--- title: "Incident AI for HR | Workplace Investigation Software" description: "Run HR and employee relations investigations with cited allegation summaries, side-by-side accounts, fairness checks and draft letters. Findings stay with your people." url: https://mineguard-ai.com/solutions/incident-ai/hr updated: 2026-11-12 language: en publisher: MineGuard AI --- [Safety investigations](https://mineguard-ai.com/solutions/incident-ai)[HR investigations](https://mineguard-ai.com/solutions/incident-ai/hr) New: Incident AI for HR investigations # Workplace investigations built for fairness, cited and ready for scrutiny 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. [Book a demo](https://mineguard-ai.com/demo) See how it works Animated Incident AI screens for an HR investigation: an allegation summary built from the evidence, each person’s account set side by side, a neutrality review that rewrites loaded wording, and a letter of allegations generated as a draft. - Findings stay with your people - Statements cite their sources for you to check - Access limited to the named case team - A person approves every step The problems we solve for HR ## HR investigations are judged on the process as much as the outcome A sound decision can still be challenged on how the investigation was run. These are the places where that happens. ### Accounts that do not match Statements, emails and chat messages disagree. The differences are easy to miss, or to settle too early. ### Wording that reads as a verdict One sentence that states an allegation as fact can undermine an entire report. ### Corroboration that is thinner than it looks Two witnesses can look like corroboration when both heard it from the same person. ### Process steps nobody recorded The respondent was told and given time to respond, but the case file does not show it. ### Letters retyped from notes Each letter is written from scratch, so its wording drifts from what was actually alleged. ### Sensitive files with wide access Case material sits in shared drives and inboxes, with no record of who has opened it. How it works ## From the first complaint to the final letter 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. Stage 1 Case framing ### Start from what is alleged, not what is assumed 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 allegation cites the file it was drawn from, and the page where it can - You approve the summary before any later step can use it Animated Allegation Summary screen: evidence files come in, the people on the case are listed by role, and three allegations appear in neutral wording with their sources, then the investigator approves the summary. Stage 2 Evidence analysis ### Prepare, run and record every interview 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. - Questions are marked as AI-drafted or added by the investigator - See at a glance which interviews are still to do Animated Interview Plan screen: a list of interviewees, the objective and topics for one witness, an open question with the answer being recorded, and the interview marked complete. ### See every account side by side 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. - Differences it finds are flagged for you to test - Findings are entered by an authorised person, never by AI Animated Version of Events screen: three accounts of the same allegation appear side by side with their sources, a conflict between two accounts is flagged, a missing account is noted, and the finding is left for an authorised person. ### Set the conduct beside your own procedure Incident AI finds the relevant clause in the procedures you supply and places the reported conduct beside it, with both sides cited. - Uses your procedures, not a generic standard - Whether the standard was met stays with the investigator Animated Conduct vs Procedure screen: a clause from the workplace behaviour procedure is placed beside the reported conduct, each with its sources, and the decision on whether the standard was met is left to the investigator. Stage 3 Evaluation ### Know what each allegation rests on 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. - Helps flag allegations resting on a single untested account - Helps flag corroboration that traces back to one origin Animated Corroboration Check screen: a table counts the accounts, documents and procedures behind each allegation, one allegation is flagged as resting on a single untested account, and a second witness account is traced back to the complainant. ### Catch wording that reads as a decision 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. - Checks for seven kinds of loaded wording - You choose whether to change it or let the wording stand Animated Neutrality Review screen: a sentence that states an allegation as established is flagged with the reason, a neutral version of the same point is shown, and the investigator accepts the neutral wording. ### Check the process before it reaches a decision-maker The procedural fairness check compares what your case file records with what your procedure requires. Readiness for decision then lists what is still outstanding. - Designed to report on the record of the process, not on the conduct - Separates what blocks the report from what only needs recording Animated Procedural Fairness Check and Readiness for Decision screens: each procedural requirement is marked as recorded, not evidenced or to confirm, then the outstanding items are listed and one is closed. Stage 4 Documents and reports ### Letters and reports from approved work only 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. - Every draft needs human review before it is issued - Reports are built from approved outputs and can be reproduced later Animated HR correspondence screen: the investigator selects two allegations, a Letter of Allegations is generated from the organisation’s template as a Word draft marked for human review, and the full report is offered as PDF and 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. Built for fairness ## The AI prepares. Your people decide. HR investigations affect people’s jobs and reputations. Incident AI is built so the judgement stays with the people accountable for it. Incident AI prepares Your people decide A neutral summary of each allegation, cited to the evidence Whether an allegation is substantiated Each account, side by side, with conflicts flagged Which account to accept, and why The procedure clause beside the reported conduct Whether the standard was met Draft letters from your own templates What is sent, to whom, and when ### What Incident AI is designed not to produce - A finding on any allegation - A confidence or likelihood score - A legal conclusion - A disciplinary recommendation 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. Confidential by design ## Restricted to the people on the case HR investigations run in the same Incident AI workspace as your safety investigations, but they are kept apart. - **Named case teams** An HR case is visible only to the people explicitly assigned to it. Being an Incident AI user is not enough. - **Separate from safety** HR cases have their own workflow, tools and reports, and cannot be opened through the safety screens. - **Audit and access history** Each case shows who currently has access and a record of the investigation activity. - **Approvals that stay honest** If the evidence or an earlier step changes, later outputs are marked as needing review. Approved work is not replaced without being flagged. - **Controlled templates** Letter templates are uploaded, previewed and published as fixed versions by an authorised administrator. - **Reproducible reports** 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](https://mineguard-ai.com/trust#sovereign) Common questions ## Incident AI for HR investigations ### What is Incident AI for HR investigations? 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. ### Does it decide whether an allegation is substantiated? 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. ### What kinds of matters is it for? 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. ### Who can see an HR case? 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. ### Can we use our own procedures and letter templates? 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. ### Is this a separate product from Incident AI? 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. ## See an HR investigation run from complaint to letter We will walk your team through a complete case: the allegations, the interviews, the checks and the correspondence. [Book a demo](https://mineguard-ai.com/demo) [Incident AI for safety](https://mineguard-ai.com/solutions/incident-ai) HR investigations are switched on for each organisation. Ask us about yours.