You Know Enough to Teach AI

You know enough about AI to help somebody who is a few steps behind you.

And get paid to do so.

Your feed makes that difficult to believe. It is full of researchers, developers and people who spend every working day testing new models or building businesses with 50 agents. You see them constantly, so their level starts to feel normal.

It’s not.

We are the weird ones. Those of us paying attention to AI are in the VAST minority.

The video cites recent A16Z data showing only 2.2% of American households pay for AI. Plenty of people tried ChatGPT years ago, got a rubbish answer and went back to the way they had always worked. Others still use it as a slightly smarter Google.

Helping others up to where you currently are is an extremely valuable skill. You’ll help people who are behind but also be able to (if you want) charge a good rate for this service.

Your edge is your history

Let’s say you work in recruitment.

Ten years in recruitment taught you what a good candidate looks like, which details people miss and what a client asks next.

You know how the work ticks. What’s important and what can be ignored. And where AI comes into all of this. That judgement is difficult for a generic AI trainer to copy. Sure they know the AI stuff but they don’t know the actual work. Their value is limited.

Let’s say you work in real estate/property. You go into a property company and do some exercises with them. Someone in property will learn more from an exercise about replying to a buyer's enquiry than a demo about writing pirate poetry. You know what a good reply looks like in your industry. You know when the output is a bit crap. You can explain the mistake in language the room already understands.

In this example a beginner types reply to this enquiry and gets four weirdly enthusiastic paragraphs, plus a discount code that does not exist.

They come to you and they say “AI is crap”.

Nope. The answer they got from AI was crap. But that is (in part) because they gave crap instructions.

Your job is to ask your student what a useful reply would do. What does that look like? Do they have examples of past replies? What can we extract from those?

And then we use this to write out proper instructions for the AI and then check the result together. Did it ask the right questions? Did it make anything up? Can the student correct it without starting again?

Sounds extremely basic. And it is. But for 95%+ of people out there this is where we start with them.

Know enough for this room

You know a lot more than the average Joe about AI. But we need to formalise that knowledge so that you can teach. We need to refine your skill and knowledge so you can be genuinely helpful to others.

There is still a standard. One YouTube video is nowhere near enough preparation for selling yourself as an AI expert. We have enough of that already!

You need enough knowledge to teach the task reliably: do it yourself, explain it clearly, help when AI goes wrong and know its limits. This is the basic sanity check I run before teaching anything. You’d be surprised how many don’t!

Start with people who know you

How do you get your first workshop?

Not by building a large social media following - that’s a terrible way. Believe me! It’s far too much hard work.

Nope, start simple. Send out some messages - text, Whatsapp, LinkedIn, whatever.

I've been using AI to help with [a task relevant to their industry]. Are people on your team using it? Would you like a short practical session?

That’s it. Super simple.

Your first room can be a former client, a team you used to work with, a professional group or a local co-working space. Five people around a table is enough. You are there to facilitate practice in a small room.

This is to help you get your sea-legs and practice your material. To learn what sort of questions people ask. To be prepared for those and to make the experience valuable for your students.

The full video covers what to teach, how workshop pricing works, how to run the first small session and how to approach the first client.

To the Task,

Kyle