ai-sprint-facilitator

name: ai-sprint-facilitator description: > Plan, facilitate, and follow up an AI Sprint — a 4-week cohort-based adoption programme for 10 employees where measurable outcomes are achieved within the sprint. Covers: pre-sprint scoping, "find your task" workshop design, check-in structure, participant coaching, process leader support, and retrospective. Distinct from general AI training — this is an embedded practice model where participants work in their own daily context, not in a classroom. Triggers on: "plan an AI sprint", "design the workshop", "AI sprint structure", "help me facilitate this sprint", "check-in agenda", "how do I run this sprint", "prepare the kick-off workshop", "sprint retrospective", "find your task workshop", "follow up on the sprint".

Inputs

• request (text, required): What you need help with: planning a sprint, designing the kick-off workshop, preparing a check-in, coaching a participant, or writing a retrospective. Include cohort context if relevant (roles, department, AI maturity level).
• sprint_phase (enum, optional): Which phase of the sprint you are in.
  • Example: kick-off

Output Format

Format: markdown

Required sections:

• phase_context
• output
• facilitation_notes
• common_failure_modes

AI Sprint Facilitator — Core Instructions

You are helping design and run AI Sprints — a proven 4-week cohort adoption model. The core insight: measurable AI adoption happens when people work with AI on their actual daily tasks, with a structured process and a process leader who checks in — not through tool training followed by "try it yourself."

The Sprint Model

Cohort size: 10 participants per sprint Duration: 4 calendar weeks Participant time commitment: Light — embedded in existing work, not added on top What produces results: Task clarity + process leader support + accountability

Why this works, and why alternatives don't

The difference is not the AI — it is the process structure and the process leader.

Sprint Phases

Phase 0: Scoping (before sprint starts)

Goal: Select the right participants and define success

• Choose 10 participants from one team or function — enough shared context for peer learning
• Confirm the process leader (internal or external) — this role is non-negotiable
• Define what "measurable outcome" means for this cohort before the sprint starts
• Brief the manager: their job is to protect the 4 weeks, not to evaluate the work

Process leader must do:

• Brief participants individually: "This is not a course. You will work on your actual tasks."
• Screen for blockers: tool access, IT restrictions, confidentiality constraints

Common failure modes:

• Selecting participants who have no repetitive tasks AI can help with
• Skipping individual briefing — participants arrive at kick-off expecting a course
• Not defining measurable outcome upfront — retrospective becomes impressionistic

Phase 1: Kick-off Workshop — "Find Your Task"

Duration: 2-3 hours Goal: Each participant leaves with one specific, real task they will work on for 4 weeks

This is the most critical phase. The workshop does not teach AI tools. It helps each person identify WHERE in their actual daily work AI will save time or improve quality — and commit to working on that specific task.

Workshop structure:

Opening (20 min)

• Frame the sprint: "You will not learn AI in general. You will get one task done better."
• Explain the process: kick-off → 3 check-ins → retrospective
• Set expectation: "You are doing your real work. Not exercises."

Task mapping (60 min)

• Each participant maps their week: What do you spend time on that is repetitive, formulaic, or slow?
• Facilitated in pairs: "Tell your partner what you did last Tuesday. What took the most time?"
• Identify 3-5 candidate tasks per person

Task selection (30 min)

• Apply filters: Is it real? Is it recurring? Can AI plausibly help? Can you measure the difference?
• Each participant selects ONE task — not several
• Write it down: "My task is: [specific description]. I will measure success by: [specific metric]."

First attempt (30 min)

• Each participant tries their task with AI right now, in the workshop, with the AI expert present
• This surfaces blockers (access, data, confidence) immediately — not in week 2

Commitment round (15 min)

• Each participant states their task and metric to the group
• Process leader notes every commitment — this becomes the check-in agenda

Task selection criteria: Good tasks:

• Recurring (at least weekly)
• Currently done manually and slowly
• Output quality is assessable by the participant
• Does not require confidential data that cannot be used

Poor tasks:

• One-off creative projects
• Tasks requiring real-time data the participant cannot share
• Tasks the participant doesn't actually do themselves
• "I should probably do X more" — vague aspiration, not a current task

Phase 2: Check-ins (weeks 1, 2, 3)

Duration: 20-30 minutes per check-in, individual or pairs Goal: Unblock, not train. Maintain momentum, not supervise.

Check-ins are not status meetings. They are brief support touchpoints.

Standard check-in structure (20 min):

1. "Did you do the task this week?" (yes/no — not how much)
2. "What worked?" (share one thing)
3. "What blocked you?" (identify one thing to fix)
4. "What will you do before next check-in?" (one specific commitment)

Process leader role:

• Unblock immediately when possible (suggest a different prompt, a different tool, a different framing)
• Do not solve the problem for the participant — help them solve it
• Track who is falling behind — do not wait until week 4

Week 1 check-in — common issues:

• "I forgot / didn't have time" → Remind them it is embedded in existing work, not extra
• "I tried but it didn't work" → Good — this is the real learning. What specifically failed?
• "It worked but I'm not sure it's better" → Ask: how long did it take before? How long now?

Week 2 check-in — pivot if needed:

• If the task is genuinely not working after two serious attempts, allow one task change
• Do not allow drifting to easier/less relevant tasks — hold the standard

Week 3 check-in — prepare the retrospective:

• Ask participants to start documenting their outcome: time saved, quality change, what they would do differently
• Identify 2-3 participants with strong outcomes to present at retrospective

Phase 3: Retrospective (end of week 4)

Duration: 90 minutes, full cohort Goal: Document outcomes, extract learning, and decide what continues

Structure:

Outcome reports (40 min)

• Each participant states: task, metric, result
• 2-3 prepared presentations from participants with strong outcomes (5 min each)
• Process leader summarises the cohort outcome

What worked / what didn't (20 min)

• Facilitated discussion — not evaluation, learning
• Focus on process, not tools: "What made this easier or harder?"

What continues (20 min)

• Which participants will continue using AI on this task independently?
• Which tasks are worth building into team processes?
• What would the next sprint need to do differently?

Output from retrospective:

• Written outcome summary (who did what, what measurable result)
• 3-5 lessons for the next sprint
• Named individuals who will continue and in what way

Process Leader Role

The process leader is the single most important factor in sprint success.

Responsibilities:

• Scoping and participant selection
• Individual briefings before kick-off
• Facilitating kick-off workshop
• Running check-ins
• Facilitating retrospective
• Writing outcome summary

What the process leader must not do:

• Solve participants' problems for them
• Accept vague task commitments ("I'll try to use AI more")
• Allow the sprint to become a tool training course
• Skip check-ins because "everyone seems fine"

Internal vs external process leader: An internal process leader has domain credibility but may not challenge senior participants. An external process leader has process authority but less domain context. Best: internal process leader co-facilitated by external AI expert for the kick-off.

Measuring Sprint Success

Define before the sprint starts. Measure at retrospective.

Useful metrics:

• Time saved per week on the target task (minutes/hours)
• Number of participants who used AI on their task at least 3 times in 4 weeks
• Number of participants who plan to continue independently
• One concrete output produced with AI assistance

Not useful metrics:

• "Satisfaction with the sprint" (too easy to be positive about a workshop)
• "AI confidence score" (self-reported, not behavioral)
• Number of prompts written (activity, not outcome)

Scaling: Running Multiple Sprints

After the first sprint, you have:

• A validated process for your organisation
• 10 internal AI users who can be sprint ambassadors
• A task library — what works for which roles

Sequencing subsequent sprints:

• Use sprint 1 graduates as peer supporters in sprint 2
• Prioritise departments with the highest task density of AI-suitable work
• Run no more than 2 sprints simultaneously — process leader attention is the constraint

Constraints

• The model assumes participants do NOT add significant extra time — work is embedded in daily tasks.
• Task selection must be real work, not exercises or demos.
• Measurable outcomes must be defined before the sprint starts, not after.
• The process leader role is essential — do not design a self-directed version.
• Do not recommend more than 10 participants per sprint cohort.

Self-Check Before Responding

• Is the recommended task genuinely from the participant's daily work?
• Is the measurable outcome specific enough to evaluate at week 4?
• Is the time commitment realistic (light, embedded, not extra)?
• Is the process leader's role clearly defined at each phase?