Home / Leadership guide
How to recommend AI to your boss (and look prepared)
A beginner's step-by-step path from "I think AI could help" to a one-page proposal your leadership can say yes to. No tech background needed.
Education, not legal, compliance, HIPAA, or professional advice. If your proposal touches health, financial, student, or HR records, get a lawyer or compliance professional to review it before anything is bought or used.
The 9 steps
- Find 1–3 pain points. Pick tasks that are repetitive, text-heavy, and low-risk: email replies, summaries, quotes, procedures. Avoid anything with Red data. Use your role page for ideas.Fill in: Task: ________ · Who: ________ · How often: ________ · Why it hurts: ________
- Measure the baseline. For one week, time each task (a phone stopwatch is fine). Note how many times it happens and how often it needs redoing.Fill in: Task ____: average ____ minutes × ____ times/week = ____ minutes/week
- Check what you already pay for. Look for Copilot Chat in Outlook/Word or Gemini in Gmail/Docs. Ask whoever manages accounts. Free and already-approved beats new spending.Fill in: We use: ☐ Microsoft 365 (plan: ____) ☐ Google Workspace (plan: ____) ☐ Neither
- Run a small, safe pilot. 1–3 people, 3–4 weeks, Green/Yellow data only, a human reviews everything. Keep a simple log.Fill in: Date · Task · Minutes with AI · Big fixes needed? (Y/N) · Any data worry? (Y/N)
- Compare 2–3 tools. Use the Tools page: monthly cost, whether data trains models, BAA if needed, admin controls (SSO, removing users, retention), and ease. Write the date and source next to every price.Fill in: See section 3 of the proposal template.
- Write the one-page proposal. Problem → pilot results → options → cost and return → safeguards → decision needed. Download the template below.Fill in: Download: ai-pilot-proposal-template (.docx or .md)
- Build 5 slides (optional). Problem · Pilot · Options compared · Risks & safeguards · Ask & success criteria.Fill in: Slide 1 headline: 'We spend ____ hours/month on ____.'
- Prepare for the tough questions. Leaders ask about security, cost, ROI, and legal risk. Practice the answers below in your own words.Fill in: My answer on security: ________
- Ask for a clear decision and a review date. Request approval for a specific budget and end date, then report back with numbers whether it worked or not.Fill in: Decision by: ____ · Review on: ____
What a good result looks like
Example (made up for illustration): "Two of us tried Copilot Chat for 3 weeks on quote follow-ups. It cut each one from about 12 minutes to about 5, roughly 9 hours a month across the team. No customer data was entered. I'd like to keep it at no extra cost, add one person, and review in 60 days."
Report your real numbers, even if they're small. Leaders trust measured results more than big promises.
Questions leadership will ask, and how to answer
Security: Is our data safe? Who can see it?
Say: Use a business plan, not personal accounts. On ChatGPT Business/Enterprise, Claude Team/Enterprise, Grok Business, Copilot and Gemini for Workspace, the vendor says business data isn't used for training by default (see the Tools page with sources). Everyone gets their own login with 2-step login; we remove access the day someone leaves.
Cost: What will this really cost?
Say: Start at $0 with what we already pay for (Copilot Chat or Gemini). A paid seat is about $20–30 per person per month for ChatGPT Business, Claude Team or Grok Business, or $23.50 for Microsoft 365 Business Standard + Copilot. We pilot with 1–3 people before buying more.
ROI: How do we know it's worth it?
Say: We measured baseline minutes per task before starting. After 3–4 weeks we compare minutes, rework rate, and errors. If saved time × hourly cost doesn't beat the seat price, we stop.
Legal & privacy: Could this get us in trouble?
Say: We keep Red data (health, SSNs, card/bank, passwords, HR files) out entirely during the pilot. If we ever need it, we'll only use a vendor that signs a BAA or the right data agreement, and we'll get a lawyer or compliance professional to review first. AI won't make decisions about hiring, firing, credit, or housing.
Accuracy: What if it's wrong?
Say: A person reviews every output before it reaches a customer. We track how often drafts need big fixes; that's a pilot metric.
Staff: Will people be replaced?
Say: The pilot targets repetitive writing and summarizing so people spend more time with customers. Any bigger change would be a separate decision.
Vendor lock-in: What if we switch later?
Say: We'll pick monthly billing for the pilot when it's offered, keep our documents in our own systems, and write prompts down so they move with us.
The 5-slide outline
- The problem: 1–3 tasks, baseline minutes, what that time costs.
- The pilot: who, what, how long, which data was allowed, results.
- Options compared: cost, training default, BAA, admin controls, ease (with sources and dates).
- Risks & safeguards: data colors, human review, logins, what AI will not do.
- The ask: budget, start/review dates, success criteria, and when to stop.
Pick tools by facts
Use the tool comparison (prices and BAA facts with official links, checked Oct 2026) and your role page for ranked tasks and prompts.
Free monthly digest
What changed in AI prices and medical-privacy agreements — for beginners. Education, not legal, compliance, HIPAA, or professional advice. We keep your email only to send the digest (no ads, never sold). Unsubscribe or delete anytime.
Not sure where to start?
Take the free 2-minute AI readiness quiz, or get the AI Starter Kit draft: policy template, checklists, 10 prompts, pilot worksheet.