The Autocomplete Leader

LLM's are just probabilistic word generators, but how often are we so different?

Understanding that LLMs function as probabilistic word generators - assessing word-by-word what word most likely comes next - is practically useful.

It helps explain why, with a little practice, we can easily spot AI generated text: it sounds like the literal average of everything you've ever read.

It also helps build intuition for when using these tools can be beneficial and when it is not. If you're trying to organise incoherent thoughts, get a quick topic overview or to assess a document compared against a prescribed standard - Mr. AIChatbot is your friend. If you want to forge unlikely connections between concepts, develop useful analogies, or convey the visceral feeling of a moment, we're better keeping the pen ourselves. Mr. AIChatbot has never felt the rain on its ankles.

Knowing this, we can see how AIs stringing together single word after single word on the basis of mathematical operations can easily create the illusion of intelligence. Inside, we may feel some relief that our biological superiority is not threatened.

Recently, however, I've been thinking about the other side of the equation. As humans are we always so different? Is our conscious stream of thought - and the ways that we respond to the world, to each other, to pressure - as different as we would like to think?

How much of the time are we operating from learnt scripts – learned patterns of behavioural, emotional and language responses?

And what does that mean for leaders and organisations grappling with change?

Are humans so different?

Stephen Covey opens The 7 Habits of Highly Effective People with a moment from Viktor Frankl's life. From 1942 to 1945, psychiatrist Viktor E. Frankl was captured and held in Theresienstadt, Auschwitz, Kaufering (a Dachau satellite camp), and then Türkheim before finally being released. He published his account in Man's Search for Meaning, which he wrote in 9 days in 1945.

As we can (probably not) imagine, there he suffered indignity, deprivation and humiliation. There is a moment where Frankl, as he is naked and alone in a bare room, realizes something. He realizes that no matter what the guards can do to him, it is entirely in his control how he responds. That in between stimulus and response there exists "the last of human freedoms."

It also made him ask: for the rest of the time — the ordinary days, the unremarkable moments — how much are we really on auto-drive? How much of what we say, think, and feel is not chosen but simply repeated — the accumulated average of everything we have experienced and been conditioned by before?

The Corporate Autocomplete

We each have our own default operating system - shaped and moulded by our deep experience, formative environments, early professional conditioning and learnt survival behaviours. Neurons that fire together wire together. They determine how we see others, how we interpret situations, our assumptions about the world, our emotional reactions. They are operationalised through scripts, long and short, that we fall back on to navigate through the day. Most of the time they are invisible, at least to ourselves.

These ready responses and scripts play a valuable role. They manage cognitive and emotional load, protect self-esteem, provide a working map of the world. Without them, the ordinary business of getting through a day would be exhausting.

Stress accentuates this. Think about the last time you felt stress, or the need to perform well in front of others. Chances are, you had to battle against a familiar stress response pattern you recognised. Speaking more, speaking loudly, not speaking at all. Under pressure, uncertainty, or threat - our brains look for something to fall back on. Whether the script that surfaces is the corporate 'party line' or some personal defence mechanism the result is the same.

The consequence is that to others, and sometimes to our later selves, we can seem absent. On autopilot. In truth, we are dealing with something other than what is in front of us.

A few weeks ago I wrote about the 'Mona Lisa' team - smiling through discussions of how AI could replace 80% of their current jobs. In that moment of existential anxiety, the last thing people in that room need is an autocomplete corporate response from the front of the room. Or worse still, an emotionally-driven response born of the manager's own fear dressed up as authority. The people in that room don't need more information, the latest strategy talking points or 'strong leadership'. They need to know someone is actually in the room with them.

We worry about AI-generated responses. Yet most people can recognise an auto-generated corporate response instantly. The language is different, but the feeling is the same. Nobody is actually there.

Getting out of the river

Performance anxiety in response to uncertainty is normal. The body and mind will surface patterns from the past. Response patterns that once worked, served up again whether they fit or not.

Karl Steptoe, a sports performance psychologist who has worked with elite golfers and Premier League footballers for over a decade, builds his research and practice on this observation. His approach is not to try to deny or replace negative thoughts and emotions, but to change our relationship with them. Humans have the capacity for metacognition – we can move from being the commentator, dragged along in the current of our own reactions, to being the spectator, watching from the riverbank.

These ideas are not new. Zen Buddhism and more recently mindfulness practice are based on the same simple insight. "Notice what arises without being carried away by it" is the way the founder of the particular flavour of Zen I practice put it, way back in 1227.

The velocity of change today places leaders under intense pressure, meaning their pre-programmed default responses are more likely to be triggered than ever before. Just when it is the last thing their teams need.

The difference between an LLM and a human is not that we don't run on patterns.

It's that we can notice we're doing it. And in that noticing – Frankl's gap, Steptoe's metacognition, Dogen's instruction – we create a gap where we choose how to respond.

I've been enjoying a segment on a local radio show. Every Monday the presenter asks the same two guests the same four questions: what delighted you over the weekend, what annoyed you, what did you learn, what are you looking forward to this week. I've started asking myself the same. And from now I'll add a fifth.

Which moments this week do I definitely want to be present for?