At a glance
- Service leaders are hiring into a shortage, with 52 percent of new diesel-technician demand unmet in 2026.
- Few treat the veteran’s knowledge as an asset to capture before it retires.
- The best evidence says AI helps newer workers most, but it comes from support desks; nobody has measured it in a service bay yet.
- Capture every fix on the work order, put it on the new hire’s tablet, and protect how they learn.
You can’t hire the veteran technician you need. The TechForce Foundation’s 2026 report puts demand for new diesel technicians at 26,692 a year, with 14,011 of those roles unfilled.
So the question for a VP of Service becomes how fast a new hire can work like an experienced one, and what that takes.
AI helped the newest workers most in the best study we have
The strongest evidence comes from a support desk. Economists Brynjolfsson, Li and Raymond studied 5,179 customer-support agents at one firm as an AI assistant rolled out.
On average, agents with the assistant resolved 14 percent more issues per hour. Novice and lower-skilled agents gained 34 percent.
The ramp-up result matters more to an operations leader. Agents with two months on the job and the assistant performed as well as agents with more than six months who didn’t have it. Attrition fell too.
The BCG field experiment with 758 consultants found the same shape. On tasks inside AI’s frontier, below-average performers improved 43 percent and above-average performers 17 percent. Across these studies, AI spreads what the best people know to everyone else, and shortens the time it takes to learn it.
Nobody has run this study in a service bay yet
There are no randomized studies yet of field technicians, plant operators, or maintenance crews using AI.
A support desk is a fair analogue for part of a technician’s day: diagnosing from symptoms, finding the right procedure, writing up the job. The rest of the job is not in the data: hands on a machine, a safety call in the field, a customer standing beside the truck.
So measure your own baseline before you start. Track how long a new technician takes to work alone, and how often their jobs need a second truck roll.
The knowledge has to be in the system before an agent can share it
An agent can only pass on what someone recorded. Today a newer technician facing an unfamiliar fault often calls around until a senior tech picks up. That fix lives in two heads and a phone log.
Recording the fault, the fix, and the part on every work order changes that. With the history in 
Dynamics 365 Field Service, a knowledge agent built in 
Copilot Studio can read it with the manuals. It puts the likely causes and the right part on the technician’s tablet before the truck leaves the yard.
We have built this pattern: a technician virtual assistant on Field Service for a national facility-maintenance services provider. The pattern works only as well as the history underneath it, which is why capture comes first. Our piece on first-time fix at the dispatch desk covers the intake side.
Speed without understanding is the hidden cost
AI can make a new hire faster without making them better. In a 2026 Anthropic randomized trial, 52 mostly junior software engineers learned a new programming library, with or without AI help.
On a quiz about what they had just built, the AI group averaged 50 percent, against 67 percent for those who coded by hand. The way people used AI mattered. Those who asked for explanations and followed up scored 65 percent or more; those who handed off the work scored under 40 percent.
The entry-level pipeline is under pressure too. Stanford’s Digital Economy Lab finds about a 19 percent employment gap for workers aged 22 to 25 in the most AI-exposed jobs, concentrated where AI automates rather than augments. The evidence is descriptive and contested: PwC finds AI-exposed entry-level job postings up 35 percent.
Design it as an apprenticeship. The agent explains why a part fits and what the history shows. New technicians still practice without it, and a veteran signs off their early jobs.
The veteran’s role changes, and the technician keeps the call
Capturing a veteran’s knowledge changes the veteran’s job. It shifts toward reviewing what goes into the knowledge base, coaching, and signing off new technicians’ work.
What stays human matters as much as what the agent drafts. The repair call, the safety call, and the warranty call stay with the technician. The agent brings the history; the person decides what to do with the machine in front of them.
Where this apprenticeship model doesn’t fit yet
If your work orders don’t record the fault, the fix, and the part today, an agent has nothing to read. Start with capture, and let the history build before you judge the answers.
Leave mixed reality out of the plan. Microsoft ends support for Dynamics 365 Guides and Remote Assist on December 31, 2026, so don’t build a knowledge program around them now.
Finally, where the work is safety-critical and regulated, the agent informs and the person signs. That line doesn’t move with better tools.
What service leaders should do this year
Measure how long new technicians take to work alone. Without a baseline, you won’t know whether the money you spend helps.
Record the fault, the fix, and the part on every work order. It’s the knowledge base your newest technician will need.
Put the history on the tablet, with the reason behind each answer. Understanding is what makes the second year better than the first.
Give your veterans the reviewer’s role. They decide what goes into the knowledge base and sign off the early jobs.
Keep hiring at entry level. Use AI to bring juniors up faster.
Deciding which tasks an agent should support starts with mapping the jagged frontier. Making it stick takes an operating model built around the people. More on our approach: the human side of AI.
How we know this
This point of view comes from Ludia’s Field Service delivery and its Helios mapping work, described without client names.
- Technician figures: TechForce Foundation, Technician Supply, Demand and Opportunity Report (May 2026).
- Support-agent study: Brynjolfsson, Li and Raymond, Quarterly Journal of Economics (2025). 5,179 customer-support agents at one firm; issues resolved per hour.
- Consultant experiment: Dell’Acqua, Mollick et al., “Navigating the Jagged Technological Frontier” (Sept 2023); published in Organization Science (2026). 758 BCG consultants; gains on tasks inside the frontier.
- Skill formation: Anthropic, “How AI assistance impacts the formation of coding skills” (2026). Randomized trial, 52 mostly junior engineers.
- Early-career employment: Stanford Digital Economy Lab, “Canaries in the Coal Mine” update (Aug 2026). ADP payroll data; descriptive, not causal. Counterpoint: PwC Global AI Jobs Barometer (June 2026).
- Mixed reality: Microsoft Learn, Dynamics 365 Guides and Remote Assist end of support.
