adopt-hands-on-session-plan git:20261011.dce569cA

Immutable. This exact content never changes and is served at /api/v1/blob/73a39eb53831d7ce.

---
name: adopt-hands-on-session-plan
description: Writes a timed plan for one hands-on AI session where each person brings a real task, with a live demonstration, a practice round, a compare-and-check round, facilitator what-if notes and a no-screens fallback. Use for "run adopt-hands-on-session-plan", "hands-on AI session", "AI workshop plan", "plan an AI training session", "AI session where people bring their own task", "AI workshop agenda", "what if the tool is wrong live", "AI session without screens", part of the Claude Guide for AI Adoption Pack by Polar Bear.
---

# Hands-On AI Session Plan
## When To Use
The last AI session was a webinar where one person shared a screen and nobody touched the tool. You have one session to run and you want every person to leave having worked their own task with AI and checked what came back. It answers: what happens in the room, in what order, and what do I do when it goes wrong?

## When Not To Use
If you are deciding what the whole path should cover across several units, use AI Basics Curriculum: this skill plans one session only. If people cannot name a task they would use AI for, run AI Use Case Map first, because a session with nothing real to work on turns back into a demo.

## Inputs
- The unit or topic, who attends (roles and group size), the time you have, the room or call setup and the tool access people will have.
- Two or three real tasks from these roles, one real example the facilitator can work live, and your organisation's never-put-in list.
If you have none of this, I start from the roles, the session length and one real task, and mark the output as a first draft.

## Approach
The plan follows the first principles of instruction (Merrill, 2002): learning is built around a real problem, it activates what people already do, it demonstrates, it lets people apply, and it carries into their work. The judgement is in the proportions: application gets the largest block, and the demonstration includes a moment where the output is wrong and gets caught. The failure this prevents: a facilitator runs a flawless rehearsed demo, the room nods, and the first time anyone meets a wrong answer is alone at their desk with a deadline.

## Workflow
1. Ask at most three questions: which roles and how many people, how long the session is, and what tool access each person will have on the day.
2. Problem: write the pre-session message. Each person brings one real task, sanitised, sent in advance; say what never to bring (anything on the never-put-in list) and offer a spare task for anyone who arrives without one.
3. Activation: plan a short round on "how do you do this task today, and where does it hurt?". It surfaces the criteria people already use to judge the work, which you will need in the check round.
4. Demonstration: script one real example the facilitator works live, with the reasoning said aloud, including one point where the output is wrong and the facilitator catches it against the source. Do not hide the miss; it is the lesson.
5. Application: everyone works their own task. Inside this block, plan the compare-and-check round: pairs compare outputs on the work and check them against the source material. Feedback is on the output, never on the person who produced it.
6. Integration: each person writes what they will try this week and what they need for it. They choose it and keep it; the facilitator collects needs as themes, with no names. Then write the facilitator's what-if notes and the no-screens fallback: the same five phases on printed outputs.

## Output Format
```markdown
# Hands-On AI Session Plan
Session: [unit or topic] | Roles: [roles] | Group size: [you set] | Length: [you set] | Facilitator: [role] | Co-facilitator: [role or none]
Pre-session message. Bring: [one real task, sanitised] | Never bring: [items from the never-put-in list] | Spare task for anyone without one: [task]
## Timed plan
| Phase | Time | What people do | What the facilitator does | Materials |
|---|---|---|---|---|
| Problem | [min] | [state their real task] | [confirm each task is safe to use] | [tasks sent ahead] |
| Activation | [min] | [say how they do it today and where it hurts] | [note the quality criteria people name] | [board] |
| Demonstration | [min] | [watch and name what was checked] | [work one real example, catch the wrong output] | [example and its source] |
| Application | [min] | [work their own task] | [circulate, unblock access] | [own task] |
| Compare and check | [min] | [pairs compare outputs and check them against the source] | [ask "where is that from?"] | [source material] |
| Integration | [min] | [write what they will try this week and what they need] | [collect needs as themes, no names] | [card each person keeps] |

## What-if notes
| If | Then |
|---|---|
| The tool gives a wrong answer live | [stop, check it against the source in front of the room] |
| Someone has no access | [pair them, log the access gap for [role]] |
| Someone brings a task that is off limits | [swap to the spare task, point to who to ask] |
| A sceptical question | [take it seriously, test it on a real task] |
| The question about jobs | [hand it to the prepared answer from [named leader], do not improvise] |

## No-screens fallback
[The same five phases on printed outputs: what is printed, who brings it, what the check round uses]
## Decision
[Facilitator] confirms timings and the live example by [date]. [Sponsor or session owner] confirms tool access and the room by [date]. Each person decides what they try afterwards.
```

## Done When
- All five phases appear in order, application with its check round is the largest block, and the demonstration names the point where the output is wrong and how it gets caught.
- The pre-session message says what to bring and what never to bring.
- The what-if notes cover the five cases, and the fallback works with no screens.

## Quality Bar
- Every person touches their own task; a session where only the facilitator types has failed.
- The check round compares outputs with source material, never people with each other.
- The jobs question is handed on to its owner, never answered from the front of the room. Timings are placeholders you set.
- Each person chooses the task and what they try next; the plan records no attendance score and no per-person follow-up.

## Next
Run adopt-ai-practice-tasks (AI Practice Tasks) to give people something to do with it between sessions.

## About the makers
This pack is made by Polar Bear, a consultancy built by ex-McKinsey founders with a dream to make AI work for People, not instead of them. We help our clients build people systems and AI-first ways of working, and we run our own company on Claude. If your team has outgrown the self-serve version, message Pauline (linkedin.com/in/paulinebertry).