Human drivers absorbed unpriced ground operations: parking, staging, fueling, cleaning, and local judgment. When the driver disappears, those tasks do not. They reappear as line items in an AV fleet's operating cost. XoomPark coordinates the off-depot ground-services layer those tasks land in.
A human driver did the driving. The driver also did everything around the driving: found a place to wait, parked between rides, fueled the car, kept it presentable, decided where it was safe to stage, and handled small problems before they became dispatch tickets. None of that work appeared on a balance sheet, because one person performed it for free as a condition of the job. Remove the driver and the driving gets automated. The ground operations do not. They get itemized.
What ground operations did human drivers absorb?
Drivers absorbed staging, waiting, parking between trips, fueling or charging, interior cleaning, vehicle presentation, local site judgment (where it is legal and safe to stop), and first-line exception handling. In a human fleet these are invisible because the driver does them off the clock, inside the fare. In an autonomous fleet there is no person to absorb them, so each task becomes a discrete operation a fleet must locate, schedule, pay for, and prove happened. This is the core economic insight behind XoomPark: the driver was a bundled ground-operations worker, and unbundling that role makes the bundle's cost visible.
| Bundled in the human driver | Becomes in an AV fleet |
|---|---|
| Knows where to wait between rides | Reserved staging capacity at a known site |
| Parks the car overnight at home | Contracted overnight storage with access rules |
| Fuels or charges on personal time | Scheduled charging-adjacent queueing slot |
| Wipes the seats, clears trash | Light-cleaning service with evidence of completion |
| Judges if a curb is safe and legal | Pre-qualified private-site access permission |
| Handles a stuck or stranded car | Recovery holding and exception workflow |
Why robotaxis make these costs visible
Robotaxis convert a labor cost into an operations cost, and operations costs do not disappear when labor does. The industry framing of autonomy as "removing the cost of the driver" is incomplete. It removes the wage, but it also removes the unpriced bundle of services the wage quietly covered. A human Uber or Lyft driver absorbs the cost of parking, the time spent staging near demand, the fuel stop, and the quick interior reset, all inside their effective hourly rate. The hidden costs of Uber drivers were never zero. They were paid by the driver and hidden from the platform.
When a fleet operator runs autonomous vehicles, those same tasks must be procured. The vehicle still needs somewhere to sit when it is not earning. The private car this displaces is famously parked about 95% of the time (Donald Shoup, The High Cost of Free Parking, citing the US DOT Nationwide Personal Transportation Survey; Streetsblog USA), and even a high-utilization robotaxi still spends a meaningful share of every day idle between fares. A car that is not on a trip is occupying space somewhere, and space in a dense city is not free. Driver labor replacement does not eliminate that line item. It relocates it from an invisible personal subsidy to a visible operating expense.
Plain-English definition
The driver ground-ops layer is the set of physical, between-trip operations a human driver used to perform as an unpriced part of driving: staging, parking, fueling, cleaning, site judgment, and small problem-handling. As fleets remove the driver, this layer has to be rebuilt explicitly as scheduled, paid, and audited services. XoomPark is the coordination layer for that work across distributed private sites: reservation, access, workflow, SLA, evidence, and audit. The property defines permission. The fleet validates capability. XoomPark coordinates everything in between.
Original analysis: decomposing the driver's invisible ground-ops bundle
| Driver task (unpriced) | AV-fleet operation | Where the cost reappears |
|---|---|---|
| Idle staging near demand | Reserved staging capacity | Per-hour space cost at a qualified site |
| Parking between/after shifts | Overnight storage | Contracted slot + access management |
| Refueling / recharging | Charging-adjacent queueing | Scheduled slot + dwell-time coordination |
| Interior cleaning and reset | Light cleaning | Per-event service labor + evidence capture |
| Curb and site judgment | Private-site access permission | Site qualification + access-rule enforcement |
| Handling a stuck/stranded car | Recovery holding | Exception workflow + holding capacity |
The methodology note that matters: each row was unpriced precisely because one person did it adjacent to the paid task, so no separate transaction ever recorded it. Charging is the clearest illustration. A human driver charges or fuels on personal time and absorbs the dwell. An AV cannot. The vehicle physically occupies a space for the full charging session, and that dwell is real, schedulable time that must be coordinated against finite slots. A typical DC fast charge from roughly 10–20% to 80% takes about 20–45 minutes (US DOT, EV charging speeds), so every charging vehicle holds a slot for a meaningful block of time. Multiply any one task family across a metro fleet and the "we just removed the wage" model breaks. The driver was not only labor. The driver was the ground-ops coordination layer, performed by a human, for free. XoomPark rebuilds that layer explicitly.
Why it matters for AV fleet operating costs
The economics of autonomy depend on what you assume about ground operations. If you model an AV fleet as "the human fleet minus wages," you overstate the margin, because you have deleted a worker without re-pricing the dozen tasks that worker performed. Ground operations after the driver disappears are not a rounding error. They are the difference between a clean unit-economics story and an honest one.
This matters most to three groups. Investors underwriting AV economics need to know whether a business plan has priced the ground-ops layer or assumed it away. Fleet operators scaling beyond a single depot need distributed capacity they can trust, schedule, and prove. Real-estate and parking owners sitting on underused private sites need a way to turn that space into AV-ready ground-service capacity without becoming an autonomy company themselves. The driver ground-ops layer is where these three meet.
How it works
XoomPark treats each between-trip task the driver used to absorb as a coordinated service event at a qualified private site. A fleet finds qualified capacity, reserves it, the vehicle checks in under defined access rules, the service happens, evidence is captured, and a billable, auditable session record closes the loop. The fleet keeps doing autonomy. The site keeps owning the property. XoomPark owns the coordination, records, and proof.
This is the permission-versus-capability distinction in practice. HD mapping tells an AV where it may be able to navigate. It does not tell a fleet whether a private lot will allow the vehicle in tonight, what the vehicle is allowed to do there, whether a charging-adjacent slot is open, or what evidence proves the cleaning actually happened. That is a coordination problem, not a mapping problem, and it is exactly the layer the driver used to cover from the front seat.
Who needs the driver ground-ops layer
Three groups have a direct stake in making this layer visible and financeable. Fleet operators(the Waymo, Zoox, Nuro, and Avis Budget Group ground teams of the world, plus fleet-ops partners like Moove and Transdev) need distributed capacity they can trust without owning it. Waymo reported 2,500 robotaxis on its 5th-generation system to NHTSA in late 2025, with roughly 800–1,000 in San Francisco, ~700 in Los Angeles, and ~500 in Phoenix (NHTSA filing data; trade-press tracking). At that scale, ground operations are not a rounding error. Real-estate and parking owners hold the capacity and want to monetize idle space without becoming AV experts. Investors need to know whether a business plan has priced the ground-ops layer or assumed it away.
What XoomPark does and does not do
| XoomPark does | XoomPark does not do |
|---|---|
| Discover and qualify private sites as AV-ready ground nodes | Own or operate chargers or an EV charging network |
| Hold reservation, access, and session records | Perform vehicle maintenance or repairs |
| Run check-in/check-out and exception workflow | Drive, dispatch, or operate the autonomous fleet |
| Track SLA and capture ground-service evidence | Build HD maps or certify AV safety |
| Produce billing and audit records | Replace fleet operators (Moove, Avis, Transdev, ABM, Waymo ops) |
| Coordinate across distributed private-site capacity | Run a consumer parking app or parking marketplace |
Not for you (yet)
If you run a single fixed depot and every vehicle returns to that one building for staging, charging, cleaning, and storage, you do not need a distributed node network yet. The driver ground-ops layer becomes a real cost center when you operate off-depot: when vehicles spend their idle hours away from base and you need trusted, scheduled, audited capacity spread across a city. If you are pre-deployment, mapping-only, or still modeling autonomy as "human fleet minus wages," this is a thesis to underwrite, not a service to buy today.
Frequently asked questions
What ground operations did human drivers absorb?
Drivers absorbed staging, parking between trips, fueling or charging, interior cleaning, vehicle presentation, local site judgment, and first-line exception handling. All of it was unpriced because one person performed it as a condition of driving. In an autonomous fleet there is no person to absorb it, so each task becomes a discrete operation that must be located, scheduled, paid for, and proven.
Are the hidden costs of Uber drivers really removed by automation?
No. Automation removes the wage, but it does not remove the work the wage quietly covered. Parking, staging time, fueling, and cleaning were paid by the driver and hidden from the platform. In an AV fleet those same costs reappear as visible operating expenses that a fleet must procure and coordinate.
Why are depots not enough for robotaxi ground operations?
Depots handle concentrated maintenance and charging, but vehicles spend idle hours far from base, near where demand is. Sending every vehicle back to one building for every reset wastes miles and time. Distributed private-site capacity, coordinated for reservation, access, workflow, and evidence, covers the between-trip ground operations a single depot cannot.
Does XoomPark replace fleet operators?
No. XoomPark does not drive, dispatch, maintain, or operate vehicles, and it does not replace operators such as Moove, Avis Budget Group, Transdev, ABM, or a fleet's own operations team. It coordinates the ground-services layer: reservation, access, workflow, SLA, evidence, and audit across distributed private sites.
Related reading
Read about AV ground services
The driver was the ground-ops layer. Now that layer has to be built on purpose. Read about AV ground services to see how XoomPark coordinates reservation, access, workflow, SLA, and evidence across distributed private sites.