Common questions

The things safety directors ask first.

Including the ones where the honest answer is no. If a question you have is not here, ask it directly — hello@chatir.io.

Our data is a mess. Is that disqualifying?
No — it is the normal starting condition. CSV or Excel is fine, your column names are fine, and narrative detail matters far more than a tidy template. A sample review tells us whether there is enough depth to proceed before anyone commits.
Will this be used against my crews?
The analysis is built to describe conditions, not people. Findings are framed in systems terms, and the interview is deliberately non-accusatory — that is a design constraint, not a marketing line.
Can you take PHI?
Not currently. Deidentified operational data only. The contractual and technical safeguards needed to support PHI are being built, including BAAs with applicable vendors, and the current boundary is published rather than buried.
Will this reduce our incidents?
We do not promise that, and you should be skeptical of anyone who does. What the record can establish is which conditions recur and which travel together. Whether a control changes anything is a question the next ninety days answer, not the report — which is why every candidate control arrives with the observability question attached: who would sign this off, and what could you look at to tell whether it was ever actually in place? The report poses those. Answering them is the first thing a working session does.
Does this find root causes?
No, and it declines to claim it does. Recurrence and co-occurrence describe the record; neither establishes cause. What you get is the set of conditions that keep appearing, each over the population it was counted in, which is what tells you where a structured control analysis would be worth doing.
How is this different from the incident module we already have?
Most incident software records what kind of event happened and closes the file. The question it cannot answer is what was going on around the event, because there is usually no field for three of the five condition domains — and nothing a crew writes can ever be filed under a field that does not exist. The free instrument audit measures that gap on your own form in about ninety seconds.
How much of our time does this take?
For the diagnostic: one approved historical export, a sample review, and a working session. There is no configuration project, no integration, and nothing for your crews to learn — the analysis runs on the record you already have, exactly as it is.
What if the analysis does not find anything?
That happens, and it is reported rather than dressed up. A record can be too thin to support a conclusion, and saying so is more useful than a finding manufactured to justify the engagement. In practice the thinness itself is usually the finding, and it points at the reporting instrument rather than the operation.
Is our data used to train models?
No. Your incident data is used solely to produce your report, under strict access controls — not shared with other organizations, not sold, and not used for model training.
Who actually reads the report?
It is written for the person who has to defend a decision in a room — usually safety, operations or risk leadership. Every figure carries the population it was measured over and the report states plainly what it could not settle, because a document that only makes claims cannot be checked by the people being asked to act on it.
Do we have to buy the platform to get value from this?
No. The diagnostic is a complete standalone engagement with a fixed scope and no platform commitment. If ongoing use makes sense, the platform is scoped at the decision review based on your organization and how you actually work — not before.
Is the AI Interviewer making safety decisions?
No. The AI Interviewer conducts the interview and surfaces conditions and patterns; it does not determine cause, assign fault, or decide what you should do. Findings are provisional and systemic until a person with context and authority says otherwise. The product equips the decision and deliberately stops short of making it.

The cheapest way to find out

Fifteen questions about the form your crews already fill in, and a read on which contributing-condition categories it cannot see. Free, no signup, and your answers never leave the browser.

Take the instrument audit