The most expensive thing in your company isn't a bad bet. It's the fifty good ones your process wouldn't let you try.
AI can change the cost of exploring an idea. The opportunity is to find tests that are small enough to run, useful enough to learn from and appropriate for the risks involved.
A CEO of a $400M industrial services company walked me through his AI prioritization process recently. Four months. 60 ideas. Narrowed to 12. Business cases. Stage gates. Steering committee approval.
That process made me ask whether every idea needed the same path to a decision. Some needed a full business case. Others might have benefited from a smaller test first.
Recheck the cost of exploration
A draft or prototype can sometimes answer a question that would otherwise occupy a meeting. It still needs a clear purpose, suitable data and someone who can evaluate the result.
The math of testing ideas changed. Most organizations' governance structures haven't.
Strategy was never about imagination
For two decades, strategy functioned primarily as a selection mechanism. Kill ideas. Focus. Protect the core. Don't waste resources.
That wasn't wrong. It was correctly calibrated to a world where exploration was expensive. Testing a new market took quarters and millions. So you had to pre-select ruthlessly.
"The constraint on innovation was never imagination. It was the cost of testing imagination."
When a useful test becomes cheaper, the approval process deserves another look. Keep the review needed for consequential decisions, and create a lighter path for bounded experiments.
A product example
Duolingo’s 2025 language report records 148 new courses for speakers of more than 20 languages. That is a concrete expansion of what people could learn. The question for another business is which underserved need a new capability could help it address.
The useful lesson is to connect a capability with a customer need and a result you can observe. More experiments only help if they produce better evidence.
The question has changed
When execution is expensive, the winning question is: "what can we build?"
When exploration gets cheaper, another question becomes more valuable: what problems can we uniquely see?
Your proprietary data. Your customer relationships. Your domain depth. The things competitors can’t replicate. Those become the moat. Not the build.
Audit your own infrastructure
Look at your operating model with fresh eyes. Ask, honestly:
How much of your stage-gate process exists because failure used to be expensive?
How many committees exist to save money that it would no longer cost to spend?
Which planning decisions could be informed by a small test before the next formal review?
How many business case templates ask for certainty you could just go build and find out?
Every one of those answers is a drag coefficient on your ability to compound.
What to do this week.
Cancel one meeting. Give the hour back to your leadership team as an open inventory.
List the ideas you killed, the markets you passed on and the products you shelved. Pick one where a bounded prototype could answer a real question. Set the scope and the evidence you need before you start.
You're not going to get all of them back. But the one you pick is probably more interesting than anything currently in your roadmap.
From the portfolio
Board of Innovation runs an AI Transformation Studio that helps mid-market and Fortune 500 companies turn their shelved idea backlog into a pipeline.
AlignAI provides the infrastructure to track, govern, and scale AI experimentation across the organization, so when exploration goes wide, visibility doesn't break.
Caution should be proportionate to the decision. Make room for inexpensive learning, and keep the judgment that consequential work requires.
What is waiting costing you right now?
