Presenting, Deploying and Evolving Your Plan

A well-built plan that nobody understands or approves will stay in a drawer. This final lesson helps you present your plan to your manager or your team, set out a realistic 30/60/90-day roadmap, finalize your usage charter and indicators, and organize the monitoring and learning that will keep your practice reliable as tools and rules evolve.

You now have a scope, documented workflows and measured results. The last step is to turn them into a short, convincing and honest presentation, and then into a plan you can actually carry out.

Presenting to your manager or team

Camille has fifteen minutes with her director. She knows that long technical explanations will lose the audience. She builds her presentation around five questions that any decision-maker has in mind.

1. What problem are we solving? The tasks that consume time or cause errors, shown with your inventory.
2. What do you propose? The three to five workflows and the scope, one line each.
3. What did you find? Your measured results, honestly presented, with sample sizes and limits.
4. What are the risks, and how are they handled? Data rules, human validation, red lines, incident procedure.
5. What do you need from us? Approval of tools, time, a decision on a specific point, a budget, training.

A short structure works well: one slide (or one page) per question, plus a one-page appendix with the workflow sheets. Lead with the problem and the results, not with the technology.

A ready-to-use prompt: prepare the presentation

I must present my AI adoption plan to [audience, e.g. my director] in [10 to 15] minutes. Here is my material:
- Problem and tasks: [summary]
- Retained workflows: [list]
- Results so far (measured): [summary with sample sizes and limits]
- Risks and safeguards: [summary]
- Requests: [what I need]

Propose a storyline of 5 slides with a one-sentence message per slide, the key points to say, and the likely objections from this audience with a suggested honest answer to each. Tell me where my evidence is weak so that I do not overclaim. Do not invent figures or benefits.

Rehearse the objections: "What about confidentiality?", "What if it makes a mistake with a customer?", "Is this going to replace people?", "What does it cost?", "How do we know it works?". Your documents contain the answers, which is the whole point of preparing them.

The 30/60/90-day plan

A roadmap makes your plan concrete and limits the risk of doing too much too quickly. The plan below is a model to adapt; the dates and content are yours.

Period	Focus	Typical actions	Decision point
Days 1 to 30: prepare and pilot	Safe foundations and first workflows	Get approval of tools and charter; configure settings; take baseline measures; train the pilot group; run workflow 1 and 2 with full human review	Day 30: are the data rules respected? Are the first results usable? Adjust prompts.
Days 31 to 60: extend and refine	Broader use, better prompts	Add workflow 3; improve prompts from test results; share the prompt library; hold a short feedback session; log incidents and near-misses	Day 60: compare before and after; keep, adjust or stop each workflow
Days 61 to 90: consolidate and report	Stability, evidence, next steps	Re-measure; update the charter; finalize documentation; present results and a recommendation to management; plan the next quarter	Day 90: decide what becomes standard, what to extend, what to drop

Write the plan with owners and dates for each action, and a margin for the unexpected. An empty owner column is a sign that the plan will not happen.

Finalizing the charter

Bring together the usage rules you drafted in chapter 5 into one page that fits your context. For an individual or small team, it can be short. Here is a model.

MY (OUR) AI USAGE CHARTER
1. Tools: I use only [approved tools]. I request approval before using a new one.
2. Data: Green data in any approved tool; orange data only anonymized and in approved tools; red data never, unless specifically authorized.
3. Validation: I read and verify everything before sending, publishing or using it. I am responsible for the result.
4. People: I do not use AI to decide about, rank or evaluate people.
5. Sensitive fields: legal, tax, medical, safety and financial points are validated by a qualified person.
6. Transparency: I mention the use of AI when [rules or situations] require it, and I never present AI-generated content as something it is not.
7. Traceability: I keep prompts and versions for [which workflows] and a short log of incidents.
8. Incidents: I stop, report to [contact] and correct.
9. Learning: I review this charter every [3 months].

Have the charter reviewed by the people responsible for data protection, security or legal matters, where they exist. For personal data, keep in mind that the GDPR and, where applicable, the European regulation on AI set obligations on organizations; the CNIL and EUR-Lex are good official sources for the exact texts.

Follow-up indicators

Choose a small dashboard that you can really update. For each workflow:

• Usage: how often is it used, and by whom?
• Time: average time per task compared with baseline.
• Quality: average grid score and number of corrections.
• Satisfaction: a simple rating from users and recipients.
• Risk: incidents, near-misses and rule exceptions.
• Cost: licenses and time invested.

Review it at fixed intervals (for example every month during the pilot, then every quarter). If an indicator is never read, remove it.

Monitoring and continuing to learn

Tools, practices and regulations change quickly. A plan that is not maintained gradually becomes outdated or unsafe. Build habits that keep you up to date without taking over your time.

• A regular slot: thirty minutes each month to review results, update prompts and read about changes in your tools' terms or settings.
• Reliable sources: official publications of data protection authorities (such as the CNIL), EU institutions (EUR-Lex for texts, the European Commission for guidance), your professional body and your tools' documentation. Be selective about commentary and marketing.
• Peer learning: exchange prompts and failures with colleagues or your professional network.
• Revisit the basics: from time to time, return to the lessons on verification, bias and privacy: they matter more than any new feature.
• Re-assess regularly: tasks change, so repeat the audit from chapter 1 once or twice a year.

Bad use versus good use

Bad practice	Good practice
Presenting only the impressive examples and the technology.	Presenting the problem, measured results with limits, risks and clear requests.
A roadmap with no owners, dates or decision points.	A 30/60/90 plan with a person and a date for each action and a keep/adjust/stop decision.
Writing the charter once and never revisiting it.	Scheduling reviews and updating the charter when tools, rules or practices change.

Your turn: complete your plan

Exercise. Finish your adoption plan. (1) Write your five-slide storyline using the prompt above and list three objections with answers. (2) Fill in your 30/60/90-day plan with owners and dates. (3) Finalize your charter on one page. (4) Choose your dashboard (maximum six indicators) and its review dates. (5) Write your monitoring routine: your monthly slot and three reliable sources. Present the plan, even informally, to one person and note their questions.

See a solution

Marc has no director, but he presents his plan to his accountant and his apprentice. His storyline: (1) "Paperwork eats two evenings a week"; (2) "Three workflows: quote descriptions, site reports, customer messages"; (3) "On ten reports, the average time with the prompt, including checking, was lower than by hand; two reports needed corrections"; (4) "I never put names, addresses or photos in the tool, and I verify prices, standards and dates myself"; (5) "I ask my accountant to confirm the rules on invoicing and data records".

His 30/60/90 plan: days 1 to 30, set up the business account, write the charter and test two workflows; days 31 to 60, add customer messages and ask his apprentice to try the site report; days 61 to 90, compare results, decide what to keep and write the next quarter's plan. His accountant asks how he keeps customer data out of the tool, which leads to one more line in his charter: "customer details are added by hand at the end".