Incident AI helps HR and employee relations teams investigate complaints, grievances and conduct matters, from the first allegation to the final letter. It helps organise the evidence, sets each account side by side and helps you check the process. It is designed not to make findings: that stays with your people.
A sound decision can still be challenged on how the investigation was run. These are the places where that happens.
Statements, emails and chat messages disagree. The differences are easy to miss, or to settle too early.
One sentence that states an allegation as fact can undermine an entire report.
Two witnesses can look like corroboration when both heard it from the same person.
The respondent was told and given time to respond, but the case file does not show it.
Each letter is written from scratch, so its wording drifts from what was actually alleged.
Case material sits in shared drives and inboxes, with no record of who has opened it.
Eleven guided steps in four stages. Each one is drafted from your evidence, cited to its source and approved by a person before anything builds on it.
Upload the complaint, statements, messages and rosters. Incident AI drafts a list of the people, allegations, dates and sources, with each allegation worded neutrally for you to review.
Each person gets their own plan: an objective, topics to cover and open questions drafted from the evidence. Record the answers as you go and they become the record of interview.
For each allegation, Incident AI sets out what the complainant, the respondent and each witness say, with the source of every account. Conflicts and missing corroboration are recorded, not resolved.
Incident AI finds the relevant clause in the procedures you supply and places the reported conduct beside it, with both sides cited.
Before the report is drawn, see how many accounts, documents and procedures stand behind each allegation, and where the support is thinner than it looks.
The neutrality review reads your approved outputs for sentences that state an allegation as fact, prefer one account or lead a witness. It shows the same point written neutrally.
The procedural fairness check compares what your case file records with what your procedure requires. Readiness for decision then lists what is still outstanding.
Generate a Letter of Allegations or Show Cause correspondence as a Word draft, from your own templates and the records you select. Export the investigation report as PDF or Word.
The workflow also records how the matter arrived, builds a source-traced timeline, and draws out key learnings about process, culture and training, kept separate from any individual finding.
HR investigations affect people’s jobs and reputations. Incident AI is built so the judgement stays with the people accountable for it.
These limits are built into the product, not just its instructions. AI output is screened for verdict language and confidence scores, and blocked when they are detected. Findings are entered separately by an authorised person, with their own audit trail.
Incident AI uses generative AI. Its drafts can contain errors or miss things, so a person reviews each output before it is approved. It supports your investigators and does not replace their judgement or legal advice.
HR investigations run in the same Incident AI workspace as your safety investigations, but they are kept apart.
An HR case is visible only to the people explicitly assigned to it. Being an Incident AI user is not enough.
HR cases have their own workflow, tools and reports, and cannot be opened through the safety screens.
Each case shows who currently has access and a record of the investigation activity.
If the evidence or an earlier step changes, later outputs are marked as needing review. Approved work is not replaced without being flagged.
Letter templates are uploaded, previewed and published as fixed versions by an authorised administrator.
A report is created from the current approved outputs and stored as a snapshot, so it can be produced again exactly.
Hosting, database and AI inference can run in your region or your own cloud tenant.How sovereign deployment works
It is an investigation type inside Incident AI, built for HR and employee relations teams. It guides a workplace investigation through eleven steps, from recording how the matter arrived to checking it is ready for a decision-maker. Statements cite the evidence they draw on, so a reviewer can check them, and a person reviews and approves each step.
No. Incident AI organises the evidence, sets the accounts side by side and flags conflicts and gaps. It is designed not to make findings, give confidence scores, reach legal conclusions or recommend disciplinary action. Findings are entered by an authorised person in a separate step.
Workplace complaints, grievances and conduct matters where an organisation needs to investigate allegations about how someone behaved at work, then give the respondent a fair chance to answer them.
Only the people explicitly assigned to that case. HR cases are not visible to other Incident AI users, including safety investigators in the same organisation, and each case keeps a record of who has access and what was done.
Yes. The procedure comparison and the fairness check work from the procedures you upload to the case. Letters are generated from your own Word templates, which an authorised administrator uploads, previews and publishes.
No. Safety and HR investigations are two investigation types in the same Incident AI workspace, with the same sign-in and evidence handling but separate workflows, reports and access. HR investigations are switched on for each organisation.
We will walk your team through a complete case: the allegations, the interviews, the checks and the correspondence.
HR investigations are switched on for each organisation. Ask us about yours.