The Two-Sentence Problem Test
The two-sentence problem test is the first gate on any AI build: describe the problem in plain words with no vendor vocabulary. If you can't, you don't have a problem yet — you have marketing residue.
The two-sentence test is the first gate on any AI build decision: describe the problem in plain words, and if you can’t do it without a vendor’s vocabulary, you don’t have a problem yet — you have marketing residue. It costs nothing, takes two minutes, and kills more doomed builds than any other move in the framework.
Two clean sentences? Run the full decision at Build, Buy, or Skip?. Then pressure-test any specific tool you’re weighing with Tool, Toy, or Trash?.
The Two-Sentence Problem Test
Fill in both sentences. No jargon. No vendor terminology.
Nothing you'd be embarrassed to say in front of a sharp
twelve-year-old.
The problem I'm trying to solve is ______________________.
The way I currently handle it is ______________________.
Why it works A good pitch doesn't argue with your understanding of your problem — it replaces it. Once "I want my staff to share notes about customers" has become "we need to close our customer intelligence gap," the vocabulary is doing your thinking, and it was written by someone with something to sell. Two side effects of the test are features, not bugs: if the honest version sounds embarrassingly small, that's the test working — most real problems are small, and small problems have small, cheap solutions. And if you can't fill in the second sentence at all, you've been thinking about solutions before you've ever looked squarely at the problem.
Then what Two clean sentences → run the full decision at Build / Buy / Skip. Can't strip the borrowed words → stop; spend a week with the actual work — the people, the data, your own confusion — until the plain version comes. Small honest version → solve the small version. Don't inflate it back to a size that justifies the engagement.
These prompts put the framework from Before You Build in your hands. More about the book →
Frequently asked questions
What is the two-sentence problem test?
It's the first gate on any AI build: state the problem you're solving and how you currently handle it, in plain words with no vendor terminology. If you can't, you don't have a defined problem yet — you have marketing residue.
What if the honest version sounds too small?
That's the test working. Most real problems are small, and small problems have small, cheap solutions — don't inflate it back to a size that justifies a vendor's engagement.