en llegar a 100 millones de usuarios en dos meses y actualmente cuenta con 300 millones de personas activas semanalmente. Paralelamente, en China, Deepseek ha experimentado un crecimiento exponencial similar, posicionándose como el principal competidor asiático y atrayendo millones de usuarios de la noche a la mañana. En perspectiva esto supone una adopción de un 2700% más rápido que Facebook en llegar a los 100 millones de usuarios, o un 887.5% más rápido que Instagram en alcanzar 10 millones de usuarios al día. Como los porcentajes son inasumibles por nuestra mente, podemos comparar las velocidades de ir caminando con la de ir conduciendo un coche de fórmula 1. Y a esa velocidad de adopción se une el impulso de iniciativas de financiación como el proyecto Stargate donde Estados Unidos invertirá 500.000 millones de dólares en cuatro años.

Este uso masivo ha generado múltiples cambios en cinco aspectos de nuestra sociedad. El primero, la automatización y productividad donde el 75% de los trabajadores globales afirma usar estas soluciones en sus labores diarias, según el Índice de Tendencias Laborales de LinkedIn y Microsoft de 2024. El segundo, la creación y transformación de empleos; esta innovación redefine roles, suplanta algunos y presiona la demanda de otros para determinadas necesidades nuevas que han surgido en las empresas. El tercero, el impacto sectorial de las herramientas donde ciertas industrias se han visto más afectadas que otras. El cuarto aspecto atañe a los cambios en habilidades y competencias de los empleados que tienen que adaptarse a las nuevas necesidades del mercado laboral. Y el quinto, la percepción y adopción de estas herramientas donde existe una gran incertidumbre de los trabajadores porque sus puestos de trabajo puedan eventualmente correr el riesgo de desaparecer.

Este es el tablero en el que jugamos como profesionales, empresas y países. Y es aquí donde es perentorio pararse a reflexionar para trazar la mejor estrategia del éxito personal, profesional e institucional. No todos contamos con las mismas herramientas ni recursos para competir en igualdad de condiciones, pero David Collingridge publicó en 1980 el libro “El Control Social de la Tecnología”, un concepto fundamental para evaluar el impacto de una tecnología. El dilema de Collingridge es imprescindible entenderlo a la hora de regular la tecnología y la innovación. Lo que plantea es un problema de tiempos. Por un lado, si regulamos una tecnología demasiado pronto, cuando aún está en desarrollo y no ha alcanzado una adopción masiva, no podemos prever todas las consecuencias negativas de dicha regulación. Esto podría dar lugar a una regulación ineficaz o incluso innecesaria. Por otro lado, si tardamos demasiado tiempo para regularla, entonces la tecnología ya estará muy extendida y será difícil de modificar sin costos significativos, conflictos políticos o resistencia de la industria.

Este dilema cobra especialmente sentido hoy en día con propuestas de regulación como la Ley de Inteligencia Artificial de la UE, cuyo objetivo es equilibrar los supuestos desafíos de los sistemas de IA según su nivel de riesgo: inaceptable, alto, limitado y mínimo. Aquí la UE busca establecer unas salvaguardas para un uso fiable y responsable de la IA con el objetivo de garantizar la transparencia, pero existen tres grandes escollos insalvables a esta misión aparentemente loable: uno, las empresas líderes en este tipo de tecnologías son privadas y extranjeras; dos, la ventaja estratégica de Estados Unidos y China pueden establecer regulaciones locales para que sus empresas avancen sin restricciones en detrimento de las europeas, como demuestra el apoyo gubernamental chino a Deepseek y por último, el poder de las Big Tech reside en ofertar distintos servicios por regiones relegando a Europa a una posición de desventaja.

Podemos afirmar sin equivocarnos que estas tecnologías son muy incipientes, pero su gran penetración las hace más difíciles de limitar por vía regulatoria. Incluso pensar que la vía ganadora es un desarrollo tecnológico propio, como la herramienta recientemente lanzada por el Gobierno de España, ALIA. Esta herramienta lingüística de código abierto representa un esfuerzo por desarrollar IA en español y promover la soberanía digital, pero su relevancia práctica y capacidades están muy por detrás de las de otras soluciones del mercado.

En este escenario, donde además la geopolítica ha ganado un gran peso, las empresas y profesionales de la UE parten con una desventaja clara, y es importante definir cuanto antes una estrategia ganadora para fomentar el uso de las tecnologías más relevantes, consolidadas y funcionales. Sólo así podrán existir innovaciones originadas en Europa con las que competir con otras regiones.

Publicado originalmente en Linkedin en abril del 2025.

I’ve been following this initiative closely since its inception, documenting the milestones, debating the skeptics, and analyzing the data. Now feels like the right moment to step back and examine what this experiment tells us about the future of urban mobility.

The Numbers That Matter

Let’s start with what the data actually shows after 12 months:

Traffic reduction: 27 million fewer vehicles entered the congestion relief zone—an 11% reduction. That’s roughly 73,000 fewer cars per day on Manhattan’s streets below 60th Street.

Revenue generation: Over $550 million in net revenue, enabling the MTA to advance $15 billion in critical transit infrastructure projects, including the Second Avenue Subway Phase 2 and accessibility upgrades at nine stations.

Speed improvements: Traffic moves 23% faster on crossings entering and exiting the zone during rush hour. The Holland Tunnel alone saw rush hour delays drop by 65%. Lincoln Tunnel MTA express buses travel nearly 24% faster than in 2024.

Transit ridership: Subway trips entering the zone increased 9%. Express bus trips up 7.8%. Local bus trips up 8.4%. All MTA modes hit post-pandemic record highs.

Safety: Crashes within the zone declined 7%. Traffic injuries down 8%. Traffic fatalities down 40% through July 2025.

Air quality: A peer-reviewed Cornell University study published in npj Clean Air found that daily maximum PM2.5 concentrations declined by 3.05 μg/m³ within the Congestion Relief Zone during the first six months—a 22% reduction from projected levels without the policy.

Economic activity: Broadway had its best season in history with $1.9 billion in ticket sales. Sales tax receipts in NYC were up 6.3% through November—three times higher than neighboring Westchester County.

Governor Hochul put it simply: “We changed how people in this great city and the region live, how they breathe, how they act.”

The Science Behind the Air Quality Claims

One of the most debated aspects of congestion pricing has been its environmental impact. Early official reports were cautious, stating air quality improvements were “inconclusive.” But rigorous academic research now tells a different story.

The Cornell study, led by researchers from the Systems Engineering Program and School of Civil and Environmental Engineering, analyzed daily PM2.5 data from 42 air quality monitors across the New York metropolitan area over 518 days.

A quick note on what PM2.5 means: these are particulate matter particles 2.5 micrometers or smaller—about 30 times thinner than a human hair. Their tiny size is precisely what makes them dangerous: they penetrate deep into lungs and can enter the bloodstream, contributing to respiratory disease, heart disease, and premature death. Vehicle exhaust, especially from diesel trucks, is a major source.

The researchers used predictive models incorporating meteorology, baseline pollution levels, and neighborhood demographics to estimate what air quality would have been without the policy.

The findings are significant:

• Within the CRZ: 22% reduction in daily maximum PM2.5 (3.05 μg/m³ decline)

• Across all five NYC boroughs: 1.07 μg/m³ reduction

• Broader metropolitan area (CBSA): 0.70 μg/m³ reduction

Perhaps most importantly, the effects grew over time. Treatment effects in the CRZ rose from 0.8 μg/m³ in Week 1 to 4.9 μg/m³ by Week 20, indicating accumulating behavioral adaptation as commuters, trucking companies, and delivery services adjusted their patterns.

This magnitude exceeds prior evidence from Stockholm (5–15% reduction) and London’s Ultra Low Emission Zone (about 7% decline). The researchers attribute the comparatively larger effects to New York’s intensity of travel demand and the modal flexibility provided by its extensive transit system.

What Commercial Fleet Data Adds to the Picture

In July 2025, Altitude by Geotab released a complementary analysis focused specifically on commercial vehicles, a critical piece often overlooked in passenger-focused reports.

Using data from over 16,000 vehicle trips comparing January–May 2024 to the same period in 2025, they found:

• 44% of Lower Manhattan roadways saw faster commercial vehicle travel times between 8 AM and 6 PM

• Noticeable decrease in large trucks (medium- and heavy-duty vehicles) entering the zone

• Corresponding increase in multi-purpose vehicles (vans, SUVs, pickups) for last-mile delivery

The Cornell study corroborates this, showing heavy-duty truck entries fell by approximately 18% and that this shift was closely associated with air quality improvements, given trucks’ disproportionate contribution to urban pollution.

One nuance worth noting: despite improved traffic flow, Geotab’s analysis found no substantial change in idling behavior. This suggests room for additional interventions—like coordinated delivery scheduling or incentives for zero-emission freight vehicles—to amplify environmental benefits.

What the Data Confirmed and What Questions Remain

Before congestion pricing launched, opposition was fierce. Politicians predicted economic devastation. Drivers complained about the $9 toll. Staten Island residents worried about spillover effects.

Most concerns proved unfounded. The “it will kill business” argument collapsed under Broadway’s record $1.9 billion season and a 6.3% jump in sales tax revenue. The “it’s regressive” claim ignored that most working New Yorkers take transit, not cars—and improved bus speeds benefit precisely those populations.

At the six-month mark, traffic delays were down 25% in the zone and 9% across the entire metropolitan region, including parts of New Jersey and the Bronx. Critics who predicted gridlock would simply shift elsewhere were wrong. The Cornell study confirmed this: no evidence of emissions displacement, just net reductions across the broader region. As the researchers put it, congestion pricing works as “a system-wide behavioral intervention rather than a geographically bounded one.”

But legitimate questions remain. The Regional Plan Association found illegal parking complaints jumped 17% just outside the toll zone—a reminder that policy interventions require systemic thinking. And a separate NBER study reached different conclusions about air quality, finding “little-to-no effect”—a sign that the science will continue to develop.

These wins are real. But they’re not the whole story.

Congestion Pricing Isn’t Enough

Back in October 2025, I argued that congestion pricing alone is not a sustainable mobility strategy. The MTA was selling $230 million in debt amid ongoing legal battles. The program had raised significant revenue, but relying on a single funding stream introduces real risk.

One year in, that concern hasn’t disappeared—it’s intensified.

Here’s what NYC still needs:

Resilient, diversified funding: Congestion pricing is a great tool. But legal challenges and political shifts can stall entire transit upgrades. New York defended the policy in court and won, but other cities watching this experiment should note: one lawsuit could derail years of planning.

Public trust and communication: Many residents still don’t fully understand how these funds are used. The MTA responded by announcing signage that says “This was paid for by congestion pricing”—a smart transparency move. But the broader communication strategy needs work. When people see tangible improvements linked directly to the toll, support solidifies.

Equity-centered design: Where’s the conversation about ensuring affordability for low-income commuters? What about improving alternatives in transit deserts? Congestion pricing benefits neighborhoods with good transit coverage. But for communities in eastern Queens or the South Bronx with limited options, the calculus is different. Access-A-Ride ridership increased 21%—a positive sign—but equity must remain central.

Mode integration and true MaaS: NYC still lacks a seamless multimodal experience. If congestion pricing changes driving habits, the alternatives need to work together—across subway, rail, bikes, buses, ferries, and more. One app. One payment. One user experience. That’s the next frontier.

The Bigger Picture: Lessons for Other Cities

New York’s experiment offers a template, but not a blueprint. Every city’s mobility ecosystem is different.

What works in Manhattan—with its dense transit network, high pedestrian traffic, and cultural acceptance of not owning a car—won’t translate directly to Los Angeles, Houston, or Phoenix. But the principles hold:

Pricing works. When you make driving into congested areas more expensive, fewer people drive. This isn’t controversial economics—it’s supply and demand.

Revenue enables investment. Congestion pricing isn’t just about reducing cars; it’s about funding the alternatives. The $15 billion in transit improvements that congestion pricing unlocked will serve New Yorkers for decades.

Data drives refinement. Real-time monitoring allows policymakers to adjust. When Staten Island expressed concerns about spillover effects, the MTA tracked traffic volume there. When critics claimed air quality wasn’t improving, Cornell researchers measured actual particulate matter across 42 monitoring stations. Evidence-based policy beats ideology.

Effects compound over time. The Cornell study’s most exciting finding may be the temporal dynamics: air quality improvements within the CRZ intensified nearly fivefold between Week 1 and Week 20. This suggests that early post-implementation evaluations may understate long-term gains, as behavioral adjustments—trip consolidation, increased transit use, peak-hour avoidance—accumulate gradually.

Political will matters. Governor Hochul paused the program in June 2024, then restarted it at a lower toll rate in January 2025. She faced lawsuits from the Trump administration. She weathered criticism from all sides. One year later, she’s celebrating at press conferences with dramatically reduced opposition. Leadership means making difficult choices and sticking with them.

What Comes Next

The question now isn’t whether congestion pricing works—the data settled that. The question is what we build on top of it.

The Cornell researchers identified three priorities that align with what I’ve been arguing:

Reinvestment is critical. Allocating toll revenues to support outer-borough transit, subsidized fares, and active-mobility infrastructure extends benefits beyond the CRZ and ensures citywide accessibility.

Freight management deserves explicit integration. The data confirms that heavy-duty truck activity remains a dominant driver of localized air-quality variation. Differentiated tolls, incentives for zero-emission freight vehicles, and coordinated delivery scheduling could amplify environmental benefits.

Adaptive management is essential. As treatment effects grew over time, continuous air-quality monitoring, dynamic rate adjustments, and iterative policy design will be vital to sustaining gains and preventing rebound effects.

For other cities watching, the lesson is clear: bold policy delivers results, but only when paired with systemic thinking. Congestion pricing is a critical piece of the puzzle but not the whole picture.

I’ve been building digital mobility ecosystems for nearly a decade now. What excites me about New York’s experiment is that it proves cities can make transformational changes when they commit to data-driven, outcome-focused policy. The peer-reviewed research is now catching up to what real-time data showed early on: this works.

The work isn’t done—equity concerns need addressing, freight integration needs refinement, and multimodal experiences need to become truly seamless. But the direction is right.

What’s your city doing to reduce congestion and improve transit? I’d love to hear about other initiatives that are moving the needle.

References:

My Previous LinkedIn Posts on This Topic:

• Campoamor, G. (December 2025). “NYC Congestion Pricing Air Quality Study.” https://www.linkedin.com/posts/guillermocampoamor_urbanmobility-congestionpricing-mta-activity-7393671592846802944-tfjC

• Campoamor, G. (October 2025). “Congestion Pricing Isn’t Enough: What’s Missing in NYC’s Mobility Strategy.” https://www.linkedin.com/posts/guillermocampoamor_congestionpricing-urbanmobility-sustainabletransport-activity-7371143068873408512-wpS7

• Campoamor, G. (July 2025). “Six Months In, Governor Hochul Highlights Success of Congestion Pricing.” https://www.linkedin.com/posts/guillermocampoamor_six-months-in-governor-hochul-highlights-activity-7355870077293842432-auPI

• Campoamor, G. (May 2025). “NYC Congestion Pricing: Insights from Geotab’s Latest Analysis.” https://www.linkedin.com/posts/guillermocampoamor_urbanmobility-mobilitytech-smartcities-activity-7341042696205025281-R5ku

• Campoamor, G. (January 2025). “Congestion Pricing Launches in NYC.” https://www.linkedin.com/posts/guillermocampoamor_urbanmobility-smartcities-transportation-activity-7289628816166039552-IMTt

Academic Research:

• Fraser, T., Park, Y.G., Lu, D., Tayarani, M., Deng, H., & Gao, H.O. (2025). “A first look into congestion pricing in the United States: PM2.5 impacts after six months of New York City cordon pricing.” npj Clean Air. https://doi.org/10.1038/s44407-025-00037-2

Government & Official Sources:

• New York State Governor’s Office. (January 5, 2026). “Less Traffic, Better Transit: On Its First Anniversary, Governor Hochul Celebrates Transformational Success of Congestion Pricing.” https://www.governor.ny.gov/news/less-traffic-better-transit-its-first-anniversary-governor-hochul-celebrates-transformational

• New York State Governor’s Office. (July 5, 2025). “Six Months In, Governor Hochul Highlights Success of Congestion Pricing: Traffic Is Down, Business Is Up.” https://www.governor.ny.gov/news/six-months-governor-hochul-highlights-success-congestion-pricing-traffic-down-business-and

• MTA. (January 5, 2026). “ICYMI: Less Traffic, Better Transit: On Its First Anniversary, Governor Hochul Celebrates Transformational Success of Congestion Pricing.” https://www.mta.info/press-release/icymi-less-traffic-better-transit-its-first-anniversary-governor-hochul-celebrates

Industry Analysis:

• Altitude by Geotab. (July 25, 2025). “Congestion pricing: Faster traffic and fewer trucks, but no change in idling?” https://altitude.geotab.com/news-and-releases/nyc-congestion-pricing-analysis/

• Regional Plan Association. (June 2025). “Congestion Pricing: Faster All Around.” https://rpa.org/news/lab/congestion-pricing-getting-around-faster-all-around

News Coverage:

• THE CITY. (January 5, 2026). “One Year In, Congestion Toll Yields Gains for Manhattan and MTA.” https://www.thecity.nyc/2026/01/05/congestion-pricing-hochul-mamdani-mta/

• Streetsblog NYC. (January 6, 2026). “How Kathy Hochul Learned to Stop Worrying and Love Congestion Pricing.” https://nyc.streetsblog.org/2026/01/06/how-kathy-hochul-learned-to-stop-worrying-and-love-congestion-pricing

• amNewYork. (January 5, 2026). “Congestion Pricing: Hochul celebrates Manhattan toll program’s ‘extraordinary’ results on its one-year anniversary.” https://www.amny.com/news/hochul-celebrates-congestion-pricing-anniversary/

• amNewYork. (January 5, 2026). “Congestion pricing: 27 million fewer vehicles entered Manhattan’s Central Business District in first year.” https://www.amny.com/news/congestion-pricing-27-million-fewer-vehicles/

• Smart Cities Dive. (July 7, 2025). “New York congestion pricing at 6 months: the results.” https://www.smartcitiesdive.com/news/new-york-congestion-pricing-6-month-results/752256/

• Bloomberg. (October 20, 2025). “MTA to Sell $230 Million of Debt Amid Tolling Plan Legal Fight.” https://www.bloomberg.com/news/articles/2025-10-20/mta-to-sell-230-million-of-debt-amid-tolling-plan-legal-fight

Advocacy & Policy Organizations:

• New York League of Conservation Voters. (January 2026). “One Year In, Congestion Pricing is Winning.” https://www.nylcv.org/news/one-year-in-congestion-pricing-is-winning/

• New York Lawyers for the Public Interest. (January 2026). “NYLPI Celebrates One Year of Congestion Pricing.” https://www.nylpi.org/nylpi-celebrates-one-year-of-congestion-pricing-with-governor-hochul-mayor-zohran-mamdani-and-advocates-at-rally/

Rather than merely assessing physical infrastructure, transport adequacy examines the full mobility experience. It considers factors such as ease of reaching destinations, travel conditions, time and cost efficiency, and the broader impact on individuals’ lives. This paradigm shift provides a more nuanced understanding of transportation effectiveness.

Why Transport Adequacy Matters

Mobility is not a one-size-fits-all experience. Two individuals living on the same street may have vastly different perceptions of transportation adequacy due to physical limitations, time constraints, or safety concerns. A nearby bus stop might offer seamless mobility for one person but pose accessibility challenges for another. More than just movement, transportation affects overall quality of life. Key considerations include:

• Time efficiency – Does commuting allow sufficient time for personal and professional commitments?

• Financial impact – Are transportation costs manageable?

• Safety and comfort – Does travel feel secure and convenient?

• Social inclusion – Does mobility enable access to essential services and opportunities?

The Key Components of Transport Adequacy

To understand transport adequacy, researchers identified several crucial elements that shape mobility experiences:

1. Subjectivity & Individual Variation – Accessibility depends on personal circumstances.

2. Travel Conditions – Safety, convenience, and comfort influence the experience.

3. Impact on Well-Being – Mobility should not negatively affect finances, health, or social life.

4. Access to Opportunities – Reliable transport fosters employment, healthcare, and education access.

5. Skills & Capabilities – Digital literacy and navigational ability play a role in transport use.

6. Emotional & Psychological Factors – Perceived risks and stress levels impact travel choices.

7. Daily Life Consequences – Mobility influences overall life satisfaction and community engagement.

8. Urban Context – Location-specific factors (e.g., bike lanes, transit connectivity) shape mobility experiences.

Measuring Transport Adequacy: The Transport Adequacy Scale (TAS)

Recognizing the challenge of quantifying subjective mobility experiences, researchers developed the Transport Adequacy Scale (TAS)—a structured tool for evaluating transportation adequacy.

The TAS comprises eight key indicators, with respondents rating statements such as:

- “There is always a transport option available to me at the times I need it.”

- “I am able to live my life as I want to because of my transportation options.”

Rated on a five-point scale, the TAS provides insights into transportation gaps, particularly for vulnerable populations. Initial testing in Rotterdam and Utrecht demonstrated its reliability in capturing diverse mobility experiences.

The Future of Transport Planning with TAS

By incorporating transport adequacy assessments alongside traditional metrics, cities can adopt a more holistic approach to transportation planning. This enables:

• Targeted improvements – Identifying and addressing disparities in transport access.

• Better resource allocation – Ensuring infrastructure investments enhance actual mobility experiences.

• Impact measurement – Evaluating how transportation changes affect diverse community segments.

Rather than focusing solely on infrastructure expansion, transport adequacy ensures that mobility systems genuinely enhance people’s lives. By balancing physical accessibility with human experiences, urban planners can create inclusive, efficient, and sustainable transportation networks.

A special thanks to the researchers from Utrecht University for their dedication and insights into this critical topic. Their work is shaping the future of mobility and making transportation more inclusive for all.

Originally published on LinkedIn on February 2025