Before we let AI take over missiles, I’d rather we figure out how it should restock ketchup.

Most concern of AI taking control over “making the decision” is about matters of life and death. Targeting weapons, deciding which way a car should swerve to avoid an accident.

Fair. But the AI decisions actually showing up in business today are smaller. And, as an industry, we’re still getting them wrong.

Take more mundane situations where AI can automate a large portion of tedious tasks done manually today.

Retail for example. Working at a grocery store to locate out-of-stocks or misplaced shelf tags and products is done today by eyeballing and barcode scanning. Similar by-hand activity for designing shelf assortments for what to carry and how to have enough of the popular items without over-investing in the wrong stuff.

I am seeing up to 70% of such tasks, using AI agents, being automated. The issue is how to set it up and how you know it is making the right decisions. And, where it is a “team effort” between AI and people, how do you add value without getting in the way.

This requires a combination of change management, how to get people to embrace the role of AI in changing how they work, and good governance, how to have guard rails to avoid AI ordering 100 cases of what is already in the backroom.

There are good answers to this. AI can independently audit the AI. Good design for AI can avoid getting in the way of experienced workers while super-powering junior employees. Good AI can absorb corrections and do better in the future. 

I am witnessing these techniques working well in practice, crucial for people to build trust in AI, to welcome and embrace rather than fear.