A conclusion you can defend in front of your board.
Kapari ends every test of your decision with a level of reception, from “outright rejection” to “clear support”. That conclusion comes from a fixed rule, not from an AI's mood: you can show where it comes from.
What it changes for you.
You present a price increase. Someone asks where the recommendation comes from. You show what each group said and the rule that made the call. Nobody can dismiss it with "the AI said so."
- An AI gives each voice its words: a loyal customer, a shop-floor supervisor, a neighbor. Each one reacts to your decision.
- Kapari counts who supports it, who opposes it, and whether a whole group digs in.
- The rule decides, the same way every time. Rerun the test in three months: it will say the same thing.
We pitted an AI against the rule. The rule won.
Across eight decisions, the AI tasked with concluding found nothing the rule had not already seen. We removed it.
Research has shown it since 1954: a rule applied the same way every time beats the best expert, who changes their mind from one day to the next. The reactions are still written by the AI, which is why Kapari sits a public exam on 100 real decisions from the past.
Kapari runs your decision past simulated voices so you walk in prepared. Your team and your customers still have the final word.
For the curious: the full rule
The overall score of the voices runs from -100 (very hostile) to +100 (very favorable). A voice counts as favorable from 25 up, as opposed at -25 or below. The spread of opinion is 0 when every voice says the same thing. A group counts as opposed when its score is -12 or below without the group being split itself (internal spread below 45). The public demo gives a simplified conclusion, which takes into account neither whether the topic is sensitive nor the risk found by the analysis.
Hold off
Any one of these is enough:
- score of -15 or below;
- at least 60% of the voices opposed;
- an opposed group of at least 30% of the voices, with a high risk (at least 2 risks rated high by the analysis).
Adopt
All at once:
- score of 25 or above;
- at least 50% favorable, at most 25% opposed;
- spread of opinion below 55;
- no opposed group of 30% or more;
- no high risk backed by at least 30% opposition (with 25% at most, it stays a point to watch).
Adjust
Every other case, plus:
- a sensitive topic (politics, social issues);
- a risk review that could not run: "Analysis unavailable, caution."
Below 6 voices, when less than 60% of the voices answered, on a sensitive topic, or when group weights come from public data, the reception stays in words: Adopt then requires a spread below 40 and no high risk (2 risks rated high). When you change group weights, a very tense reception also counts as a high risk: score of -15 or below, opposition of 45% or more, spread of 55 or more, or an opposed group of 30% facing a spread of at least 35. An automated test checks this page against the code.p77u