You ran a decision through the test bench. Sooner or later the question comes. From an employee at the all-hands, from a reporter, from someone on your own board. "You tested this on fake people?" This page makes sure that question never catches you flat-footed. Here is why your use of Kapari holds up, and the exact words to say so.
This is Kapari's position, and it becomes yours the moment you use it: a panel of simulated voices is there to pressure-test against, never to skip the real people. The all-hands still happens. If there is a union, it still gets its say. The public comment period still runs. The board still deliberates. Kapari comes before all of that. It helps you spot objections, prepare answers and ask better questions. Real people give the real answers.
The charge you are worried about, "you simulated people instead of listening to them", only sticks if the simulation replaces the listening. Used the way it is built, Kapari does the opposite. You walk into the room better prepared, with sharper questions. Say so, and say it early. Transparency lands hardest when it comes before the question.
Four facts anyone in the room can check for themselves.
Kapari collects nobody's answers. One AI builds profile types from public data. Another has them react to your decision. No employee, no customer, no resident is ever contacted, and none of your team's personal data goes into the simulation.
How the panel gets built, how the verdict is computed, how each signal is produced: it is all written down in the how-it-works walkthrough. You are not defending a black box. You can hand over the link.
Kapari gives you a range of plausible reactions and a qualitative verdict: Adopt, Adjust, Hold. Stability is checked across three passes. On politically sensitive topics, no percentage shows up at all. Nothing that could pass for an opinion score.
Kapari is graded against 16 documented US decisions and the real criticism each one drew, checked one by one. Average coverage: 74%, nearly three in four. The worst score, 33%, sits on the exam page exactly as it came out.
Take them straight or bend them to your situation. The whole move is two steps: say what the tool did, then say what it did not do.
"Before we brought this to the board, we ran it through Kapari, a test bench for decisions. A panel of AI-simulated voices reacted to the plan so we could surface objections and blind spots early. Nobody real was interviewed. This is not a poll and not a measure of opinion, and the method is public. What you are about to see is qualitative input. The decision still gets made around this table."
"To get ready for this, the leadership team used a qualitative simulation tool. An AI builds plausible reactions from public data, so we could anticipate your questions and come in better prepared. No employee was interviewed and no HR data was used. This does not replace the conversation we are having right now, and it changes nothing about how we work with your reps. The method is public and it is online."
"We used a qualitative simulation tool to explore plausible reactions before the public comment period opens. It is not a poll and not a measure of opinion, and the method is public. No resident was interviewed. The exercise helped us come in with the right questions. The comment period, and it alone, is what gathers residents' voices and weighs on the decision."
Say it exactly like this, no hedging. This is the half of the answer that defuses the accusation.
It is not a poll. Nobody is interviewed, no sample is drawn, no opinion score comes out of it.
It is not a survey. No questionnaire ever reaches a real person. The voices are simulated, and labeled that way everywhere.
It is not a prediction. Kapari does not tell you what will happen. It shows you what is worth preparing for.
It does not claim to be representative. The weights describe how the panel was built, and they are labeled that way on screen.
It is not the conversation itself. It does not replace an all-hands, a public comment period, or a council debate.
It is not a black box. The method, the signals and the exam are all published, bad grades included.
Every PDF export, every Excel workbook and every share link carries a transparency notice built to be read out loud. It says what the document is, what it is not, and where to find the method. When the question comes, the fastest answer is to hand over the document.
How the panel gets built, how the verdict is computed, what each signal means and what it does not say. See how it works →
16 documented US decisions. Every real criticism checked one by one, the good scores next to the bad ones. The 33% score is right there on the page, and that is exactly why the rest is believable. See the exam →
Kapari assembles a panel of simulated voices and maps how their reactions break down, so you can decide before you announce. It is not a poll, not a survey, not a measure of opinion, not a prediction, and no real person is interviewed. The decision and the conversation stay entirely human.
Run yours through the bench. A panel of voices reacts, you read the range before you announce, and you walk out with the words to explain it.