AI Isn’t the Hard Bit

We’ve been having a lot of conversations with expertise-business owners about AI lately. Founders, consultants, established experts trying to work out what this moment means for them and how to respond to it.

There’s a familiar arc to most of these conversations. Someone shows us what they’re doing with ChatGPT. They mention a tool they’ve been trialling. They talk about the workflows they’ve started to streamline and the ones they haven’t worked out how to crack yet. There’s excitement, usually. And underneath it, almost always, a quiet worry that they’re not moving fast enough.

But the longer we sit in these conversations, the clearer something becomes. The interesting story isn’t the tools. It’s what the tools are surfacing.

What’s Working

Plenty is working, and it’s worth saying so. Experts are using AI to compress hours of analysis into minutes. They’re running discovery calls through summarisers and getting back something useful. They’re drafting tenders faster, sketching proposals quicker, pulling themes out of feedback in a fraction of the time it used to take.

The more capable end of the spectrum is starting to play with proper automation. Tools like n8n and Claude Skills are making it possible to build workflows that would have needed a developer two years ago. People are getting genuinely excited about what they could build next.

And there’s a quieter benefit running underneath all of this: AI is exposing the cracks in the business. Workflows that worked when there were three of you and a handful of clients are starting to creak. The act of trying to automate something forces you to actually map it. And most people, when they map it for the first time, realise it’s a mess.

That’s not a bad thing. It’s the work that should have been done years ago.

What’s Hard

There’s a less comfortable side to these conversations too.

Inside teams, the gap in capability and attitude is wide. Some people are leaning in hard. Others are quietly nervous about what it means for their role and are hoping it’ll all blow over. Most owners we speak to are still figuring out how to bring everyone along without forcing it.

There’s hesitation about scaling automation beyond simple tasks, and it’s usually traceable to the same thing: people can’t see what the AI is actually doing. A workflow they can’t observe is a workflow they don’t trust. The built-in AI features bolted onto existing platforms tend to suffer from the same problem — they help with surface tasks but never quite earn enough confidence to take on anything that matters.

Meanwhile, the world outside is moving. Tenders are starting to ask whether AI was used in the response. People are putting that question into AI to work out how to answer it. The irony writes itself, but it points to something real: the relationship between expertise and the tools that mimic it is becoming a live commercial issue, not a theoretical one.

What’s Genuinely Ugly

Then there’s the stuff that, frankly, is just a mess.

Subscription bloat is rife. Most of the experts we speak to are paying for somewhere between three and seven AI tools, each one bought to solve a specific task, none of them really integrated, several of them barely used. The monthly fees mount up and the return is fuzzy.

We see whole teams sharing a single ChatGPT account, with the AI cheerfully confusing one person’s coaching client with another person’s puppy training plan. Context is held together by prompt history rather than by any deliberate structure — which works fine until the conversation gets long enough that the model starts forgetting, and then everyone is hunting through old chats to reconstruct what was decided.

And almost everyone has a story about investing serious time setting something up, only for the tool or the model to shift before the benefit lands.

The Pattern Underneath

If you step back from all of this, the same thing is true across every conversation.

AI isn’t the hard bit. The hard bit is the thinking, structure, and discipline it forces you to confront.

The experts struggling with AI aren’t struggling because the technology is too complex. They’re struggling because the technology is asking them to be clear about things they’d been comfortable being fuzzy about. What’s our process, actually? Whose voice is this? What do we know that nobody else knows? Where does our value actually live?

These were always the right questions. AI just makes it harder to keep avoiding them.

The ones making real progress aren’t the ones with the most tools. They’re the ones using this moment as a forcing function — to map their workflows, codify their thinking, get clear on their voice, and build foundations that the technology can actually amplify.

Because that’s what AI is. An amplifier, not a magic wand. It makes whatever is underneath it louder. If what’s underneath is sharp, clear thinking and well-structured expertise, the amplification is extraordinary. If what’s underneath is mess, you get amplified mess.

This is the work we’re doing with Expert OS clients right now. If that’s the kind of work you want to be doing with your own expertise, join the waitlist.