I first wrote about hidden risks in strategy execution in my 2017 blog, Essence of Decision and the Cuban Missile Crisis, from which I drew the following lessons:

• Organisations have a creed that motivates their people – think Nike’s ‘Just do it!’.
• The picture at the top of an organisation is distorted by the way information has been processed on its way to the top.
• Ingrained systems of work affect HOW strategic initiatives are implemented.

These collectively ensure there are going to be unintended consequences when any strategic initiative plays out. Both good and bad. And when it comes to AI, unintended bad consequences can be a frightening prospect 😬

How should you think about these three lessons for your AI deployment?

• Your organisation’s creed. If you are a can-do organisation, the challenge is being gung-ho when the risks are so much less known and understood. If your creed is slow and steady, the fast pace of AI could leave you out of the game, perhaps permanently.
• The distorted picture. AI distorts information. Worse, even its creators can explain only part of how it reaches its conclusions. As James Kavanagh, founder of AI Career Pro, said at the recent RMIA National Conference words to the effect, “AI models are black boxes made up of black boxes.”
• Ingrained systems of work. AI isn’t implemented in a vacuum. It is absorbed into existing workflows, systems, incentives and controls. And those legacy ways of working may matter more than the technology itself.

So, the real challenge of AI deployment is not just choosing the right tools. It is understanding the organisation those tools are entering. Culture, distorted information and ingrained systems of work will shape what AI becomes in practice. Ignore them, and the risks will not be accidental. They will be built in.

My tip: Develop mental models of each key area of the business so you can better understand what makes that area tick. If you would like some help with that, here is Chapter 5 of my book, Team Think: How Teams Make Great Decisions in which I cover this approach in some detail.