Thinking About Offering AI Coaching — What Would You Want Help With?

[bizvids]

Hi everyone!

I’ve done several courses on AI implementation and have put AI to work in my own business, saving me a lot of time. Now I’m thinking about doing some AI consulting to help other business owners.

My question is this…

Would any members be interested in some 1 : 1 coaching on using AI? If so, what would you like to know about?

I’m not trying to sell anything here, just trying to gauge interest.

Thanks.

 

[fisicx]

No thanks.

Almost all the AI advice I have sought has been flawed in one way or another.

AI does have its uses but there should always be a caveat: check, check and check again. Never ever take the advice of anything AI tells you without verification.

 

[GLAbusiness]

Not for me, thanks.

Have you thought about this topic:

Make sure you ask what you mean

Disclaimer: This post is the result of a collaboration with Claude. It is a distillation of my own opinions, in my own words. It took 11 prompt iterations to get here — which rather proves the point.

---

Why does an LLM always give you the same number when you ask for one between 1 and 100?

It's not a bug. It's a semantics problem.

Here's what's actually happening:

User: "Give me a number between 1 and 100"
User expectation: a random number
LLM logic (paraphrased): "Aha, I've seen this question many times. It's probably best to be consistent in my answers. Let me check what I said last time. The answer is 73."

The LLM isn't malfunctioning. It's answering the question you asked, not the one you meant. "Give me a number" reads as a request for a preference — and LLMs are rewarded for consistency.

This unlocks something important about working with AI:

Precision is iterative — you can refine tone, format, and content as you go
Semantics needs a moment's thought upfront — are you asking for an opinion, a fact, or generation?
Ideally, LLMs need to improve their semantic parsing — the burden shouldn't be on the user at all

In the meantime, "good enough" is an underrated standard. A sufficiently unpredictable number serves almost every real use case. Perfect randomness is a mathematical concept; what most people actually want is an unbiased result.

The same applies to "what's your favourite colour?" vs "pick a random colour." Same surface structure, completely different intent.

And the 42 answer? That's Douglas Adams' fault. "42" is so over-represented in training data as the archetypal answer to a numerical question that it has become the least random number an LLM can produce. The joke has eaten itself.

The ideal fix isn't to burden users with prompt engineering theory. It's for the LLM to silently recognise the ambiguity and resolve it — returning an unbiased result without explaining why it can't be truly random.

Semantic analysis of a prompt should be the LLM's job, not the user's.

#AI #PromptEngineering #LLM #ArtificialIntelligence #UX

 

[Paul Kelly ICHYB]

@bizvids your approach is a common solution to a problem that doesn't exist.

The problem isn't that people don't know much about AI, but they do not know what to use it for - read some of the many recent threads that highlight that.

You need to identify a problem, describe how AI can solve it, then explain how to do it.

 

[Data Swami]

Hi everyone!

I’ve done several courses on AI implementation and have put AI to work in my own business, saving me a lot of time. Now I’m thinking about doing some AI consulting to help other business owners.

My question is this…

Would any members be interested in some 1 : 1 coaching on using AI? If so, what would you like to know about?

I’m not trying to sell anything here, just trying to gauge interest.

Thanks.

Quick question in relation to your background are you already involved and part of the coaching world?

Also getting a decent understanding of your experience of AI would be helpful. You say youve done a couple of courses what were they in relation to? AI in its business applications is wide and varying so a "coaching" setup would need that wide viewpoint. What i mean by this its not just ChatGPT vs Claude its everything inbetween.

Having been doing AI implementations alot its very much never a conversation about AI its a business problem over everything and whether certain things fit rather than AI for AIs sake. I will always be truthful and state where and when AI should be used for a client in their business processes.

So from my viewpoint AI coaching sort of still sits in that business coaching frame rather than a standalone coaching. Unless of course you are more going into the AI coaching side of things for other implementers. There are quite a few Skool communities from those who are creating youtube videos on creating n8n workflows etc. Theyve done pretty well out of it but its a constant content stream they have to produce ensuring they have free resources to offer while keeping their more tailored support in their paid coaching ecosystem. I was part of a few and I wasnt a huge fan as it missed alot of the business problem focus and more on the oh look at this cool tool we've built. But it is a good area to learn actual building skills.

And one final point is not to fall for the way that the AI Charlatans go about things either. They have given this industry a bad name with their prompt to change everything or one GPT to rule them all. The new age snake oil salesmen