Bobcat announced on 10 November, from a West Fargo, North Dakota dateline, that its AI-powered Bobcat Service Assistant is in daily use across its North American dealer network. The release, issued by Doosan Bobcat, says the tool “is now being used by about 600 technicians each day across Bobcat’s North American dealer network.”
Announcements in this sector usually describe a product that is about to be used — this publication reads a great many of them. This one describes one that already is, and attaches numbers to it, which makes it a different kind of document and a more useful one.
What Was Announced
Bobcat Service Assistant is a conversational interface over the company’s own service corpus. The release describes it as bringing together relevant service information, diagnostic guidance and knowledge from Bobcat’s archive of historical service cases, so a technician can ask a question instead of searching a manual. It is pointed at dealer service departments rather than at machine owners or operators.
It did not arrive from nowhere. The release says it was “First introduced as the Service.AI prototype at CES 2026.” The thing that was a named prototype in January is, by this release, a deployed system in November.
The release also sets the limits itself. The assistant is “designed to complement technicians’ training and expertise,” and Bobcat’s established service procedures and safety requirements “remain the standard for all repairs.” That is a narrower claim than this publication is used to reading in agent announcements, and it is consistent with where the tool sits: between a technician and a document, not between a technician and a decision.
The Four Numbers
Bobcat gives four figures, and they measure four different things. Two are about use, and two are about what the use displaced.
About 600 technicians each day. Reach, across the North American dealer network. The release does not give the size of that network or a total technician headcount, so the figure stands on its own rather than as a share.
Nearly 1,000 conversations per day. Over the first 60 days following launch, the release says, technicians averaged nearly 1,000 conversations a day with the assistant. That is intensity rather than reach, and it describes a different and earlier window than the 600 figure — the first 60 days, rather than now.
About a 40% reduction in the use of traditional troubleshooting guides in My Machine, measured “compared to the pre-launch monthly average for the year.” This is a substitution figure: it says one existing resource is being opened markedly less. The release does not say what My Machine is, and this publication did not find it described on Bobcat’s public-facing pages, so a reader cannot tell how much of a dealer technician’s reference work that 40% covers.
More than a 25% reduction in troubleshooting cases submitted for additional support, measured “compared to the prior year.” This is a deflection figure: it says fewer cases were escalated for additional support at all.
All four are Bobcat’s figures, reported here as such. This publication has seen no independent audit of them. What is worth noting is that both displacement figures arrive with their comparison stated — a monthly pre-launch average in one case, the prior year in the other. Vendor percentages in this market frequently arrive without one.
The Timeline Is the Gap
The release does not say when the Service Assistant launched. Construction Briefing’s report on it, filed the same day by Andy Brown, repeats the figures and does not supply a date either. This publication searched for one in the wider coverage and did not find it. So the first-60-days window is real but unanchored in the public record: a reader cannot place it in the calendar, and therefore cannot tell how much of the year the 25% comparison covers.
That is a limit on what the numbers can be made to say, not a flaw in them. The deployment figure — 600 technicians a day — is the one that needs no timeline to be legible.
The Service Bay, Not the Cab
It is worth being precise about which part of the equipment business this is. The AI stories this publication has covered from the machine makers have mostly been about the machine: Caterpillar and FieldAI putting physical AI on inspection rather than on machine control, HD Hyundai and XiteSolution building a digital twin around autonomous excavation in Korea, Kajima running FieldAI-powered robots across more than ten Japanese sites.
Service Assistant is in a different building. It is aftermarket — the dealer’s workshop, the warranty claim, the diagnostic tree, the call to the factory when the tree runs out. That layer has a characteristic this publication keeps noticing: it is a layer where a language model has relatively little to invent and a great deal to retrieve. A fault code usually has a correct answer somewhere in a manual or a past case, and the hard part is finding it rather than reasoning about it.
The closest comparable story in the archive is Filmore, which put equipment-dealer demand data behind a Copilot agent in Microsoft Teams in September. That was the sales side of the same dealership. This is the service side, and unlike Filmore’s early access, Bobcat is reporting from after the rollout.
Why Published Deployment Numbers Matter Here
The adoption surveys this publication has covered keep landing in the same place. Bluebeam’s 2027 AEC Technology Outlook put 28% of US AEC firms running AI in production, barely moved from 27% a year earlier. ServiceTitan’s commercial trades survey found that 62% of commercial contractors had tried AI and 15% of those using it could show clear ROI. The gap in both is between trial and consequence.
A manufacturer publishing how many of its dealer technicians use a tool each day, and what the tool displaced, is an attempt to report on the consequence side. The figures are self-reported and the baselines are the company’s own choices. But they are specific, and they describe work that decreased rather than capability that exists — which is the thing the surveys say the industry cannot yet demonstrate.
The Dealer Account
The release carries one worked example, from White Star Machinery, an independent Bobcat dealer that sells and rents Bobcat equipment across Kansas and Oklahoma. Shop foreman Trevor Simons consulted the assistant after a field technician had already checked several possible causes of a fault; it surfaced another item to examine, which helped locate the issue. Simons estimated that finding the same information through traditional service resources could have added 30 to 60 minutes. The release notes, in its own words, that individual results will vary.
Simons’s framing of why that matters is the plainest line in the document: “Information is one of the best tools a technician can have.” Eric Nardini, service director for Bobcat of New Hampshire, Boston, South Shore and Cape Cod, puts the same point as a question of latency — “The speed of an answer or suggestion is huge.”
Rob Waletzko, Bobcat’s senior director of service, frames the product as helping dealer technicians “find answers faster and with confidence.”
What Is Not in the Release
The release names no model, vendor or architecture. It carries no accuracy figure, no error rate and no account of what happens when the assistant is wrong — which, for a tool whose output a technician may act on with a torque wrench, is the number a cautious buyer would want next. It gives no pricing, and no statement of whether dealers pay for access.
Nor does the release say anything about machine-side data. Bobcat’s own site describes Machine IQ as a telematics service whose paid Health & Security tier sends fault-code notifications with troubleshooting steps — which looks like an obvious feeder for a diagnostic assistant. But the release does not describe a connection between the two, and this publication is not asserting one.
What the release does establish is narrower and harder to argue with: a prototype shown in January is being used by about 600 technicians a day in November, and the company is willing to say in public which of its older resources that use replaced.
