Why the doorstep still decides this job
Light truck drivers run vans and small trucks on local routes. The driving is only part of it. A typical day means loading the vehicle, sequencing stops, carrying packages to a door, getting a signature or a receipt, and sorting out the stop that does not go to plan. Software can handle a lane and a traffic light. It cannot find the service entrance behind a construction fence.
That is the honest reason the work holds. Automated driving aims at the highway and the mapped street. Delivery work ends on private property: driveways, loading docks, apartment lobbies, stairwells. Those spaces are not mapped, not standard, and change by the day. A damaged carton, a refused order, a customer who needs the pallet moved six feet, a dog in the yard — each one is a small judgment call made in seconds.
The market backdrop is steady rather than shrinking. The Bureau of Labor Statistics counted about 983,300 light truck drivers in the United States, with median pay near $44,860 a year, and projects employment growth of roughly 6.3% between 2025 and 2035 (BLS, 2025). Will delivery drivers be replaced by AI across that window? The task mix on this page suggests pressure on parts of the day, not on the job as a whole.
What AI does, what it assists, and what stays with the driver
The tasks our data puts in the “AI does it” group come to 0% of task time. These are the desk-like pieces of the route: building the stop order for the day, and keeping the delivery records and logs that used to be written out by hand. Routing software has done this at scale for years, and it does it well.
Assisted tasks account for 23%. Turn-by-turn navigation with live traffic is the obvious one, and the driver still decides where the van actually fits. Reporting delays and mechanical problems is another: prompts, checklists and camera systems flag the issue, and a person confirms what is really wrong with the load or the vehicle.
The rest, 77% of task time, sits with people. Loading and unloading cargo by hand is the clearest example. So is the customer side of a stop: collecting signatures or payment, checking that an order matches the manifest, and handling the complaint on the spot. Our coverage score, which measures the share of task time AI can handle today, comes out at 14 out of 100; you can read how that is built on the coverage method page.
How strong is the evidence here
Our quality-parity grade for light truck drivers is D. That grade means no study has yet tested an automated system against a working driver on this job’s real tasks, so we publish no parity number for it. Pilots of driverless vans and sidewalk robots exist, but a pilot is not a measured comparison.
What would settle it is specific: a trial where an automated delivery system runs full routes end to end, door included, with completion rates, failed stops, damage and human hand-offs all counted against drivers on the same routes. Until a result like that is published, the number would be guesswork. Our rule on that is set out in the quality-parity method.
Good to know: a grade that says “not measured” is not the same as a grade that says “AI is close” — it means the test has not been run.
When the picture could change
Most likely after 2044 (8 in 10 of our scenarios). The replacement-year method explains how that window is produced and what it does and does not claim.
Two things could pull it earlier. First, mobile robots are the robotics tier that matters for this job, and the hardware for them is getting cheaper and more capable each year. Second, delivery can be redesigned around machines: lockers, curbside hand-offs and secure drop boxes remove the hardest part of the stop, which is the door itself.
Two things push the other way. A large share of these tasks are physical and happen off the public road network, where automated driving has the least support. And the cost picture is lopsided: an automated van is a capital purchase plus service, while a driver is hired per hour, which makes the swap hard to justify on low-density routes with mixed freight. Rules on driverless operation and liability add another delay.
How to stay needed on the route
Lean into the parts of the day that machines keep handing back. Handling and checking cargo well — loading for the stop order, spotting damage before it leaves the dock — is one. Customer-facing stops, where you confirm what was ordered and fix the problem on the doorstep, is another. So is dealing with the exceptions: the blocked dock, the wrong address, the refused delivery.
Two skills travel well from here. One is working with the routing and telematics tools rather than around them, including knowing when the system’s plan is wrong for a street you know. The other is documentation and compliance: clean records, inspections and chain-of-custody detail, which carry weight in freight, medical couriers and regulated loads.
If you are weighing a move, nearby work scores differently. Compare this page with heavy and tractor-trailer truck drivers, driver/sales workers and taxi drivers, or put any two side by side on the job comparison tool. The wider picture sits on the motor vehicle operators family page and in our trucking sector overview.
For the headline figure, this job scores 80 out of 100 (higher is safer) on Still needs a human; the full scoring approach is on the methodology page. If the hardware side interests you, our guide to humanoid robots and physical jobs covers what machines can and cannot lift today.