Platform · Discussions

Twenty minutes with your people, asked by something that listens.

Structured AI discussions that adapt to each role and each answer, in French or English, natively. They surface the real process, the undocumented workaround, and the knowledge that lives only in someone's head, without ever asking twice what the system already knows.

Available

How it works

AI Discovery Discussions, step by step.

Prepare

The discussion briefs itself on what's already known, the scan, the org structure, previous answers, so no one repeats themselves.

Discuss

A 20-minute adaptive conversation, in French or English. Follow-ups go where the answers lead, not where a form says.

Respect the boundary

Decline a sensitive area and it moves on. Anonymous mode is available, and its limits are stated honestly.

Structure

Every answer becomes claims with sources, never a summary that loses who said what.

Evidence tiers

Where each claim stands.

Declared

What a person tells us, attributed and consented, the raw material of every discussion.

Corroborated

The same claim from two independent people, or one person plus a system signal.

Verified

Confirmed against a system of record: logs, grants, configurations.

Stated limits

What it will not do.

Will not ask again what it already knows, Ask-Once is a hard rule.

Will not push after someone declines a sensitive area.

Will not attribute anonymous answers to an individual, ever.

Will not score, rank, or evaluate the person being interviewed.

Features

3 available · 1 in preview · 1 on the roadmap.

Feature	Status
Adaptive interview engineFollow-ups driven by answers, role, and evidence gaps.	Available
Native FR/EN question banksAuthored per language, never machine-translated.	Available
Ask-Once memoryCross-references everything already collected.	Available
Anonymous modeWith honest guarantees, stated in plain language.	Design-partner preview
Voice discussions	Roadmap · Q3 2027

Available, ships today Design-partner preview Dated roadmap

Questions we actually get

Asked by real skeptics.

Will my people actually be honest with an AI?

More than you'd expect, and we design for the rest. Discussions are 20 minutes, respect declines, offer anonymous mode, and every participant sees exactly what is collected. The cross-check against systems exists precisely because no single source is fully trusted.

Is this a survey?

No. A survey asks everyone the same questions and loses the thread. These discussions follow the answers, when someone mentions a workaround, the next question is about the workaround.

What languages?

French and English, natively. Question banks are authored in each language by Quebec-French writers, not translated.

See this pillar inside a real diagnostic.

The two-week engagement uses every pillar above, and hands you the evidence-linked output.

Book a scoping call See the diagnostic