What I Would Fix First in AI SaaS Problem Interviews

If an AI SaaS team fixes one habit in their problem interview process, fix this: stop opening with the AI question. Open with the workflow. Almost every other AI-specific failure follows from getting that one decision wrong.

Why This Is the Bottleneck

The moment you say "tell me how you currently use AI," you have invited the customer to give you a position statement instead of a workflow description. Position statements are noisy. Half of them are about how AI is overhyped. Half are about how AI is underhyped. None of them tell you what your customer actually does on Tuesday afternoon.

Workflow descriptions are dense with signal. Where they pause, what they redo, what gets handed off, where the spreadsheet lives, what they wish would just disappear. Whether AI is in that picture is a fact you discover, not a topic you import.

The Specific Discipline

Decide before the call that you will not say the word "AI" until the customer says it first. If they describe their workflow without ever mentioning AI, do not steer them there. The absence is data: this workflow is not AI-touched today, and you can make of that what you will.

If they bring up AI, you can follow them down the thread. But you do not import the topic. The customer's spontaneous mention is the difference between data and theater.

What This Unlocks

Once you stop steering, the conversation gets richer. You hear about the bash script someone wrote that you assumed they would describe as "AI tooling." You hear about the contractor they hired and what the contractor produces by hand. You hear about the manual review step that exists because trust is broken somewhere.

None of those would have come out if you had asked the AI question. The customer would have answered the AI question, and you would have moved on, and the actual workflow texture would have stayed hidden.

What If They Ask What You Are Building?

Customers will ask. Especially in AI SaaS, where everyone is curious. The right answer is not "an AI tool that does X." The right answer is what you are actually doing right now: research. "Honestly, I am still figuring out the direction. Can we keep talking about how this currently works for you?"

Almost everyone says yes. The few who push will accept the answer if you keep the conversation focused on them. The deflection works because it is also the truth. You are still figuring it out. The interview is part of how.

The Hidden Win

The interviews where you do not lead with the AI frame produce something founders rarely expect: a sharper sense of where AI does not belong in the workflow. That is just as valuable as knowing where it does belong.

The customer who tells you they have an AI step but do not trust it for the final review is telling you that the boundary your product has to respect is exactly there. Cross it and the product fails. Stay inside it and you are useful.

How to Train Yourself Out of the Habit

Pin the rule on your sticky note: "do not say AI first." Practice with a friend before the first real call. Have your co-founder review your transcripts and call you out when you slipped.

The first three calls will feel awkward because the AI frame has been so heavily reinforced everywhere else. By the fifth, it will feel natural. By the tenth, you will wonder why you ever opened with the AI question.

The One-Sentence Version

If you remember one thing about AI SaaS problem interviews, remember this. Open with the workflow, not the technology. Most other AI-specific failure modes dissolve if you hold that line.