Why the route still belongs to a person
Courier work is mostly movement. Someone drives or walks a route, carries documents and parcels between offices, labs and homes, finds the right door, and hands the item over. Software can plan that route and track the package. It cannot climb the stairs, deal with the locked side entrance, or decide that the lab specimen needs to go back because nobody signed for it.
The physical share is the heart of it. About 65% of the tasks in this job are physical, and the robotics tier that matches them is mobile robots: machines that move through open, changing spaces rather than a fixed cell. That is the hardest tier to deploy. Sidewalk robots and vans with autonomy stacks exist, but they work best on tidy, repeatable routes, not on a day that mixes a hospital basement, a construction site office and a third-floor walk-up.
The other half of the answer is paperwork that already moved. Recording what was received and delivered, sorting items by destination, and sequencing stops are tasks software handles well, and dispatch systems have been doing them for years. That is where the erosion happens first: fewer clerical hours around the route, not fewer routes.
What AI does, what it helps with, and what it leaves alone
Start with the tasks AI can take on its own. Route sequencing and delivery records sit here: picking the order of stops, logging times, and updating a status that a customer can see. That group covers 0% of task time in this job. It is the back office of a delivery, not the delivery.
Next, the tasks where AI assists a working courier. Sorting items for a run, flagging an address that looks wrong, and reading a manifest are faster with software in the loop, but a person still makes the call. Our split puts 23% of task time in that assisted group, and the share behind it is the one measured by our coverage score, which stands at 13 out of 100.
Then the rest. Driving the vehicle, loading and unloading, carrying items into a building, obtaining signatures or payment, and handling the exception when the recipient is not there all stay with people, and that group holds 77% of task time. Those tasks are why the headline Still needs a human score reads 81 out of 100 (higher is safer).
What the evidence actually shows
There is no direct test of AI against couriers on their own tasks yet. Our evidence grade for this job is D, and a D grade means not measured, so we publish no parity number against a working courier. The quality parity method explains why a grade is withheld rather than guessed.
What would settle it is specific: a published field trial of autonomous delivery over a mixed urban route, with completion rates, failed-delivery rates, handover time and cost per drop set beside the same route run by a courier. Pilot announcements are not that. Claims from delivery-company executives about robots taking over are not that either. Until somebody publishes the comparison, the honest answer is that the hard part has not been measured.
The market numbers around the job are measured. The Bureau of Labor Statistics counts about 68,640 US couriers and messengers and a median wage near $39,200 a year, with employment projected to grow roughly 8% from 2025 to 2035 (BLS, 2025). Growth plus unmeasured automation is a normal combination in delivery work: parcel volume rises while the hours per parcel fall.
When this could shift
Most likely after 2036 (8 in 10 of our scenarios). The replacement-year method sets out what that window does and does not cover.
Two things could pull it earlier. The first is cost: the tooling side of this job is cheap compared with a salaried courier, so even partial autonomy on simple routes pays back fast. The second is route shape. Campuses, hospitals, business parks and dense apartment blocks are repeatable environments where mobile robots already do laps, and a pilot that works there can scale quickly.
Two things hold it back. Handover is one: signatures, ID checks, chain-of-custody for medical and legal items, and the judgment call when a door is locked. Infrastructure is the other: sidewalks, stairs, elevators, gate codes and weather all break machines that work fine in a demo. Add local rules on sidewalk robots and drones, which vary city by city and move slowly.
Good to know: the parts of courier work most exposed today are the desk tasks around the route, which is also where entry-level hours used to sit.
How to stay needed
Lean into the tasks machines keep failing at. Take on time-critical and chain-of-custody runs, such as lab specimens, legal filings and pharmacy transfers, where a documented handover matters. Make yourself the person who solves failed deliveries rather than logging them. And learn the buildings on your patch: access, docks, desks, who signs.
Two skills travel well. One is route and dispatch software: knowing how to work with the system, correct it and read its exceptions. The other is a commercial license or specialist cargo handling, which widens what you can carry and what you get paid for.
If you want to see where the work sits next to it, three close jobs are worth a look: Postal Service Mail Carriers, Shipping, Receiving, and Inventory Clerks and Dispatchers. You can also read the wider picture for transportation and warehousing, browse the rest of the other transportation workers family, or set two jobs beside each other on the compare page.
Our guide to robots and physical jobs covers what mobile machines can and cannot do outside a warehouse, and the full scoring approach is set out in the methodology.