Claude for Dealers / Learn / 90-Day AI Rollout Plan
Leadership · GMs and Dealer Principals
The 90-Day Dealership AI Rollout Plan for GMs and Dealer Principals
Most dealership AI efforts die the same way. Someone gets excited, ten people get logins, nothing is written down, month-end hits, and by day 45 two people still use it. This plan is built to survive a busy month.
- Name one owner
- Write the one-page policy
- Accounts on a store plan
- First three knowledge files
- Pick the department that asked
- Ship three Skills
- Two champions, daily use
- Baseline the numbers
- Measure against baseline
- Second department
- First dashboard from an export
- Weekly 20 minute review
- First agent with a daily deliverable
- Connectors, read-only
- Document everything
- Decide the next 90
A dealership AI rollout that lasts follows four phases over 90 days. Days 1 to 14: name one owner, write a one-page policy, put staff on store-controlled accounts, and create three knowledge files. Days 15 to 30: ship three Skills in one department with two champions. Days 31 to 60: measure against a baseline, add a second department, build one dashboard. Days 61 to 90: build the first agent, add read-only connectors, document everything. The owner needs four to six protected hours a week.
- One named owner with four to six protected hours a week
- One department first. Pick the one that asked
- Three Skills before any agents
- Baseline your numbers before you start
- A 20 minute weekly review is what keeps it alive
- Written policy before the first login
Before day 1: two decisions only the dealer can make
- Who owns this? One person. Not a committee, and not "IT." The best owners are operators who are already the person others ask: a marketing director, a BDC manager, a controller, a sharp assistant manager. Give them four to six protected hours a week and say so out loud in a manager meeting.
- What is the goal in one sentence? "Every internet lead gets a quality first response in five minutes." "Every declined service line gets three follow-ups." A goal you can count. Not "use AI more."
Days 1 to 14: foundation
- One-page policy. What data never goes in. Human review before anything reaches a customer. Which accounts are allowed. Use AI Guardrails for Dealerships as the template.
- Store-controlled accounts. Move people off personal accounts. See Which Claude Plan Does a Dealership Need.
- Three knowledge files. Store facts, brand voice, and hard rules. How to write them: Markdown Knowledge Files for Dealership AI.
- Owner training. The owner works through How to Learn Claude and Projects, Skills, and Agents Explained.
Exit test: the policy is signed by the dealer, every pilot user is on a store account, and a new hire could read the three knowledge files and describe your store accurately.
Days 15 to 30: one department, three Skills
Pick the department whose manager asked for it. Enthusiasm beats strategy here. Then ship three Skills, no more.
| Department | Skill 1 | Skill 2 | Skill 3 |
|---|---|---|---|
| BDC | First-response draft | Aged-lead re-engagement | Thread summary for managers |
| Service | Inspection write-up | Declined-service follow-up | RO history summary |
| Used cars | Vehicle descriptions | Monday aging list | Recon chase |
| Marketing | Campaign brief to copy | Review replies | Answer pages for AI search |
Two champions use the Skills every day. The owner sits with them twice a week and fixes what is annoying. Record the baseline numbers in week one, before the Skills change anything.
Exit test: both champions would complain if you took the Skills away.
Days 31 to 60: prove it, then expand
Compare to baseline. Share the result in the manager meeting, including what did not work. Then add the second department, and build one dashboard from an export using this guide so the results are visible without anyone assembling them.
Start the weekly review now. Twenty minutes, same time every week: what shipped, what broke, what is next. This meeting is the single best predictor of whether the program exists at day 90.
Exit test: one number moved and you can show it. Two departments have daily users.
Days 61 to 90: make it stick
- First agent. Wrap the proven Skills into an agent with a daily deliverable. One agent. Not four.
- Read-only connectors. Documents first, then calendar and email. See the connectors guide.
- Documentation. Every Skill, prompt, and knowledge file in one shared folder with a one-page "how this works." Name a backup owner.
- Vendor review. Run the build vs buy worksheet ahead of your next renewals.
The scorecard
| Measure | Baseline | Day 30 | Day 60 | Day 90 |
|---|---|---|---|---|
| Weekly active users | __ | __ | __ | __ |
| Skills in daily use | 0 | 3 | 6 | 8+ |
| The goal metric (your one sentence) | __ | __ | __ | __ |
| Hours saved per week, champion estimate | 0 | __ | __ | __ |
| Policy incidents | 0 | __ | __ | __ |
What kills rollouts
- One owner with protected time
- One department, then the next
- Skills saved and shared, not retyped
- A counted goal
- A weekly 20 minute review
- Everyone gets a login on day one
- No policy until something goes wrong
- Prompts live in one person's head
- “Use AI more” as the goal
- No meeting, so month-end wins
The shortcut
The fastest way through days 1 to 30 is to put your owner in a room for two days with operators who have done it, and have them come home with the Skills already built. That is what the Claude for Dealers workshop is for. The plan above works without it. It just takes the full 30 days.
Common questions
How do I start using AI in my dealership?
Name one owner, write a one-page policy, put pilot users on store-controlled Claude accounts, and ship three Skills in one department. Do not give everyone a login on day one. This page lays out the full 90-day sequence.
Who should lead AI at a dealership?
An operator with curiosity and credibility, given four to six protected hours a week. Marketing directors, BDC managers, controllers, and strong assistant managers make good owners. Do not assign it to a committee or default it to IT.
Which department should go first?
The one whose manager is asking for it. If nobody is asking, start with the BDC, because the work is writing at volume and results show up within days.
How long before we see results?
Champions usually feel time savings in the first week of daily use. A counted business result, such as response time or follow-up completion, typically shows by day 45 to 60 if you recorded a baseline.
How much does a dealership AI rollout cost?
Software is the small part: roughly $20 to $125 per user per month depending on plan. The real cost is the owner's four to six hours a week. Budget for that time or the rollout stalls.
Do we need an AI policy first?
Yes, and one page is enough. It should cover what data never goes into AI tools, human review before anything reaches a customer, and which accounts staff may use. Get it signed before the first login.
Compress Days 1 to 30 Into Two
Send your AI owner. They come home with the Skills built. Atlanta, January 5, 6, and 7, 2027.
Register Now →