Rollout, Handover & Keeping the Customer's Trust
Roll out in steps, train people, hand over cleanly and feed what you learned back to your team.
A project ends when real people use the solution and the customer can run it alone. Roll out in steps: shadow, assist, then auto on safe cases. Train users with their own examples. Hand over with a checklist. Keep updates honest, and tell your own team what you learned.
- Shadow mode
- The AI runs but people still do the work, so you can compare.
- Handover
- The guides and checks that let the customer run it without you.
- Champion
- A respected person in the team who helps others adopt the tool.
The project is not finished when the code works. It finishes when real people use the solution, the result is measured, and the customer can carry on without you. This lesson covers rollout, training, handover and the trust that brings the next project.
Roll out in steps
Avoid a big switch-on day. Move in steps. Start with a few friendly users. Then widen to one team. Then the wider group. Watch the success number at each step.
- Shadow mode: the AI runs but people do the work as before. Compare results.
- Assist mode: the AI drafts and people approve each result.
- Auto mode: the AI acts alone only on safe, well-tested cases, with spot checks.
Most projects should stay in assist mode for a while. Move a task to auto mode only when the test results and the spot checks are good.
Train the people
A tool nobody trusts will not be used. Train the users in plain terms. Show what the AI does well. Show where it fails. Teach them how to check its work and what to do when it is wrong. Give them a short guide and a place to ask questions.
- Run a short live session with real examples from their own work.
- Give a one-page guide with screenshots.
- Name a champion in the team who can answer first questions.
- Make it easy to report a bad answer in one click.
Watch it run
After launch, watch the system. Track the success number, the error reports and the cost. Review a sample of results every week. Fix problems fast, and tell users what you fixed. People return to a tool that listens.
Write the handover
Leave the customer able to run it. A good handover has a short guide, a map of how it works, a list of settings and a plan for common problems.
HANDOVER CHECKLIST [ ] How it works: one diagram and one page of plain words [ ] Where everything lives: code, prompts, settings, keys (names only) [ ] How to check it is healthy: dashboard, alerts, who gets them [ ] How to fix common problems: step by step [ ] How to change a prompt or rule safely, and how to test the change [ ] Costs: what it costs to run each month, and what changes it [ ] Who owns it now: names and backup names [ ] Review date: when we look at results together again
Keep the customer's trust
- Send a short update on the same day every week: done, next, blockers.
- Bring bad news early, with a plan. Never hide a problem.
- Say no to scope changes kindly, and offer the next phase.
- Report results in the customer's own numbers, such as hours saved a month.
Feed the product
Your own team needs what you learned. Write down the repeated requests, the missing features and the fixes you built by hand. If three customers ask for the same thing, the product should have it. This is how forward deployed work turns into lasting value for your company.
- ✓Roll out in steps: shadow mode, assist mode, then auto mode on safe cases only.
- ✓Train users with their own examples. Make it easy to report a bad answer.
- ✓Watch the success number, the errors and the cost after launch, and fix problems fast.
- ✓Hand over with a checklist, keep weekly updates honest, and tell your own team what you learned.
Plan a rollout and handover
- Choose the project from your spec. Write a three-step rollout plan with the user group at each step.
- Write a one-page training guide for a new user, including what to do when the AI is wrong.
- Fill in the handover checklist as far as you can for your idea.
- Write the weekly update you would send in week two: done, next and blockers.
Q1Why use shadow mode first?show answer
The AI runs without affecting real work, so you can compare its results with people's results and find problems safely.
Q2What belongs in a handover?show answer
How it works, where things live, how to check health, how to fix common problems, how to change prompts safely, costs, owners and a review date.
Q3How does forward deployed work help your own company?show answer
You record repeated requests and fixes. If several customers need the same thing, it becomes a product feature.