AI in the Background, Humans at the Table.

By Bruce Lloyd, The United Kingdom and Frank Peters, France.

I arrived with my head full of programming. Bruce arrived with questions. Somewhere in the untidiness between the two, we found something worth talking about.

I walked into my conversation with Bruce badly prepared.

Not deliberately. I had simply been engrossed in something completely different. For several days I had been building and testing a piece of software around ChatGPT, trying to turn a complicated prompt into a practical tool for finding the right people for future Brida Tables.

The night before, I had asked it to research organisations across eleven European cities in nine countries. It came back with 210 organisations and 768 people.

That had rather occupied my mind.

Bruce, meanwhile, had been doing almost the opposite. He had been asking AI about literature, philosophy, creativity, God and the future of humanity. I had meant to read more of what he had published before we spoke.

I hadn’t.

So there we were. Bruce with what he once called his “intellectual indulgence”. Me with my programming. No agenda capable of holding the two together.

Which, as it turned out, was rather the beauty of it.

Deep, but light.

We began somewhere else entirely. Bruce had noticed more personal stories appearing on LinkedIn — people talking about illness, ageing, family pressures and the ordinary difficulties of getting through life. There seemed to be more human pain on view.

I wondered whether there was actually more of it, or whether technology had simply made it more visible.

Bruce made a distinction I liked. The macro problems around us — AI, politics, power, social change — may have become so enormous that people are beginning to switch off from them. Meanwhile, the micro problems of ordinary life remain stubbornly close and increasingly demanding.

We live surrounded by conversations about transforming humanity while the person next to us may simply be trying to survive Tuesday.

Bruce once described our discussions as “deep but light”. That feels about right. We are not trying to win an argument. Quite often we are trying to discover what the argument is.

Eventually I told him about my programming.

What interested me was no longer simply what ChatGPT could produce. I had begun taking some of its work to Claude, Grok, DeepSeek and Kimi and comparing the responses. They reacted differently. Some were more critical, some more neutral, some almost embarrassingly enthusiastic.

I started wondering whether there were cultural fingerprints inside the machines.

Bruce’s response was simple: ask them.

Ask the models to examine their own biases.

That is one of the things I enjoy about talking with Bruce. I arrive with a half-formed observation and he quite often gives it back to me as another question.

He has no particular need to close the loop.

He opens another one.

Arguing with the machine.

Something similar had started happening between me and ChatGPT. I was arguing with it. If its analysis of a country or business culture contradicted something I knew from experience, I told it. If I found a problem in the programming, I challenged it.

Sometimes it corrected itself. Sometimes it explained why it had gone wrong. Sometimes it effectively said: yes, you have found something I missed.

It is a peculiar experience. You know perfectly well that you are not talking to a person, yet the process begins to resemble a working relationship. You propose something. It responds. You disagree. It adjusts. You ask another question.

Bruce picked up on that immediately.

“It stimulates you to ask further questions.”

Exactly.

Perhaps that matters more than getting the answer.

I tried to explain the picture forming in my head as a Venn diagram. On one side was my very practical use of AI: solve a problem, find people, create something useful. On the other was Bruce, feeding literature and philosophical questions into different models and exploring what came back. Then there was ordinary human experience.

Where did they overlap?

Bruce said something that brought the whole thing into focus. Across many of his experiments, however different the starting point, a similar conclusion kept appearing: technology is moving faster than our ability to manage it.

So perhaps the real problem is not simply what the technology can do.

It is whether we are developing quickly enough to handle what we have created.

Then Bruce put it better.

“AI is pretty good at telling humans what they ought to do to be better humans. But it can’t do it for us.”

There it was.

AI can compare, calculate, summarise, provoke, challenge and retrieve. It can show us another way of looking at something. It can even tell us where our reasoning may be weak.

But eventually a human being has to decide what happens next.

It cannot make us curious. It cannot make us decent. It cannot make us listen. It cannot make us change our minds.

And it cannot make a conversation good.

We still have to do that.

Perhaps slop is human.

From there we wandered, naturally enough, into AI slop.

I mentioned the mountains of machine-generated rubbish now being produced. Bruce pushed back. Imagine, he said, a shelf full of books about communication. Ask AI to extract the essential learning and perhaps it comes down to a page.

What, then, are all the other pages?

I hadn’t followed that thought through before.

Humans have been recycling material for centuries. We take an old idea, put a new cover around it, build a framework, sell a course and call it fresh thinking. Nobody called it AI slop because AI wasn’t involved.

Perhaps slop is not really a technology problem.

Perhaps it is a human one, now operating at greater speed.

And that brought us back to systems.

I told Bruce about a poor sales approach I had received. The offer was such a bad fit that, instead of deleting it, I wrote back and suggested a conversation about why their system had selected me.

The woman initially agreed.

Then her system dragged her back. Her job brief would not allow her to have the conversation.

There is something gloriously absurd in this. We worry that machines are becoming too human while constructing organisations that stop human beings behaving like humans.

Systems tell us what to say, who qualifies, what counts, how long something should take.

And somewhere in all that efficiency, conversation disappears.

Bruce wondered whether AI might actually help humans have better conversations.

That sounds contradictory until you realise that AI does not have to run the conversation. It can simply sit somewhere in the background.

Which, of course, is exactly what had happened with us.

The AI wasn’t feeding Bruce and me questions. It wasn’t controlling the discussion. I had arrived unprepared. Bruce followed his curiosity. We interrupted each other, changed direction, dropped threads and picked them up twenty minutes later.

AI had influenced what both of us had been thinking before we arrived.

But the conversation was ours.

The usefulness of chaos.

Towards the end, Bruce laughed about the job of turning one of our wandering discussions into something coherent. We had jumped around enormously.

He was right.

But I am increasingly reluctant to see that as a defect.

There is a temptation, particularly with AI, to optimise everything. Define the objective. Design the prompt. Remove deviation. Produce the outcome.

Useful, certainly.

But that is not necessarily how a good conversation works.

A good conversation has friction. Somebody misunderstands. Somebody asks an awkward question. A practical discussion about software suddenly becomes a conversation about philosophy. AI slop becomes a discussion about human publishing. A bad sales email becomes a question about whether systems are making us less human.

There is chaos in that.

There is also discovery.

Just before we finished, Bruce suggested one final experiment.

Why not give the conversation to AI and ask it how we could have done it better?

So I did.

Its answer was irritatingly sensible.

The strength was the curiosity. Neither of us had arrived defending a finished position. Ideas collided. We challenged each other without needing a winner.

The weakness was almost the same thing.

We opened too many doors.

Interesting thoughts appeared and disappeared before we had really tested them. Its suggestion was not to script the conversation — thankfully — but to give the chaos a few handrails. Start with one or two questions worth returning to. Stay with the strongest friction a little longer. Occasionally stop and ask what we have actually learned.

I can live with that.

AI in the background.

There is a nice irony to the whole thing.

I arrived badly prepared because I had been spending too much time working with AI. Bruce and I then spent an hour trying to understand what AI might mean for humans. Finally, we asked AI how the humans could improve the conversation.

Perhaps we are disappearing up our own philosophical backsides.

Or perhaps something more useful is happening.

The more I work with AI, the less interested I become in whether it is simply good or bad.

I am much more interested in what happens to us while we use it.

Does it make us lazy, or does it make us ask better questions? Does it replace experience, or make experience more important? Does it remove friction, or help us find the useful friction? Does it make conversation unnecessary, or give two people more interesting things to talk about when they finally sit down together?

Bruce and I didn’t answer those questions.

I am not sure we were supposed to.

I arrived with my head full of programming. Bruce arrived with his questions. We spent an hour trying to make sense of the chaos between them.

AI stayed in the background.

Two humans sat at the table.

And talked.