There feels to be something viral about AI, as it finds its way into the recesses of our lives and organisations, evolving in response to what it finds, feeding on vulnerabilities, in our systems and in our assumptions. It is, I think, unlike any other technology we have developed, a chimera, technology combined with biological characteristics that evolves on its own terms.
And I think it requires a good deal more curiosity regarding how we deploy it.
Our attention has been focused on the drama - the potential for changing the workplace. For investors that translates into labour cost reductions, and for organised crime (which shares many of the characteristics) an unparalleled opportunity to find its way through our security protocols. On the battlefield and on the high street, recognition and analysis make us individual targets. A bullet as well as an advert can have our names on it.
At the same time, beyond the investors, it promises benefits from personalised medicine to personal assistants that can give small teams capabilities that bely their size.
Alongside what is being trumpeted, we are facing a swarm of unintended consequences that will change the workplace, but probably not in the way those of us brought up in traditional organisations expect.
I have written previously that I think using AI effectively is a craft, and I maintain that view, even as the nature of the craft continues to evolve. One of the challenges we face is that used without care, it can take us on a high speed trip to average. Used in areas where we do not have deep knowledge, it will fill in the gaps in our brief to it with what it has available. It is not just the irritating verbal slop it produces, which can vary from irritating to dangerous when the errors we do not spot give us us those unintended consequences, it is the consequence of a crisis of curiosity and imagination.
Imagine a company full of people under pressure to produce results quickly all using AI to speed their workflow without considering carefully what they are asking for, or taking the time to check what is offered back, not just for accuracy, but for quality of communication.
In a conversation with Steve Done on Monday, he did what AI is unlikely to do, by using his deep knowledge of wine to identify an unlikely parallel: In the late 1980s the wine trade invented the “flying winemaker”. Young Australians, idle during the northern harvest, were flown into French co-operatives to turn cheap grapes into modern, saleable wine. They brought new thinking, cleaned up faults and brought in hygiene and temperature control. They also brought a particular style with them, and by the 1990s the flying winemaker had become the consultant.
The most sought-after was Michel Rolland, (who died in Bordeaux in March this year). His clients hired him because he could make wine that American wine critic Robert Parker would score highly. Parker was the American critic whose 100-point scores could make or break a price, and Rolland became known as “the Parker whisperer”. Jancis Robinson described what followed as a dramatic rise in the standard of everyday wine, with “a worrying tendency towards standardisation”. Both halves matter, indicating what we are prepared to accept for the sake of a score. The evidence that Parker’s scores moved prices is strong, but what matters is how it worked: an external adviser, a critic who replaced judgement with a score, and an industry that learned to make what the metrics rewarded.
AI now puts both roles in one system. The same tool that drafts the work and is then asked whether it is any good, )and there is evidence that models recognise their own output and rate it more highly). Rolland was one man, and he had to visit each cellar in turn. AI visits every client at once.
We see similar trends in recruitment, where AI has become originator and arbitrator, and talented individuals are using templates that hide the very qualities that make them different, in search of achieving a score. If we permit AI to homogenise us, the market will make talent a free gift wrapped up in an undifferentiated résumé acquired at commodity rates.
The wine trade did find its way back. After phylloxera the French built the appellation, a legal boundary that protected what markets alone could not. A generation after Parker, the vigneron who knows one vineyard intimately is the ideal again, not the consultant. Correction came from the people closest to the vines, not from the score improving itself.
Love it or hate it, AI is not going away. For a while it offers those who use it with care, beyond the obvious, a chance to thrive, as artisans always have with new tools. But the boundaries that protect difference will not build themselves.
So what is the appellation for your work, and who is going to build it?