AI enablement
Taking a team from ad-hoc AI to using it with judgment
For AI enablement & literacy roles
A non-technical team is using AI ad-hoc. Some people lean on it, some avoid it, the output quality is hit-or-miss, and no one is sure what is safe to put in. Here is how I take that team from random to reliable. It is the same method behind the adaptive AI training system I am building; the five-question readiness check below is playable.
When to trust AI, not just how to prompt it
The method stays the same every time; the language, examples, pace, and risk boundaries are calibrated to the people in the room. It starts by finding out where someone actually is.
- 5 questionsthe readiness check
- 3 levelsCurious · Practical · Fluent
- Human in the loopon every output that counts
The method
Diagnose, set the boundaries, teach to the level, apply, and check that it stuck.
- 01 DiagnoseFind out where each person actually is before scoping anything.
- 02 Set the boundariesPlain rules tied to the team’s real data: what never goes in, what’s fine, who reviews.
- 03 Teach to the levelSame core ideas, but language, examples, and pace calibrated to the room.
- 04 Apply on real workPractice on the tasks they actually do, with human review, until it repeats without me.
- 05 Check adoptionDid it stick: usage, confidence, output quality.
01 · Diagnose, playable
Five questions, an honest read on where you are with AI, and a next step scoped to the answer. Nothing is saved or sent.
Five quick questions and I'll tell you where you're at with AI and what to do next. Answer honestly; there's no wrong score, and nothing gets saved or sent.
How often does AI actually show up in your workday?
The answer comes back mediocre. What do you do?
How do you handle an AI answer you’re not sure about?
What decides whether something goes into an AI tool?
Have you built AI into something you do over and over?
You’re here
Curious
Everyone starts here. The trick is repetition, not talent.
You’ve dipped a toe in. AI is a novelty more than a tool right now, and results feel hit-or-miss, which is exactly what you’d expect at this stage.
Your next step
Pick one task you do every week and use AI on it five times in a row. Not five different things: the same thing, five times. That’s how the pattern starts to click.
What "better" looks like
Weak
write a summary of this meeting
Tightened
Summarize the notes below for someone who missed the meeting. 5 bullets max, plain language, lead with any decisions and who owns the follow-ups. Notes: [paste]
The weak version leaves every choice to the tool: how long, for whom, what matters. The strong one answers those before it asks, so the first draft is close instead of generic.
AI Readiness Check
Curious
completed the five-question AI readiness check
“Curious is a great place to start. Go run that one task five times.”
· AI Readiness Check · mygirlsarah.com
You’re here
Practical
You’re past the hard part. Now it’s about making the good days repeatable.
You use AI and it genuinely helps, but the quality still depends on the day and the task. You have habits; they’re not yet a method.
Your next step
Write down your own three-line checklist for a good prompt: what context it needs, what constraints to set, what to check in the output. Use it for a week. Adjust it.
Tightening a prompt you’d actually use
Weak
help me write an email to my team about the new intake process
Tightened
Draft a short email to my team introducing a new request intake form. Audience: non-technical, busy, slightly change-averse. Tone: friendly, not corporate. Cover what changes, why it helps them, and the one thing they need to do differently. Under 150 words.
Same task, but now the tool knows the reader, the tone, the three things to cover, and the length. That’s the difference between a draft you rewrite and one you send.
AI Readiness Check
Practical
completed the five-question AI readiness check
“Practical and climbing. Write that checklist down.”
· AI Readiness Check · mygirlsarah.com
You’re here
Fluent
This is where it gets fun. You’re not using AI anymore, you’re designing with it.
You use AI deliberately and safely. You know when to trust it, when to check it, and what never goes in. You’re ready to build it into how a team works, not just how you work.
Your next step
Take one recurring team task and design a human-reviewed AI workflow around it: inputs, the prompt, who checks the output, where it goes. Document it so someone else can run it.
From personal habit to shared workflow
Weak
I use AI to draft our status updates
Tightened
A defined step: paste the week’s ticket updates into a saved prompt that produces a 4-section status (shipped / in progress / blocked / next), a named person edits it, it posts to the same channel every Friday.
A personal habit lives in your head and breaks when you’re out. A workflow has an owner, a review step, and a home, so it survives you.
AI Readiness Check
Fluent
completed the five-question AI readiness check
“Fluent. Sarah would want to compare notes with you.”
· AI Readiness Check · mygirlsarah.com
That’s the Diagnose step. The real thing goes further: the teaching adapts to the room, and everyone leaves having practiced on their own actual work. This is the diagnose step of the adaptive AI training system I am building.
02 · Set the boundaries
Before anyone learns a prompt, the team learns what goes in and what never does, in plain language, tied to their own data. Pick a case.
- What it is
- Internal notes, real names, no regulated data
- Rule
- Fine in an approved tool, once anything sensitive is stripped
- Why
- Low risk, high value; the names rarely matter to the output
- Who reviews
- The author, before anything leaves the draft
- What it is
- A first pass at a doc, email, or summary you own
- Rule
- Encouraged. The person stays the editor and the decision-maker
- Why
- AI drafts, people judge; the accountability never moves
- Who reviews
- The author owns the final version
- What it is
- Anything that identifies a customer, moves money, or grants access
- Rule
- Never. Not in any general tool, no exceptions
- Why
- The downside is unrecoverable and the upside is small
- Who reviews
- Not applicable; it does not go in
- What it is
- Contract terms, controls evidence, anything in audit scope
- Rule
- Only in a tool with a data agreement, and only with sign-off
- Why
- The rule is the same one that governs the rest of the work
- Who reviews
- The control or compliance owner, before use
03 · Teach to the level
The core idea, in the learner’s own work. The difference between a weak ask and one that gets a usable first draft:
Weak
write a summary of this meeting
Tightened
Summarize the notes below for someone who missed the meeting. 5 bullets max, plain language, lead with any decisions and who owns the follow-ups. Notes: [paste]
The weak version leaves every choice to the tool: how long, for whom, what matters. The tightened one answers those first, so the draft comes back close instead of generic.
The result
A team that knows when to trust AI, not just how to prompt it, with guardrails they can explain to an auditor. The five-program curriculum, teaching deck, facilitator guide, and readiness diagnostic are built and refined one-on-one; the adaptive team version is what I am building now.
A generic scenario. The five-question check and the boundary rules are the real mechanics behind the adaptive AI training system.
Open to a senior full-time role. Hiring brief →