Is autonomy the only way last mile delivery can ever make money?

I spent time recently at Curbivore in LA, one of the leading events for understanding how last-mile delivery is evolving in practice. It was a great opportunity to get hands-on with a wide range of solutions and to understand the broader ecosystem that supports last mile, including mapping, robotics and fleet operations.

What comes through is that the conversation has settled on one issue. Grocery and restaurant delivery has scaled, but the economics have not. That gap is the opportunity where the industry is focusing its efforts.

Labour is still the constraint

Labour remains the dominant cost and the main source of variability. The availability of people and demand peaks create inefficiencies. Operators can handle demand, but they often struggle to do so efficiently and consistently. This is the pressure point that sits behind much of what is being tested.

Different technologies are being used for different jobs

While there is no single solution emerging, the alignment between use case and technology is becoming clearer. Sidewalk robots are being used in dense, predictable environments. Companies such as Coco, Starship and Serve Robotics are focused on smaller baskets and repeat missions, where reliability matters.

There is also interesting work coming through from Robotis, which is developing robot workers that can operate across both indoor and outdoor environments. That includes delivery robots capable of navigating buildings, using elevators and interacting with entry points such as front doors. It extends last mile beyond the street and into the building itself. Drones are being positioned around suburban delivery, where distance and lower density make traditional models harder to run efficiently.

Platforms are becoming orchestrators

Human couriers, robots and drones are being treated as part of the same system. Platforms are responsible for deciding what gets used where. This makes orchestration a core capability and it is already being reflected in how platforms such as Uber and DoorDash are evolving their models. It depends on real-time decision-making and accurate mapping. Companies such as Mapbox are focused on practical challenges such as the final stretch to the door, where small inefficiencies add up quickly.

Reliability is replacing novelty

Once autonomous delivery systems are live, merchant feedback is broadly positive. Estimated delivery times become more predictable, handoffs improve, and congestion at the storefront reduces. From a consumer perspective, the novelty fades quickly. Reliability becomes the expectation.

Where this creates tension

The direction is clear, but the implications are not the same for everyone. For retailers and foodservice operators, autonomy has the potential to improve delivery economics over time, particularly in harder-to-serve suburban markets. That creates an opportunity to extend reach and build more consistent delivery propositions. But it also raises questions around how store networks and fulfilment models are structured to support different delivery missions. There are significant impacts for store and restaurant design.

For manufacturers, the shift is less straightforward. Platforms retain control of demand, pricing and visibility. Discovery, merchandising and media within these platforms become more important, particularly as more transactions move through them.

Both sides are dealing with a model that is still evolving. The next phase is less about the technology and more about where it gets applied first and who captures the value when it does. Autonomy is not solving the last mile on its own. It is making parts of the model work that did not work before.