AI Fluency for Small Businesses: teaching notes

"Tips and tricks date fast; the tools change every few months. This is the thing underneath that doesn't: four moves that make AI actually work for your business. Learn them once and they keep working whatever the tool does next. And the point isn't to do everything faster, it's to buy back time for the part of the business only a human can do."

"Here's the shape. We learn the framework, look at how the tool actually works, run its two loops, then put it to work on something real, and we finish with a one-page AI policy. If you lose the thread, this is the map back."

Pair the project-memory framing with the briefing-sheet move: "You find yourself re-explaining your business to AI every single time, who you are, what you sell, what you'd never compromise on. Do it once. Sit down and let it interview you about your business, then have it write that up as one document you paste in from then on. Every chat after that starts already knowing your business." (framings.md: the folder is the memory, the previewed version; 01-claude-101 slide 7, projects as a workspace that remembers.) Read the caption out loud: AI fluency is using AI efficiently, effectively, ethically and safely, and the efficiency exists to buy back time for the human work.

"Four habits, two loops. The outer loop is the bookends: Delegation, decide what's even worth handing over, and Diligence, own the final result with your name on it. The inner loop runs between them: Description, say what you want and how, and

"It's strong exactly where most of us are short on time, speed, breadth, tireless drafting. And it's bounded in ways that matter for a business: it has a knowledge cutoff, so recent or local facts can be stale. It can be confidently wrong. The best work pairs its speed and scale with your judgment, and you verify anything that would cost you money to get wrong." (05-... slide 7, strong-where-you're-weak.) Read the in-diagram line: it predicts plausible text, it does not look facts up.

Pick the route by audience (framings.md: deterministic vs probabilistic). For most small-business owners, the phone-keyboard route the deck draws: "It's the predictive text on your phone, scaled up. From what's been typed so far it guesses the most likely next word, picks one, adds it, and guesses again. That's all. Brilliant for drafting and summarising. But it's guessing what's plausible, not looking up what's true, so it can be completely certain and completely wrong." For analysts and regulated owners, the "cat sat on the mat" version. For technical peers, one sentence and move on.

"The inner loop in practice, and the place it matters most for you is research. Two halves. Going in, you describe: who you are, what your business does, the exact thing you need. Coming back, you never skip the check, is this real and current, is it right for a business my size, what's it missing? The big trap with research is that it'll happily invent a competitor or quote a stale number with total confidence. So anything that decides money, regulations, a competitor's price, a deadline, you verify at the source." Read the caption: the gap between what AI sees and what you know is where your judgment is irreplaceable.

"Same outer loop, now on your own business data, sales, feedback, customer lists. Two gates. Going in, you delegate deliberately: strip names, contacts, exact figures the AI doesn't actually need, and match the tool to how sensitive the data is, some tools train on what you type. Coming back, you own it: is this accurate to my records, is it useful, would I put my name on it? And the reassurance that matters, you connect what you trust, and a tool sees only what you can see; connecting your inbox doesn't expose anyone else's." (01-claude-101 slide 10, connectors; objections.md: IT/compliance/privacy.) Read the in-diagram line about acting fast if something goes wrong.

"Now operations, and this is all four Ds working together. Before you automate anything, sort the work. Don't ask 'can AI do this', ask 'should it?'. Three lanes: the stuff AI can just handle, your hours, your return policy, the documented FAQs; the stuff where AI drafts and you press send; and the stuff that stays human, complaints, contracts, anything needing judgement. Then for whatever you automate, describe it three ways, what it makes, the steps it follows, and how it must behave, including what it must never do. And test it on real cases before it ever touches a customer." (05-... slide 10 covers the three Description layers; this slide adds the sorting and the test-before-ship discipline.)

"The last move is to write your judgement down so your whole team works from it, not from whatever they each invent. One page, three sections, and the trick is to be honest in all three: what we use AI for, what stays human and why, and how we stay accountable when something goes wrong. This is the same idea as everything else today, the rules in your head, written down, so the work comes out your way every time, whoever's doing it. And the worry people raise, dependency, the answer is understanding: if you can explain what the AI is doing, you're fine. The time it buys back should free your team for more of the human work, not less." (framings.md: skill file as institutional memory, adapted from one expert's head to the team's shared page.)

"You don't get fluent by finishing a course, same as you don't get fluent in a language that way. You get there by doing it, badly at first, then better. So pick one real thing waiting for you this week, and run all four moves on it: decide what to hand off, describe it well, judge what comes back, own the result. Then do it again, it's a loop, not a line. And keep a running doc of what works, so next time you start warm." (05-... slide 15, fluency is built by practice. framings.md: the system improves through use.)

"Four moves, two loops, one practice. If the tools change next year, and they will, this still holds. Decide what to hand off, describe it well, judge what comes back, own the result. Run that on real work until it's saving you real time, and spend that time on the part of your business only you can do."