Find the psychos

Every new technology freaks most people out. Not everyone.

There is always a small group who see the new thing and light up. They want access, they want to pull it apart, try stupid things, break it, rebuild it and work out what it makes possible.

Find those people.

I call them psychos.

Iconoclasts, psychos, innovators, missionaries, champions, early adopters — I use the words differently, but for this argument they're synonyms.

They are energised by change while everyone else is still working out whether they like it.

These are the people you want doing AI. They were also the people you wanted doing cloud, mobile, the web, SaaS and whatever came before that.

This isn't really an AI idea. It's a change idea.

There is research behind it too.

Eric von Hippel wrote about "lead users": people sitting ahead of the market who experience needs earlier than everyone else and often invent solutions themselves.

Howell and Higgins studied "innovation champions": people who push new technologies through organisations. They found that these people networked more, influenced more and moved ideas through the organisation.

So find them, then concentrate them.

There is a temptation to spread innovative people evenly across a company. One per team, with a little bit of innovation everywhere.

I think that can be a mistake.

People doing strange new things need other people doing strange new things.

Research on technology startups has found that physical proximity increases knowledge spillovers. Studies of elite scientists show similar effects: good people make the people around them better.

Concentration of talent is good.

Give them hard problems and room to work. Then make them prove things.

The danger isn't that the psychos exist. The danger is that today's psychos become tomorrow's priests.

Every accepted idea was weird once. Every bit of organisational canon started as somebody saying: what if we stopped doing it that way?

James March wrote about this in 1991. He described the tension between exploration and exploitation.

Exploration is finding new things. Exploitation is getting better at the things you already know.

Organisations need both, but exploitation gets efficient very quickly.

The existing way has process, governance, training, metrics and people who know how to defend it. The new idea has some lunatic waving their arms saying they built something interesting last night.

That creates a problem.

The psychos can now generate new ideas much faster than the organisation can absorb them. AI makes this worse.

A small group can build a prototype, change a workflow, create an agent, test a process or automate something in days. Getting the organisation to accept that change can take months.

So the bottleneck shifts from invention to acceptance.

Can we invent something better? Increasingly, yes.

The harder question is how quickly we can decide whether the weird new thing should become normal.

Some ideas will be shit. Kill them quickly.

Some will work. Prove them quickly, then turn them into canon.

Once they become canon, let the next bunch of psychos attack them.

That is the cycle.

Find the psychos. Cultivate them and concentrate them.

Let them challenge the current canon.

Then hand the new canon to the next psycho and ask what is wrong with it.