, Autonomy is table stakes, and supply-side orchestration, which Uber is already winning, is becoming table stakes too. The next differentiator in robotaxi customer experience is demand-side. Namely, a system that anticipates each traveler’s intent (route risk, speed-versus-safety trade-offs, the right mode for the moment) and orchestrates the trip to satisfy it. I call this Context-Dependent Ride Management. It is the layer where the premium, monetizable, flagship customer experience will be built.
Introduction
I have argued across two recent pieces that the next phase of mobility will be won on customer experience, not vehicle technology. In The Flagship Experience, and again in “Robotaxis: Beyond Autonomy, It’s About Superior CX,” I made the case that success in complex service ecosystems like mobility depends on the intelligent orchestration of the entire customer journey. In “Robotaxi Value Chains: From Prediction to Reality,” I showed how value in the emerging autonomous mobility industry accrues to the orchestrator that sits above the fleet.
This piece advances that argument one step further. If value accrues to the orchestrator, then the orchestrator’s defining job, i.e., what separates a premium service from a commodity ride, is resolving what the individual traveler actually wants on a given trip. Technology is table stakes. Even orchestration of supply is becoming table stakes. The unclaimed ground is the orchestration of intent.
With robotaxi deployment expanding and ridership climbing, it is now clear that a clean cabin and a competent trip from Point A to Point B are only two components of a superior customer experience. Customer-centric mobility requires deeper thinking, and often high investment, in a layer that almost no one is building today.
Three Ride-Hailing Episodes
To understand the gap between true customer-centric mobility and what is available today, consider three personal ride-hailing experiences. Each exposes a different context dimension that today’s services cannot represent.
- Route criteria: the path the vehicle takes. I recently took a robotaxi from Century City to downtown Los Angeles on a weekday evening. I was alone in the vehicle. The route ran through the West Adams neighborhood, one of LA’s less safe areas. Stopped at traffic lights, I witnessed open-air drug use and became genuinely concerned for my safety. Because I was in a robotaxi, I could not direct it to take a safer route. As a visitor, I had no way to know the route carried that risk until I was already inside it.
- Priority criteria: what the trip is optimizing for. On a different trip, I took a human-driven TNC vehicle from Burbank airport (where my flight was cancelled) to LAX to catch the last flight to San Francisco. I asked the driver to get me there as fast as possible. We deliberately cut through some sketchy neighborhoods to shorten the trip. In this trip, my optimal criterion inverted: I knowingly traded safety margin for speed, because making the flight was the goal. No fixed objective function captures that. The right answer depends on what I am trying to accomplish, and it can vary from one trip to the next.
- Mode-and-moment criteria: whether a robotaxi is even the right choice. In San Francisco, I needed to get from the Moscone Center to the Marina District for a midday meeting on a hot day. My first choice was a robotaxi, but the pickup ETA was ten minutes. The resulting delay would have made me late, and waiting in the sun would have been uncomfortable. A TNC could reach me in two minutes. I chose the TNC. The robotaxi lost not on price or safety, but on its fit to the moment.
Today’s Robotaxi Customer Experience
The current robotaxi customer experience optimizes for one variable: shortest time to destination. It cannot represent the other goals a traveler may hold, or the contexts in which those goals must be achieved. A single-objective optimizer has no way to express “get me there fast, and I accept more route risk to do it,” or “I would rather pay more for a safer route even if it is longer,” or “if you can’t reach me in five minutes, hand me to another mode.”
This is not a user interface gap. It is an intelligence gap. Representing a traveler’s success criteria, predicting them before the traveler has to articulate them, and reasoning over a route to satisfy them are hard problems that a faster ETA calculation does not solve.
The Partnership Wave Solves Supply, Not Intent
Ride-hailing companies, i.e., TNCs such as Uber and robotaxi operators such as Waymo, are locked in an intense round of partnership formation. Uber has tied up with Waymo, Zoox, and WeRide. Waymo has partnered with Element Fleet Management, among several others. Pony.ai has partnered with ComfortDelGro.
The purpose of these partnership agreements is to improve unit economics, survive financially, offload risk, and accelerate entry into new markets. Partnering with Uber gives Waymo better fleet utilization and lowers the punishing customer acquisition cost of building a rider base. For Uber, the partnerships offer fast access to L4-ready fleets without spending billions on AV R&D.
Some of these partnerships also result in customer experience improvements. For example, hybrid fleets become reachable through a single app. Partnerships can ease the geographic fracture that today breaks the robotaxi customer lifecycle. Geofenced operators, such as Waymo, must scale city by city. The moment a traveler leaves the Operational Design Domain (ODD), the experience collapses. Uber’s upcoming Tokyo pilot relies on Wayve’s “mapless” embodied AI. Such mapless approaches promise expansion without ODD constraints.
But these partnerships also introduce a structural problem. In a partnered ride, like the Uber/Waymo rides offered in Atlanta, Uber controls the digital experience (reservations, payments, ETA) while Waymo controls the physical experience (driving style, cabin, teleoperations). If a robotaxi stalls, Waymo’s remote team must intervene while the passenger stares at the Uber app, wondering why the trip’s ETA changed. The value chain is split across two corporate entities, and the seams show.
I described this digital/physical bifurcation in the value-chains piece. With this piece, I identify a third control point that neither partner currently owns. This control point is the layer that captures the traveler’s intent and reasons over the trip to satisfy it. Whoever owns that layer owns the premium.
Supply-Side Versus Demand-Side Orchestration
It is tempting to assume the incumbents are already building this. Uber, recognizing that value accrues to the orchestrator, is investing heavily in the orchestration layer. Its Uber Autonomous Solutions suite includes a fleet-intelligence system that ingests vehicle telemetry through a supply-state machine and uses an orchestration layer to tee up actions and interventions. Its data-enriched mapping refines pickups, routing, and ETAs, avoiding specific intersections at rush hour and mitigating fleet-wide risk from weather and road closures.
This is real and impressive. But it orchestrates supply. It positions vehicles, smooths operations, and protects the fleet. A vehicle avoids an intersection to improve the operator’s efficiency, not because a rider set a risk tolerance that the route should honor. Uber is winning the supply-side orchestration.
Demand-side orchestration is a different product with a different owner. The orchestrator elicits and resolves what the individual rider wants on the upcoming trip. It may be a specific level of route risk, a particular speed-versus-safety tradeoff, a level of comfort, or modality choice. It then reasons through the available options for delivering it. My three episodes are demand-side problems. A better supply-state machine solves none of them. This is the ground that remains unclaimed, and it is where the premium customer experience will be built.
Context-Dependent Ride Management
I call the demand-side layer Context-Dependent Ride Management. It is a service that treats each trip as a small negotiation over the traveler’s success criteria, rather than as a fixed optimization of arrival time.
On the Century City trip, a service practicing Context-Dependent Ride Management would present the route’s risk level alongside the ETA and let me set the risk tolerance for which the route is optimized. Then, it should re-estimate the time to the destination accordingly. On the LAX dash, it would let me optimize for speed, with eyes open. In San Francisco, it would not simply show a ten-minute ETA. It would recognize the midday heat, cross-reference my calendar, calculate that waiting makes me late, and offer a human-driven TNC or a transit connection instead.
Understanding context also means matching the hardware to the mission. The platform should know whether to dispatch a hyper-efficient pod for a last-mile run to a transit station or a spacious SUV for a traveler heading to a business meeting. Without this layer of intent-driven orchestration, the fragmented nature of today’s partnerships leaves passengers managing the seams themselves.
Two key capabilities
Two of these capabilities are worth naming precisely, because they are harder than they look and because they are where I have spent my recent work.
The first is intent prediction. Anticipating a traveler’s criteria before they have to articulate them, and personalizing that prediction to the individual and the moment. This is the core of the experience. Context-Dependent Ride Management is, functionally, intent prediction applied to ride orchestration.
The second is explainable route reasoning. “Show the route’s risk level” sounds like a UI feature. It is actually a reasoning problem. Scoring a route segment by segment for safety, surfacing why a segment scored as it did, and doing so in a way a rider can trust requires neurosymbolic geospatial reasoning. In other words, combining structured knowledge of the urban environment with collision history, neighborhood context, time of day, and more. This is precisely the kind of explainable reasoning system my firm has been building, and it is what turns “shortest path” into “the path that matches what this rider asked for.”
The Monetization Layer
The industry has long equated “premium” with a luxury chassis. The deeper truth, the one I articulated in The Flagship Experience and implemented through the FEAT framework described in the book, is that premium means matching the offering to the traveler’s success criteria, not to the size of the vehicle. Context-Dependent Ride Management is the operational expression of that idea.
It is also a monetization layer. Intent resolution is a structured way to price-discriminate in the traveler’s favor. A risk-tolerance setting yields different routes, different arrival times, and therefore a defensible basis for different prices. A vehicle-to-mission match, pod versus SUV, calls for a tiered product. When a robotaxi is ten minutes out, the platform can offer a discount that acknowledges the longer time to get to the traveler rather than silently losing the rider to a TNC. Many travelers, particularly those riding alone, would gladly pay a premium for a service that takes their criteria seriously. The willingness to pay is the proof that this is a business, not a feature.
The MaaS Imperative and the Orchestration Layer
Context-Dependent Ride Management reaches its full form only inside multimodal Mobility-as-a-Service. The current wave of robotaxi partnerships solves for supply efficiency and risk mitigation. It does not solve for traveler intent, and it cannot reach across modalities.
True MaaS requires an orchestrator that integrates deeply into the traveler’s digital life. This means syncing calendars, drawing on real-time transit schedules, accounting for the physical transportation infrastructure, weather, and the many other signals that are available today, but remain largely unused. My San Francisco episode is the small version of this. A fully realized MaaS platform would not display a static ten-minute ETA. It would read the heat, cross-reference the meeting, conclude that waiting makes me late, and pivot seamlessly to a TNC or a transit leg. This is the Multimodal MaaS Value Chain I described in Transportation Transformation, in which robotaxis are not a standalone product but one integrated component of a digitally orchestrated network, and in which the city government itself evolves from regulator to transportation orchestrator.
Uber is building the nervous system for hybrid fleets, but on the supply side. By controlling the in-cabin interaction, fleet management, and edge-case support, it positions itself to capture high-margin supply orchestration. The demand-side intent layer, the one that knows what this traveler wants and reasons over the whole multimodal network to deliver it, is still open. That is the layer that will define the winning customer experience, and it is the layer worth building.
Conclusion
The race for the future of ground mobility has shifted. Autonomously moving a vehicle from Point A to Point B is table stakes. Orchestrating supply is becoming table stakes too, as Uber’s latest moves show. The enduring differentiator and the profitability driver is the demand-side layer. This is the system that understands the traveler’s goals and success criteria and orchestrates the trip across routes, priorities, and modalities.
Shortest time to destination is an important criterion. It is not the only one, and it is not always the most important one. The companies that ultimately win will be those that transcend rigid algorithms to deliver Context-Dependent Ride Management, anticipating passenger intent and seamlessly orchestrating a multimodal, flagship experience that adapts to the fluid reality of the physical world.
This post expands on the conversation with Ben Lorica on the same topic. You can watch the conversation on the monthly podcast I do with Ben
https://youtu.be/t43daDgTII8?si=L8KA1aMBrmLriEly