AI for independent professionals: staying the expert in the room as the tools get smarter

Augmentation, and where Anna draws the line

77, writes a lot, and she has built a polished system around it. When she publishes a blog post, an automation in Zapier picks it up, sends it through ChatGPT to draft three or four LinkedIn posts, and drops those into Buffer, where she schedules her content. The drafts never go out as the machine wrote them. As she says, they are not good enough, even with all the training in the world, because they are not how she would write it. The draft is a starting point. She edits it into her own voice before anything is published. That is an important distinction to hold onto.

• Augmentation means using AI to raise your game on work you still direct. 
• Automation means handing a task over entirely. 

Anna automates the genuinely mechanical parts, she even lets ChatGPT pick the category for a post, which it gets right about ninety per cent of the time, and that is good enough for something she no longer wants to think about. The work that carries her judgment and her voice, she keeps.

Why this favours people who already know things

Information has always been around, but it is everywhere now, and it is easy to feel you are drowning in it.
Caleb Scharf, in The Ascent of Information, puts it this way:

We are literally drowning in the raw material. Turning it into something that actually means something is the harder, more valuable move, and that is the work your experience equips you for.

Experience is an advantage here. The professionals getting the most from AI are the ones still reading, taking notes, and thinking, because that background is what lets you tell good output from generic output and steer the tool toward something useful.

Anna sees a related advantage in the material experienced people have already built up. She has been a note-taker her whole life, and what AI does well, she told me, is sift through years of collected notes and surface two things you read five years apart that turn out to be connected.

Margareta Krizova, a lifelong entrepreneur near Prague who joined me on episode 90, works the same way.
She co-founded a boutique advisory firm in 1993, spent years in mergers and acquisitions, and now mentors founders. She loves AI for brainstorming and is careful about where she draws her own boundaries, particularly around personal and financial data.

This is where your years of experience are valuable. Over a long career your mind quietly stores patterns from everything you have read, everyone you have worked with, and every problem you have had to sort out. So when something new lands in front of you, you have an advantage, because your brain has more associations to connect it to, more situations you have already lived through, and that's what helps you see quickly what matters and what to ignore.

Karl Fast and Stephen Anderson, in their book Figure It Out: Getting from Information to Understanding, help shape how I think about information. It's cheap and easy to access. Understanding is the hard part. The cost of getting from one to the other is far lower for someone who already has something to relate the new information to. AI hands you more information, faster. It does not hand you understanding.

You have spent years building a store of patterns from your experience and life, and that's what helps you turn what AI gives you into understanding quickly. Without it, the same output is just more material you have no way to evaluate and judge.

The thinking does not get handed over

Filip Drimalka goes furthest on this. In episode 67 he described helping organisations in the Czech Republic adopt new ways of working, and he has written a book The Future of No Work. 

He told me that thinking as we know it has ended, then immediately named the catch: it will still be you who is orchestrating the discussion. He uses AI as an extension of his own mind, clicking into his favourite tools at every step,
but he is always the one directing it.

When his team spent two hours stuck on how to improve an AI assistant they were building, he loaded the whole conversation into a research tool, worked through the ideas, and rewrote the instructions himself. The machine sped up the handling. The decisions stayed with him.

Margareta does something similar with a twist. She does not ask AI to hand her answers. She uploads the work of a businessman she admires, Keith Cunningham, and asks it to ask her the questions she would not think to ask herself.

Her idea about using AI, in her words, is really about asking questions. I find myself doing the same thing more and more: nutting through a problem with it rather than asking it to solve the problem for me. In writing this article I have spent hours going over paragraphs, seeking to deepen the understanding, drawing on notes that help illustrate with a bit more depth the points I am trying to make.

Peter Hatherley, who joined me on episode 54 and founded the AI company Authored Intelligence in Christchurch, gives this a useful vocabulary. He separates generative AI, where you fire off a prompt and take whatever comes back, from interactive AI, where you work a problem through with it the way you would with a consultant.

He treats it as a savant, a brilliant generalist he can reason with, rather than a servant taking orders. He calls it a co-architect, something that asks whether you really want this, whether you are happy with that, and gets into the detail with you.

All three are doing the same kind of work. They forage for raw material, sort what matters from what does not, and use AI to get their thinking out in front of them where they can see it and shape it. This is what I mean when I describe AI as a cognitive resource rather than a replacement. The tool handles the raw material. The call about what is worth keeping stays with you.

Where it quietly goes wrong

There are two ways this comes unstuck, and they are worth naming.

The first is wasted effort dressed up as progress. I run a marketing services company, and I have created and managed websites for over twenty years, so I am not shy with these tools. One afternoon I spent hours pushing AI to write the schema markup that helps a client site work with Google, testing it, watching it fail, yelling at it to do better.
In the end I found a $100 plugin that did the job perfectly in a few minutes for all my sites.

Jon Younger, who first came on episode 12 in 2023 and returned on episode 79 to talk about exactly this, has a name for the pattern. There is a stupid use of AI and a good use, and a lot of the early effort goes into the stupid kind, because it feels productive while it is happening.

The second failure is more serious because you do not notice it. Lean on AI for the hard thinking and your professional instinct slowly dulls, a kind of intuition rust that comes from not doing the difficult parts yourself. Jon worries about what this does to younger people.

The path that once may have taken twenty years of experience to build now gets short-circuited, and managers are turning to AI instead of handing junior people the problems they would have learned from. Filip sees the same thing happening in the organisations he works with.

For experienced professionals the danger is quieter, but it is real. Research has found that people given a high-quality AI tool can become careless and make worse decisions than those given a weaker tool or none at all, simply because they stop paying attention. The protection is to stay engaged with the work that keeps you sharp.

This is also why voice matters more than it used to. Anna's point is that anyone can generate a LinkedIn post in seconds, and it will read like everything else already online. The people who stand out are the ones who can edit, who have an opinion, who sound like themselves. Peter builds his tools around the same instinct. We all need to proof, we all need to edit, and the common mistake is throwing the output out without looking at it.

A simple way to decide what to hand over

Ben Legg, the former chief operating officer of Google Europe who now runs a portfolio career, explained how he approaches this in episode 87. He divides work into two kinds.

• Hamster work is running inside the machine, turning A into B as part of someone else's process
• Designer work is building the machine itself

As AI takes on more of the hamster work, he argues, the value shifts to the people doing the designing. The diagram below sorts your own tasks along that line.

Before you automate a task, the question to ask yourself is whether doing it yourself keeps your instincts sharp. If a task is genuinely mechanical, sorting a clogged inbox of newsletters, drafting a first pass, formatting, picking a category, hand it over and reclaim the time, the way Anna has.

If a task is where your value actually lives, the moment where you say that is not the real problem, the call about what matters, the writing that sounds like you, keep your hands on it and use AI to pressure-test your thinking rather than do it for you.

Filip put the whole thing into one line. It is about using AI for your own expertise. The mindset matters more than the tool you pick.

It is worth saying that the centaur and cyborg labels, which a few guests mentioned from Ethan Mollick, point at the same truth. Whether you dip into AI occasionally or work alongside it all day, what matters is that you stay the one orchestrating the work.

A last thought

An interesting theme runs through all of these conversations. None of these people worked AI out entirely on their own. Filip runs programs and teams. Margareta runs a business accelerator where founders can draw on mentors in finance, marketing, AI, and legal. Jon built an AI business with his son and sits on advisory boards. Peter tests his tools against a community. The experienced professionals getting the most from AI tend to have someone to think with, a peer or a mentor or a partner who asks the questions they are too close to ask themselves.

I wanted to mention this because advice like keep your judgment sharp can sound like a solo project, and in practice it rarely is.

Frequently asked questions

Books referenced
Stephen P. Anderson and Karl Fast, Figure It Out: Getting from Information to Understanding (Two Waves Books, an imprint of Rosenfeld Media, 2020)
Caleb Scharf, The Ascent of Information: Books, Bits, Genes, Machines, and Life's Unending Algorithm (Riverhead Books, 2021)