What Is Decision Intelligence?

Decision intelligence means designing how important decisions get made: define the real question, use data without worshipping it, listen to the people closest to the work, assign one owner, and decide in advance what would prove you wrong. Better decisions are engineered, not improvised.

I named my family's firm Educated Guess Ventures for a reason. Every meaningful executive decision is a guess — you don't get to inspect the future before you choose, and calling it "strategy" doesn't make the uncertainty disappear. The job is to make the guess more educated.

That, to me, is decision intelligence. Not eliminating uncertainty. Not pretending the spreadsheet contains the future. Building a better process for deciding when the answer is unknowable.

Why do smart executives make bad decisions?

Because intelligence isn't usually the constraint — process is. The meeting starts, the slides go up, everybody already knows which answer the boss likes, a few alternatives appear because the template requires them, someone asks for objections, and there are none. That isn't a decision process. That's organizational theater with catering.

I've watched versions of it across industries. Smart people can be spectacularly good at rationalizing the answer they already wanted. The antidote is not smarter people. It is structure: write down the question before arguing over answers, separate facts from assumptions, create alternatives that are genuinely different, give dissent a job, set the conditions under which you will admit you were wrong — and make somebody own the call.

What does "eyes on data, ears on people" mean?

"Eyes on data, ears on people" is the operating principle I run companies by. You need both because both can lie to you in different ways: people are intrinsically biased, so you need data to keep poor judgment from compounding — but data only measures what somebody decided to measure, so you need the people closest to the work to tell you what the numbers are missing.

Experience doesn't remove bias. Sometimes it just makes us more articulate when defending it. So look at the numbers. But a dashboard can tell you conversion fell; it can't always tell you why your best salesperson thinks the offer stopped landing. It can tell you satisfaction moved three points; it may not tell you that customers are furious about one specific change and your average is hiding it.

Data people dismiss anecdotes. People people dismiss spreadsheets. Both are half blind. The discipline is holding the two together: here is what the data says — now tell me what I'm not seeing.

What does a well-designed decision actually look like?

A well-designed decision has six parts: one accountable owner (a name, not a committee), the actual question written down, at least three genuinely different options, the strongest case against the favorite argued out loud, a kill criterion set before commitment, and a review date where you compare what happened with what you expected.

Two options is usually one preferred answer holding a hostage. And the review matters because outcomes can fool you: a great decision can have a bad outcome, and a reckless decision can get lucky. If you only grade the result, luck starts masquerading as skill. Grade the process too — did we ask the right question, use the evidence available, hear dissent, understand the downside, change our mind when the evidence changed? That is how an organization gets smarter.

How do you start practicing decision intelligence?

Don't launch a "Decision Intelligence Transformation Initiative." Take one consequential decision sitting on your desk right now and write down: who owns it, what question we are actually answering, three real alternatives, the strongest argument against the option we currently prefer, what would have to happen for us to admit we were wrong, and when we will review it. Do that once. Then do it again.

Decision quality becomes cultural when disciplined behavior becomes normal, not when somebody designs a logo for it.

Run the Decision Quality Audit against your last major decision — ten questions, five minutes. For the board-level version of the same discipline, read How Should Boards Govern AI?.