Who Manages Robot Fleet Operations Across Three Sites?

When A Robot Stalls At 2 A.M., Who Actually Owns It?

Right now the answer is probably four people and a group chat. The night lead at the site power-cycles the unit. The integrator's support line opens a ticket that closes in the morning. An automation engineer sees the alert on a dashboard nobody is watching. By 6 a.m. the shift ran short and no one can say whether it was the same fault as last Tuesday. That gap is the job.

A Robot Fleet Operations Manager owns the fleet as an operating asset rather than as a project with a go-live date. That means a readiness number per site that gets reported like a real metric, an allocation call when the third site's peak week collides with the first site's, maintenance and firmware windows scheduled against production rather than against the vendor's release calendar, and one incident log that treats a fault at site two as evidence about site three.

Serve Robotics scopes exactly this shape in its own posting: global allocation, readiness and regulatory alignment across markets ¹. Reflex Robotics opened a fleet operations role as its pilots with third-party logistics providers and US manufacturers turned into multi-site deployments ². Once robots stop being a pilot, somebody has to run them, and it cannot be the person who also has to build the next integration.

What Separates A Fleet Operations Manager From Your Automation Engineer

An automation engineer makes one cell work. A fleet operations manager decides which cells get units this quarter, and lives with the answer. The split is authority and time horizon, not skill level. The engineer's work ends when the workflow runs; the fleet manager's work starts there and never ends, because the fleet degrades, the sites change their mix, and the software ships.

Keep both roles. Do not fold one into the other, and do not report the fleet manager to the engineering lead who owns the rollout they are supposed to be able to delay.

The tells that separate a real candidate from a performed one are boring and reliable. Ask what their uptime definition excluded and why, and listen for whether planned downtime, operator error and blocked-path stalls counted. A candidate who has actually held a fleet number will argue about the denominator before they quote the number. Ask about a week they refused a deployment. If they cannot name one, they have never had the authority you are hiring for. Ask how many units one person can hold, and be suspicious of a clean answer: it depends on site count, fault rate, technician coverage and how much of triage is remote, and the honest candidates say so before giving you a range.

Where Do Robot Fleet Operations Managers Come From?

Nobody has ten years of robot-fleet experience, because the fleets are younger than that. The backgrounds that transfer are the ones where somebody already ran a population of machines against a service level. Airline and rail maintenance planning is the closest analogue: readiness rates, rotable spares, deferred defects, a maintenance window fought over with operations. Data center site operations is next, and it brings the incident discipline.

So does telecom field operations, where dispatch and mean-time-to-repair are the daily currency. The unexpected ones are worth naming. A rental fleet manager for construction equipment has run allocation across sites with damage disputes and utilization targets. A hospital biomedical equipment supervisor has run preventive maintenance schedules under a regulator's eye. A former Navy or Air Force maintenance chief has run readiness boards for a decade. Each of those people needs to learn what an autonomous mobile robot actually fails at, which is a few weeks of shadowing, not a career change.

What none of them can learn quickly is the habit of holding a number in front of people who want the machines used differently. If your pipeline is thin, widen it here before you widen it toward robotics graduates, whose training is in making one machine capable rather than in keeping forty available. The same reasoning applies when you staff an AI-heavy SRE function: the operational instinct is the scarce half.

Ask Them To Read Fleet Telemetry With An AI Assistant In Front Of You

The people who got good at this in the last two years did it with a model beside them, and the way they used it is diagnostic. Weak use looks like pasting a week of fault logs into a chat window and forwarding the summary.

Strong use looks narrower: asking the model to cluster fault codes and then checking two clusters by hand against the maintenance record, or having it draft the rollout comms for three site leads and rewriting the paragraph that promised a window they had not confirmed.

Put it in front of them. Give the candidate a messy export of stalls, faults and shift notes from a fleet that is not yours, an assistant, and forty minutes. Watch what they ask first, whether they check the confident claim the model makes about a root cause, and what they refuse to conclude from a sample this small. A candidate who catches the model attributing a battery pattern to a firmware version, and says so out loud, has shown you the exact judgment the job runs on at 2 a.m. This is the same evidence problem as hiring an analyst who works through models all day: the artifact tells you nothing, the session tells you everything.

Pay The Fleet Role Against Operations Management, Not Against Robotics Engineering

Benchmark carefully, because the public data for this title is thin and mixed. ZipRecruiter's aggregate for the broad "robotics operations manager" category, as of 2026, puts most US pay between $41,000 and $77,500 ³, and that range is almost certainly wrong for what is described here: it pools technician-adjacent supervision with multi-site ownership, and it lags the venture-funded postings that created the role. Treat it as a floor for the category rather than a target for the hire.

The defensible approach is internal. Find what you pay a plant operations manager or a multi-site logistics manager with the same headcount and the same authority over a production schedule, and start there, then add for the on-call load and the scarcity. Say plainly in the offer conversation which comparison you used. Candidates coming from aviation or data center operations will price themselves against their old function, and a manager who cannot tell them what the number is anchored to reads as a company that has not decided whether this role is senior. If you cannot source a number you believe, do not quote one; name the band and the comparison and let them tell you where they sit in it.

Close Them On Fleet Access And Authority, Not On The Robots

What kills this offer is almost never money. It is the candidate working out, in the third conversation, that they will be a scheduler for decisions someone else makes. Ask them to describe their first ninety days and listen for what they assume they can decide: pausing a rollout, moving units between sites, standing down a shift.

If your answer to any of those is that engineering or the site GM owns it, say so before the offer, because they will find out in month two and leave in month five. The things they care about are concrete. How many units, at how many sites, with what technician coverage on nights. Whether telemetry is theirs to query or has to be requested from the vendor. Whether they get a maintenance window with teeth.

On location: this is not a remote job for the first two quarters and pretending otherwise loses good people twice. The fleet manager should be on the floor across all three sites in rotation until they know each one's failure personality, then hold a hybrid pattern with a hard rule about being on site for a major rollout and for anything that hurt someone. Write the travel expectation into the posting as a percentage. A candidate who has run field operations will read a vague answer there as a company that has not counted the trips, which is the same doubt that a vague uptime definition creates.

See how it works >