# A practical 30-day AI rollout

> Plan a focused pilot with training, clear ownership and measurable quality before expanding to more teams.

Introduce MAIA through one useful task people already do. This suggested 30-day plan is a planning aid, not a promise that every rollout takes a month. Adjust it to your access approvals, available time and review requirements.

## Before the pilot

Choose a small group with different experience levels. Name a sponsor who can resolve blockers, a person who owns the task, an admin who handles access and a reviewer who knows what a good result looks like. One person may hold several roles.

Agree which information may be used and which actions need approval. Prepare representative, permitted examples. Tell colleagues how to get help and make participation about improving a process, not ranking individual activity.

## Week 1: establish a baseline

Pick one or two tasks using the [use-case guide](/en/docs/guides/use-cases/). Record how they work today: preparation, execution, review and correction time. Keep a few examples of acceptable finished work.

Set up [workspace access](/en/docs/first-steps-with-maia/setup-workspace-and-add-users/) and test the task as a regular member. Confirm that the source material is visible to the intended participants. A successful admin test does not establish access for the whole team.

Deliverable: a one-page pilot brief with an owner, examples, review criteria and a decision date.

## Week 2: practice together

Run a 45-minute workshop:

| Time       | Activity                                                    |
| ---------- | ----------------------------------------------------------- |
| 5 minutes  | Explain the task and the boundaries                         |
| 10 minutes | Demonstrate a request, an imperfect answer and a correction |
| 20 minutes | Let participants try their own permitted example            |
| 10 minutes | Compare results and collect questions                       |

Pair a newcomer with an experienced colleague. Give beginners the [AI introduction](/en/docs/guides/ai-basics/); give advanced users the [prompt guide](/en/docs/guides/better-prompts/) and an incomplete-source test case. Keep a short list of recurring questions for the next session.

## Week 3: make the method repeatable

Refine one shared method from the successful examples. Use a [Skill](/en/docs/maias-features/skills/) for reusable instructions or a [workflow](/en/docs/workflows/) for an ordered process. Follow the workspace's [Skills approval process](/en/docs/admin/skills/) when distributing workspace-wide or required Skills.

Assign a maintainer and keep the evaluation examples. Hold a short office hour for problems that appear in real work. Track missing access, confusing instructions and poor source quality separately: they need different fixes.

## Week 4: decide using evidence

Compare the pilot with the baseline using [quality and effort measures](/en/docs/guides/evaluate-results/). Include the time spent checking and correcting the output. Ask participants whether the result was actually useful.

* **Expand** when the method consistently meets the agreed quality bar and has an owner.
* **Improve** when the task is valuable but a fixable source, access or instruction problem remains.
* **Stop** when review effort, errors or unsuitable inputs make the task a poor fit.

Keep the next rollout small enough to support. Recheck examples after important changes to instructions, sources, models or access.

## Next steps

Use the [pilot brief and retrospective template](/en/docs/guides/templates/) to record your decisions. Find the existing MAIA product outlook in the [public roadmap](/en/docs/roadmap/public-roadmap-for-maia/) and published changes in the [release notes](/en/docs/releases/).