Essay

The Centaur’s Crossing

AI, organisational change, and the future of the supply chain planner

In May 1997, in a television studio in midtown Manhattan, a machine defeated the strongest chess player who had ever lived. Deep Blue took the match against Garry Kasparov by a single game, and the world drew its conclusion at once: human expertise was finished. The grandmasters would go the way of the lamplighters, and the rest of us would follow in due course. Newsweek had billed the match as “the brain’s last stand”.

Kasparov, who had more invested in the question than the columnists, drew a different conclusion. He had not lost, he reasoned, because the machine understood chess better than he did. It plainly did not. He had lost because it calculated differently: tirelessly, without vanity, without yesterday’s blunder sitting on its shoulder. The machine had no nerves to fail it.

For Kasparov the interesting question was no longer whether a machine could beat a man. That was settled, and sulking about it was a waste of time. The interesting question was what a man and a machine might do on the same side of the board.

I should declare an interest, because in this story I was on the machine’s side. In 2000 I joined i2 Technologies, then the reigning power in supply chain optimisation, and I spent the following years telling manufacturers across Europe that the optimiser would solve their problems, and then trying to implement it. I believed it at the time. It was, in the narrow sense, usually true: the maths was good, or got good eventually, once the data had been beaten into shape.

What the maths could not do was make anyone follow it. I have now spent the better part of thirty years watching the optimiser plan and the organisation do something else. The plan would be published on Monday; by Wednesday the planners were back in their spreadsheets, sales had promised a customer something the plan had never contained, and the steering committee was asking why adoption was low. We called this a change management problem, which is generally what you call something when you intend to deal with it later.

It took me an embarrassingly long time to ask the better question. What if we had framed the problem wrongly from the start? What if the unit of performance was never the machine, and never the human, but the partnership between them? Kasparov got there twenty years before most of my industry, and it is worth watching closely what he did with the idea.

In the essay

• Kasparov’s finding, and why the centaur lasted fifteen years in chess and will last longer in supply chain
• Chess is not planning; it is replanning. The decision, not the plan, is the deliverable
• Two true stories from thirty years of implementations: the ownerless decision and the thirty per cent forecast
• The centaur’s three moves: trust, overrule, and ask a different question
• What remains for the planner, and why it is a promotion