# Session 3 Handout: Leading the Change

<p class="meta">FAST-09 AI for Managers and Leaders · 7formations · Instructor: Wadood Chaudhary · All examples fictional · Training, not legal advice</p>

## Why this session matters

Pew Research Center found in 2025 that U.S. workers were more worried than hopeful about AI at work. Tools don't fail in small businesses for technical reasons as often as they fail because people weren't told the rules, weren't trained, or felt watched. This session gives you the three things that make adoption stick: a one-page guideline, an honest conversation, and measures that respect people.

## Key concepts

### 1. The one-page team AI use guideline

| Heading | What it says |
| --- | --- |
| Approved tools | Named tools, work accounts only; new tools go through review |
| Never allowed | Employee records, customer personal information, PHI, pay data, bids, confidential drawings |
| Tasks | Encouraged / Allowed with a check / Off limits, with examples from your team |
| Who checks what | A named role per task type |
| Reporting mistakes | Same-day report to a named role; honest reports are not punished |
| Review date | When the guideline will be updated |

It must be **consistent with the employer's policy and Massachusetts data law**. A data slip may trigger duties under G.L. c. 93H and the written information security program (201 CMR 17.00), which is why fast, blame-free reporting matters.

### 2. Talking with staff

- **Be honest** about what will and will not change. Don't promise "nothing will ever change."
- **Listen first.** Ask what worries them most.
- **Explain how AI-assisted work will be judged:** same quality standard; the sender owns it.
- **Train before you expect use.**
- **Invite ideas.** Experienced staff spot errors and good use cases fastest.

A simple structure for the change: Awareness (why), Desire (what's in it for the team), Knowledge (training), Ability (practice with support), Reinforcement (share results, fix problems). This is Prosci's ADKAR model.

### 3. Measure outcomes, not people

| Good measures | Avoid |
| --- | --- |
| Team minutes per task, before and after | Keystroke or screen monitoring |
| Errors caught in review | Reading individual prompt histories |
| Customer callbacks or complaints | Leaderboards of AI usage |
| Two-line weekly self-report | Ranking people by "AI adoption" |

Some admin tools report usage; Microsoft 365, for example, can conceal user names in its usage reports. Use team-level views and share results with the team.

### 4. The 90-day plan

| Days | Focus |
| --- | --- |
| 1-30 | Guideline, training, one pilot, baseline |
| 31-60 | Measure, adjust, add a second use case |
| 61-90 | Review against targets; keep, change, or stop; update the guideline |

Each step has an owner, a date, and a measure. The communication plan says what, who, when, and how questions are answered.

## Worked example (fictional, Cape Harbor Inn)

Pilot 1: AI drafts replies to common guest emails from an approved FAQ; staff edit and send. Baseline 6.5 h/week; target 3.5 h/week with zero wrong policy statements sent. Kickoff at the Tuesday huddle, a weekly two-line update, open Q&A at week 6, results at day 90. Measures are team-level only.

## Checklist

- [ ] Guideline fits on one page and matches employer policy.
- [ ] Never-allowed data listed specifically.
- [ ] Mistake reporting is fast and blame-free.
- [ ] Training scheduled before the pilot starts.
- [ ] Each pilot has a baseline, target, owner, and stop condition.
- [ ] Communication plan includes a way for staff to raise concerns.

## Do and don't

| Do | Don't |
| --- | --- |
| Name approved tools | Say "use whatever helps" |
| Listen before explaining | Lead with a technology lecture |
| Measure the team's results | Monitor individuals |
| Plan in 30-day steps | Roll out everything at once |
| Schedule the day-90 decision | Let the pilot drift with no review |