The Driver Check

A practical AI readiness checklist from AI, Honestly EP002 — free for you and your team.
Enter your email to read it and save a copy.

No spam. No newsletter unless you ask.

AI, Honestly — EP002 Takeaway

The Driver Check

AI will accelerate whatever direction you point it. Before your next AI initiative, someone in the room needs to be able to answer three questions.

From the episode "Automating the Wrong Things"  ·  AI, Honestly  ·  kipdavis.com/ep002.html

In enterprise AI initiatives, the most common failure mode isn't the AI. It's the foundation the AI runs on — unclear questions, unprepared data, no one with enough context to evaluate whether the output is pointing somewhere real.

The Enterprise Architect was the function that asked these questions. In many organizations, that function is being cut in the name of AI efficiency. But the questions don't disappear with the role. They just go unanswered — until they become expensive.

"The AI replacing the EA doesn't know what questions to ask. And if the person feeding it doesn't know what questions to ask either — the output is confident, well-formatted, and wrong. That's not an AI failure. That's a judgment failure that AI made invisible."

— Kyle, AI, Honestly EP002

Before your next AI initiative, run the Driver Check. Three questions. Someone in the room needs to be able to answer all three — not the AI, not the project sponsor, not the vendor. A person who knows this organization, this data, and what happens downstream when something breaks.

Question 1 of 3

Do you know what you're actually asking?

Not the prompt. The underlying question the initiative depends on. This means being able to state — clearly, specifically — what problem this AI is solving, what data it depends on, and what a wrong answer looks like.

Most AI initiative failures don't start with bad technology. They start with an imprecise question that nobody slowed down to sharpen. The AI generates an answer confidently. The answer is to a question nobody meant to ask.

Ask your team:

Can someone in this room state the specific question this AI is answering — not the use case, not the capability, but the question? Can they describe what a correct answer looks like? What a wrong answer looks like? What happens downstream if the AI gets it wrong at scale?

If nobody can answer that, you're not ready to build. You're ready to ask.

Question 2 of 3

Can the foundation support what you're building?

Gartner projects that 60% of AI projects will be abandoned due to lack of AI-ready data. IBM found that data quality issues doubled as the top obstacle to AI success in a single year — from 19% to 44%. The AI doesn't know the data is messy. It processes confidently.

Zillow's AI home valuation model worked exactly as designed. The data underneath it didn't hold. $500M loss. 25% workforce reduction. The AI was not the problem.

Ask your team:

Is the data clean, accessible, and fit for this specific use case? Are the APIs and services this AI will touch documented, rate-limited, and in known condition? Has anyone mapped what breaks downstream if this fails at scale? Is there architectural oversight — a single person with visibility into the full picture?

AI doesn't fix broken foundations. It runs on them — confidently.

Question 3 of 3

Who's driving?

Not who owns the project. Not who approved the budget. Who has enough context — about this organization, this data, these systems, these team dynamics — to evaluate whether the AI's output is pointing somewhere real?

The EA's value was never the diagram. It was knowing what the diagram couldn't show — the political landmines, the technical debt, the team that made a decision six months ago that will collide with this one. That knowledge isn't in any training set.

Ask your team:

Can you name the specific person — not a role, a person — who has enough organizational context to evaluate the AI's output critically? Who will be called when something breaks? Is that person in the room when architecture decisions are being made, or are they called in after the fact?

That person is not optional. They're the reason the other two questions get answered.

The companies that will be 24 months ahead aren't moving faster.

They're answering these questions before they move at all. They're pausing to fix their data foundations. They're keeping the people whose job is to ask whether the initiative will actually work. They're making sure someone is driving.

AI will accelerate whatever direction you point it. If you know the right questions, if your foundation is solid, if someone is driving — it's a force multiplier.

If none of those things are true, it's a faster way to go somewhere you didn't want to go. Confidently.

Further Reading