The One Workflow Pilot: a one-page multi-shop AI pilot plan for auto repair groups

This is the plan from our video on AI for automotive service groups. Same four steps, same order:

1. The workflow. Pick one thing that happens every week, has one owner and gets reviewed by a person, like stale estimate follow-up.
2. The shops. Run it in one to three shops, and include a typical one, not only your best.
3. The rules, on one page. Who owns it, what data goes in, what you'll count, what happens when it's wrong. Nothing reaches a customer until a named person approves it.
4. Your own baseline. Count two to four weeks before you start, then run the pilot for the same length and compare.

Then scale only what held up. The rest of this page is the plan filled in, with a rules sheet you can copy.

Why groups should start smaller than they want to

A group owner usually asks which AI tool should go into every shop. The better question is smaller: which one workflow, in which two shops, with whose name on it.

Groups have a harder version of the problem because they often grow by buying shops, and acquired shops arrive with their own systems and habits. In collision repair alone, Focus Advisors reported that larger operators acquired more than 450 locations in 2024. Staffing adds pressure too. The 2026 Ratchet+Wrench industry survey, with more than 430 responses from owners and managers of independent shops, says technician shortages are shaping decisions. A pilot keeps the bet small while you learn what works across shops that don't run the same way.

Step 1. Pick the workflow

A good pilot workflow passes four tests:

Test	What it means	Stale estimate follow-up
Weekly	it happens in every shop, every week	yes, every open estimate
One owner	one named role is accountable	the service advisor
Reviewable	AI prepares, a person decides	the advisor approves each message
Countable	you can count it before and after	estimates past your cutoff, contacted, approved later

Other good first picks are delayed part tracking, review replies and the Monday owner report. Shopmonkey's overview of AI in shops lists uses such as chatbots, scheduling, inventory forecasting and diagnostics. Each fails differently. A wrong appointment time is annoying and a wrong diagnosis is expensive, so start where a mistake is cheap and easy to catch.

Step 2. Pick the shops

One to three shops is enough. Include:

• One shop where the manager wants it. You need someone who will use it daily and tell you what's broken.
• One typical shop. If the pilot only works where the manager is exceptional, you've learned about the manager, not the tool.

Before day one, confirm each pilot shop's system can supply the data. Tekmetric's post on multi-shop growth lists shared customer history, standard inspections and unified reporting among the foundations. If your shops define stages or "stale" differently, agree on one definition now. Otherwise the two shops' numbers can't be compared.

Step 3. Write the rules on one page

The National Institute of Standards and Technology publishes a voluntary AI Risk Management Framework built around four functions: Govern, Map, Measure and Manage. You don't need the full framework for a shop pilot. The four words make a good rules sheet.

Copy this and fill it in:

1. Owner (Govern). Name and role of the person accountable, and who can switch it off.
2. Data in (Map). What the tool reads: open estimates, customer contact details, texting consent. What it writes, if anything.
3. What we count (Measure). The two or three numbers from Step 4, and the result that would make you stop.
4. When it's wrong (Manage). Who fixes it, how it gets logged, and how fast the vendor has to respond.
5. Approval rule. Nothing goes to a customer or changes a repair order until the named person approves it.
6. End date. The day you'll decide to scale, change or stop.

Keep the page with the vendor agreement, and give a copy to each pilot shop's manager.

Step 4. Count your own baseline

This is the step that tells you whether the pilot worked, and it has to happen before you start. For two to four weeks, count by hand or from reports:

• How many estimates went past your cutoff with no reply.
• How many of those got a call or a text.
• How many were approved later.

Then run the pilot for the same number of weeks and count the same things the same way. We have no sourced figure for what groups typically gain, and we'd be wary of any vendor quoting one without showing how it was counted. Your count is the only one that describes your shops.

A simple tracking sheet, one row per week per shop:

Week	Shop	Stale estimates	Contacted	Approved later
Baseline 1	Elm St (example)	count	count	count
Pilot 1	Elm St (example)	count	count	count

Scale what held up

At the end date, look at three things: did the numbers beat the baseline, did staff keep using it after the first week, and did the approval step catch real mistakes. If all three are yes, roll to the next shop, one at a time, with the first pilot manager coaching the next. If not, you've lost a few weeks, not a platform contract. Add a second workflow only once the first is routine.

What stays with a person

Pricing, customer relationships, hiring, safety calls on a vehicle, and any message that commits the shop to something. The AI prepares. A named person decides.

Where RevvOS fits

RevvOS runs its group work the same way: a 45-day Operating Control Sprint on one real workflow with one to three shops, starting from a Multi-Shop Operating Gap Assessment [ours]. Its all-shops view shows what needs attention in each shop with the source and last update visible, marks missing data as unavailable, and prepares drafts for a named person to review. For the wider plan read AI for automotive service groups, or see multi location auto shop AI for the daily view.