AI Strategy
Reversible or Irreversible: A Rule for Delegating Decisions to AI
A simple test for which calls to hand to a machine and which to keep.
Read
Back to Blog
AI Strategy
When Does AI Stop Being a Tool and Become the Strategy?
René joined InnoWave's "The Human Side of Innovation" to talk through the moment AI moves from an experiment on the side of the business to the thing the business is organised around, and how leaders can tell the difference.
Prof. Dr. René Bohnsack·Co-Founder, Zeroto100
July 4, 20265 min read
The Human Side of Innovation · EP2 · InnoWave. Hosted by Andreia Mitreiro (CEO, Your Trend).
In short
AI becomes strategy when it starts shaping which decisions get made and how, rather than sitting beside the work as a faster tool. In this episode, René offers three practical tests: judge AI by capability multiplied by direction, delegate decisions by whether they are reversible, and start every week by naming the one outcome AI should improve.
The conversation
InnoWave's Andreia Mitreiro asked a question most leadership teams are circling around without saying out loud. What happens when AI stops being a tool you reach for and starts being the strategy you run on? The full episode is embedded above. What follows are the three ideas from the conversation that are most useful on a Monday morning, when the slides are closed and someone has to decide what the team actually does.
Takeaway 1 — Judge AI by capability multiplied by direction
Most AI conversations inside companies are about capability. Which model, which tool, which vendor, which feature shipped this month. Capability is real and it is improving fast, but on its own it produces very little. A powerful model pointed at a vague problem returns a vague answer at scale. The value comes from capability multiplied by direction, and multiplication is unforgiving. When direction is close to zero, the product is close to zero, however strong the capability. This is why so much AI spend shows up as impressive demos and thin results. The demo proves capability. The business case needs direction, which means a specific decision or outcome the model is pointed at. Before approving the next tool, a leader can ask which decision it improves and by how much. If that answer is thin, more capability will not save it.
"Capability multiplied by direction. When the direction is close to zero, the product is close to zero, however strong the capability."
Takeaway 2 — Delegate by whether the decision is reversible
The hardest practical question is not whether AI can make a decision, it is which decisions you should let it make. René's rule is simple enough to use without a meeting. Sort the decision by how hard it is to undo. Reversible decisions, the ones that are cheap to correct if they go wrong, are exactly where AI should run with minimal supervision, because the organisation learns from volume and the cost of any single miss is low. Irreversible decisions, the ones that are expensive or impossible to walk back, keep a human accountable in the loop. This turns an anxious, all-or-nothing debate into a sorting exercise. Instead of asking whether you trust the model, you ask what a wrong answer would cost and whether you could recover from it.
Takeaway 3 — Start Monday by naming the North Star
The episode closes on a test that separates AI strategy from AI activity. On Monday morning, name the one outcome AI should measurably improve this week. If the team can name it, the work has direction and progress becomes visible. If the team cannot name it, what looks like a strategy is a collection of experiments waiting for a purpose. The value of the question is that it is uncomfortable in exactly the right way. It moves the conversation off which tools are being trialled and onto which result is expected, which is the only footing on which AI ever becomes strategic.
Why it matters
The three ideas connect. Direction gives capability something to multiply. The reversible test tells you where that direction can be handed to a machine and where it cannot. The Monday question keeps the whole thing honest week to week. Organisations stuck in pilot mode usually have plenty of capability and very little of the other two. Watch the full conversation above for the longer version, including how René sees the operating model change once AI moves from the edge of the business to its centre.