Workplace investigation software where every finding is tied to quoted evidence and tested for gaps before anyone signs
Every finding tied to quoted evidence and tested for gaps before anyone signs.
See one case, screen by screen ↓Workplace investigation software that runs employee relations investigations from complaint to finding with cited evidence. Seven agents turn the complaint into specific allegations, check investigators for conflicts, plan the witnesses, quote every interview, email, chat message and record against each allegation, test each proposed finding for gaps and bias, draft the cited report and watch every party for retaliation. The investigator writes and signs each finding, and the Head of Employee Relations approves the report.
How to choose AI solutions for back-office and enterprise work →
Updated
An employee relations investigation is central to the employer's defence when a harassment, discrimination or retaliation complaint becomes a charge. It is judged on whether it was prompt, impartial and adequate, and whether each finding follows from the evidence. Today the allegations sit in one document, the interview notes in others, and the emails and chat exports in a folder, and the investigator pulls them together by hand to write the report. On a case with 37 emails and 214 chat messages, reading every line against every allegation rarely happens, so people sample.
The weak points are the same every time. An allegation stays too vague to be found or not found. A conflict of interest surfaces after the investigator is assigned. A witness who was in the room is never interviewed. A credibility call rests on who sounded more convincing, and the respondent's explanation is never answered. And retaliation shows up weeks later as a rating, schedule or transfer change in a system employee relations doesn't watch.
Estimated split for one investigation against a 35-business-day finding clock, by hand and with the solution. Interviews stay with people, so they barely move.
Seven agents shape allegations, check conflicts, plan, quote the evidence, challenge each finding and watch for retaliation; the investigator signs.
Dana Whitfield, a Senior Demand Planner at Kestrel & Vane Distribution in Columbus, complains about her manager, Mark Halloran: an age-related remark in a team meeting, a warning about her review if she went to HR, and more. It is Wed 14 Oct 2026, business day 27 of 35. Here is how the investigator works the case with the agents, screen by screen, in the working solution.
Alicia has 11 open investigations. Today shows 4 things that need her, each with one next step, most urgent first: interview Ravi Menon before writing A2 on Dana's case, approve 3 allegations on a new complaint, review a retaliation alert, and answer counsel's 2 comments on a report due Friday. Beside it, the agents say what they found overnight, and she can ask about any of it.
Challenge agent, 07:20: “ER-2026-0219: A2 has a gap. Ravi Menon was at the 12 Aug meeting and has not been interviewed. A1, A3 and A4 passed.”
Rosa Delgado's complaint about her shift supervisor arrives on the Speak Up line and is kept exactly as received. The intake agent proposes 3 allegations — who, what, when, where — each mapped to a clause of the Respect at Work policy and to the sentence it came from. “He’s like this with everyone, ask around” names no act or date, so it is kept as a line of enquiry. The conflict agent checks 4 investigators: Ben Ford shared a site with the respondent from 2019 to 2023, Chen Wu interviewed him in a prior case, and Sofia Marquez is recommended.
“The intake agent proposes; you decide what is investigated.”
Dana's plan was approved on 10 Sep. The planning agent proposed each witness from the allegations, the org chart, calendars and room records, and says why: Luis Ortega was in room 3B on 12 Aug, Hannah Cole sent the 10:47 message. Alicia removed Erin Park with a reason. Legal hold LH-2026-061 covers the emails and chat, and the interim measures are in place: Dana's rating is frozen and her 1:1s moved to Erin Park.
The allegations are the rows; the complainant, respondent, witnesses, 37 emails, 214 chat messages and the records are the columns. Every one of the 24 cells says supports, contradicts or silent, with the quoted line and where it came from — 10 support, 7 contradict, 7 are silent. A silent cell still says what was read. The last column holds each proposed finding with the standard of proof, more likely than not.
On A2, the age-related remark, the chat cell opens message 118 of 214 in context. At 10:47 Hannah wrote to Dana: “That comment about your age was out of line. I’m sorry you had to sit through that.” The drawer shows how the agent read it, who collected it under legal hold, its hash and who has seen it. Alicia can agree, or change the reading with a reason.
“The source describes the same conduct as the allegation, in its own words. Quoted exactly; no paraphrase.”
Two witnesses and a chat message 27 minutes after the meeting match Dana's account; Mark denies mentioning age. The proposal is substantiated at 0.78. The challenge agent runs its tests: the respondent's account is addressed, credibility rests on consistency and a message the same morning, not demeanour, and the outcome is consistent with 6 similar cases. One test fails: Ravi Menon was in room 3B for the whole meeting and was never interviewed. Sign finding stays locked.
An override needs a reason code — the witness is unavailable, not material, or would only repeat evidence — and it shows in the report.
The attendance record shows six people. Alicia asks why Jo Kim doesn't count. The answer cites the record: Jo joined by video at 10:00 and left at 10:04, and Dana places the remark at about 10:20. Satisfied, Alicia schedules Ravi for Fri 16 Oct, 10:00, room 2C, with a 9-question guide that passed the leading-question check. The invitation goes with the confidentiality and no-retaliation notices, and the watch now covers Ravi for 180 days.
Challenge agent: “So Jo could not have heard it. If Ravi puts the remark earlier, I’ll raise her again.”
A1 — left off the Q3 planning reviews after asking about pay — passed all 3 tests. The calendar shows Dana was invited to 5 of the 6 reviews, and the one removal follows a move to an 8-seat room. Alicia chooses not substantiated, keeps the reasoning with its citations, picks the reason code “F2 · The evidence does not support it” and signs as the investigator.
On 2 Sep at 09:14 Mark changed Dana's mid-year rating draft from Meets to Partially meets — 13 days after HR told him of her concern, and the only one of his 9 reports' ratings he changed. Dana filed formally on 8 Sep. Because the watch counts from her first contact with HR, the agent found it on 9 Sep. Alicia upheld it, it went into the grid as evidence for A3, and Nadia Brooks froze the rating.
“The agent only raises an alert; a person decides.”
The report agent drafts scope, method, each allegation and finding, credibility and recommendations, with every sentence cited. Until a finding is signed, it reads “Finding pending” — never final. Because the case is high risk, sign-off runs from Alicia's 4 signatures to a final challenge check, to Daniel Osei's review under privilege, to Margaret Liu's approval with a reason code.
“The report agent never presents an unsigned finding as final.”
The same investigation, seen by the four people who carry it — what their weeks looked like, and what they look like now.
Seven agents shape allegations, check conflicts, plan, quote the evidence, challenge each finding and watch for retaliation; the investigator signs.
Turns a complaint into specific allegations — who, what, when, where — maps each to policy, sets a risk tier, finds related cases and proposes interim measures.
Checks each candidate investigator against reporting lines over the last 3 years, shared sites, prior cases as a party or witness, and declared relationships, showing the evidence for each conflict.
Proposes the witnesses, evidence, preservation requests and interview order for each allegation, and drafts interview guides checked for leading questions.
Reads transcripts, emails, chat messages and system records, and fills each cell with supports, contradicts or silent and the quoted line.
Independently tests each proposed finding for an attendee not interviewed, an unread source, an unanswered explanation, the credibility reasoning and consistency with similar cases. A gap stops the signature.
Drafts the report from the signed findings with every sentence cited, and drafts the outcome letters to each party.
Watches every party's ratings, pay, schedules, transfers, access and exits for 90 days, or 180 for high risk, counted from the first contact with HR.
Signs findings; Head of ER approves; counsel reviews. The agents propose; a named person decides.
Every working screen has a place to ask about it. Answers cite their source, and an action is only ever proposed for you to confirm — nothing changes without you.
Why doesn't Jo Kim count?
Jo Kim joined by video at 10:00 and left at 10:04. Dana places the remark near the end of the demo, at about 10:20, and no witness puts it earlier. So Jo could not have heard it. If Ravi puts the remark earlier, I'll raise her again.
What would make A3 inconclusive?
Two things would change it. One: Mark shows he saw Nadia's email only after the 1:1 — the archive shows he opened it on 20 Aug at 16:30. Two: a reason for the rating change recorded before 20 Aug — there is no calibration note. Neither applies today, so it stays substantiated at 0.71.
Why count from the first contact with HR?
Because retaliation often starts before a formal complaint. Dana first emailed HR on 19 Aug and filed on 8 Sep; her rating was changed on 2 Sep, between the two. Counting from the complaint would have missed it.
Why is Ben Ford in conflict?
Core HR shows he supervised at the Toledo plant alongside Greg Doyle from 2019 to 2023. The rule is no shared site in the last 3 years, so he can be assigned only with Margaret Liu's override and a reason.
13 screens from the working solution, on its sample data. Pick one to see it large.
Set by the administrator and the Head of Employee Relations. Every change is versioned and logged.
| Setting | Default | Choose from |
|---|---|---|
| Standard of proof | US and Mexico: more likely than not | Set per country · UK, Ireland, Canada: balance of probabilities |
| Clocks | Acknowledge in 2 business days · finding in 35 | EU reports under Directive 2019/1937: 7 days and 3 months, which cannot be lengthened |
| Conflict rule: reporting line | Within 2 levels in the last 3 years | Levels and years set by your policy |
| Other conflict rules | Same site in 3 years · party or witness in a prior case · declared relationships | Each on or off |
| Challenge agent before every signature | On | Cannot be switched off for one investigation |
| Confidence floor for a cell | 0.65 | Below it, the cell is marked for a person to read |
| Counsel review for high-risk investigations | On | Physical contact, retaliation, executives, or a charge already filed |
| Retaliation watch window | 90 days; 180 for high risk | Counted from the first contact with HR, not the formal complaint |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample data · “target” = the design goal, measured in the live solution · “estimated” = our estimate · “typical” = published figures (EEOC FY2024 charge data via CWC — retaliation alleged in 47.8% of charges; EEOC FY2025 results, 6 Apr 2026 — $660M recovered for 17,680 people; Worker Protection (Amendment of Equality Act 2010) Act 2023 — up to a 25% uplift where reasonable steps to prevent sexual harassment were not taken — and TULRCA 1992 s.207A — up to a 25% uplift for not following the Acas Code; EEOC Enforcement Guidance on Harassment, No. 915.064, on a prompt and adequate investigation by an impartial party, and Faragher and Ellerth (1998) on the employer's reasonable-care defence; EU Directive 2019/1937 Art. 9 on the 7-day and 3-month clocks). People and companies named on this page are fictional — characters and sample data in the working solution.
Workplace investigation software that runs employee relations investigations from complaint to finding with cited evidence. Seven agents turn the complaint into specific allegations, check investigators for conflicts, plan the witnesses, quote every interview, email, chat message and record against each allegation, test each proposed finding for gaps and bias, draft the cited report and watch every party for retaliation. The investigator writes and signs each finding, and the Head of Employee Relations approves the report.
The allegations are the rows and the sources are the columns: complainant, respondent, witnesses, emails, chat messages and records such as badge, roster and rating history. Every cell says supports, contradicts or silent, with the quoted line and its place in the source. Click a cell to read the passage in context, and agree or change the reading with a reason.
Before a finding can be signed, it tests whether corroboration is exhausted — every attendee interviewed, every source read — whether the respondent's account is answered, whether credibility rests on evidence rather than demeanour, and whether the outcome is consistent with similar cases. A gap blocks signing; an override needs a reason code and shows in the report and at approval.
The retaliation watch agent checks every complainant's and witness's ratings, pay, schedules, transfers, access and exits every night, for 90 days or 180 for high-risk cases. It counts from the first contact with HR, not the formal complaint, so it can find a change made in between. It only raises an alert; a person upholds or explains it with a reason.
Before anyone is assigned, the conflict agent checks each candidate investigator against reporting lines within 2 levels in the last 3 years, shared sites, prior cases as a party or witness, and declared relationships, and shows the evidence for each conflict. A conflicted investigator can be assigned only with the Head of Employee Relations' override and a reason.
Yes. The agents propose and cite; they never decide a finding, contact a party or witness, try to identify an anonymous reporter or change an HR record. The investigator writes and signs each finding with a reason code, counsel reviews high-risk reports under privilege, and the Head of Employee Relations approves or returns the report with a reason code.
Only the people on its access list. HR business partners see the summary, interim measures and actions but never the evidence, cases involving HR staff or executives go to the Head of Employee Relations and counsel only, and the administrator sees configuration but no case content. Every view, including the agents' reads, is on the audit trail.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your investigation standard, policies, clocks and a sample of closed, anonymised cases — and it goes live once every quality gate has passed. We will walk you through it on your own investigations first.
Every HR answer cited to the policy that applies; sensitive cases always with a person.
1-min demoAI Resume ScreeningEvery applicant read against the job, and every decision cited and checked for fairness.
1-min demoForm I-9 ComplianceAn estimated 86% less fine exposure, every correction signed and never back-dated.
1-min demoWe'll run Workplace Investigation Management on a sample of your own closed, anonymised investigations.