The High-Velocity, AI-assisted "Laboratory"

I said in my previous post that one of the ways Rumelt defines "bad strategy" is when leaders use it to avoid the 'pain of choice'. I'd take this one step further and say the only way to commit to "good strategy", and confront that pain of choice, is by being open to experimentation. We often underestimate the importance of this shift in thinking...

The pressure to focus

We know industry leaders today are facing significant market forces, while demand rises for greater impact with less resource (as always!)...

It is no longer possible to hedge your bets by running multiple, unfocused initiatives. To drive effective customer engagement in this constrained environment, one must start with a clear focus.

Ideally - a singular, testable strategy.

The good news is that this need for more experimentation arrives just as AI is opening up more powerful support avenues for this type of approach. Leaders now have the capabilities to test, learn, and iterate faster than ever before.

This is how AI can help us be more strategic.

But its critical to recognize that there are capabilities needed for "good strategy" where AI will always fall short...

Traditional strategy is often a static plan based on a best guess. In the AI era, a strategy must transform into a living experiment. The critical point is that a strategy cannot be proven right in advance. It is merely a judgement or a ‘hypothesis’ that must be tested.

Good Strategy: The first step is Diagnosis

AI can analyze data, but only human leaders can Frame the Problem. This is about asking the right strategic questions and identifying the true source of a market challenge.

The next step is a Guiding Policy

A guiding policy is an overall approach chosen to cope with or overcome the obstacles identified in the diagnosis. Like the guardrails on a highway, the guiding policy directs and constrains action in certain directions without defining exactly what should be done.

In an AI context, humans can better define the hypothesis by setting this guiding policy with a Reward Function. This is about defining what “good” looks like (e.g., maximizing prescriptions while minimizing message fatigue). 

Instead of following a fixed map, AI can optimize for this reward function — a scoring system based on business outcomes (such as sales or engagement), journey quality (avoiding message fatigue), and strategic priorities. This allows the AI to dynamically select the best channel, content, and timing for every single interaction.

The "AI-friendly" Step: a Coherent Set of Actions

This is where AI’s role becomes useful to the experimentation process, where it can act like the 'researcher' running the trials. Unlike a human who can only test one or two marketing approaches at a time, AI agents can run “what-if” algorithms and simulations. It’s where we can truly apply the “scientific method”, using hypothesis testing, meticulous measurement, and continuous refinement based on evidence.

Staying true to Rumelt´s definition of a "good strategy", AI can ensure that the set of actions taken is coherent and mutually reinforcing. By automating routine tasks (e.g. channel or content selection), AI allows a strategy to be executed with a level of coordination and focus that manual management cannot achieve.

“Test results” can also be immediate with this approach. Tools can provide audit trails to see where the AI succeeded or failed, while also allowing you to run experiments with different content arcs and audiences to see which theory yields the highest reward.

But, there is no replacement for the Human Commitment that is needed to follow a set of actions. This is where one needs to always exercise Judgement. Some actions might clash with non-quantifiable factors (e.g. market mood, competitor irrationality, ethical considerations) and ultimately one will always need a human at the helm to make the final, uncomfortable Commitment to a choice.

There it is again, that “pain of choice”. It’s unavoidable.

In summary, AI’s role will always be to elevate the human strategist. It can excel at automating the execution, measurement and iteration of a strategy. And by doing so, it could free us up to focus on the high-value, difficult tasks that matter the most. The initial Diagnosis, the final Choice, and the unwavering Commitment.