“I Don’t Want to Get Anyone in Trouble”

Two desks away

The witness sits two desks from both the complainant and the respondent. She's been asked to come to a meeting with an investigator and has spent the morning wondering what she's going to say.

The investigator explains why she's there. She answers the first two questions briefly. Then:

"Look, I don't want to get anyone in trouble. I have to work with both of them. Will they know what I said? Because honestly, I didn't really see anything."

The investigator has an email suggesting she was in the room when the key comment was made. She may not have seen anything. She may have seen a great deal and be deciding, right now, how much to share.

Witnesses have no stake — and that's the problem

Complainants come forward. Respondents have to answer. Witnesses are drawn in, often unwillingly. A witness interview in a workplace investigation therefore faces a different obstacle: the witness has little to gain by being candid and something to lose.

Four features make it difficult.

Relationships continue. The witness works with both parties and may be friends with one of them. Anything she says may affect those relationships.

She fears retaliation or reputational cost. Being seen as "taking sides" feels risky, especially if the respondent is senior.

She wants promises. Confidentiality, anonymity, assurances that nothing will happen to anyone. The investigator can't honestly provide most of them.

And her account may mix what she saw with what she heard. Office gossip, the complainant's version, the respondent's version — witnesses often blend them without realizing.

Open honestly

Explain why she's being interviewed — neutrally. "You're not in any trouble. We're speaking to several people who work in the area to understand what happened. Your account is one part of that." Not "we know you saw it."

Answer the confidentiality question accurately. "I can't promise that what you say won't be shared at all. Information goes only to people who need it for the investigation. And retaliation against anyone who participates is prohibited — if anything happens because you spoke to me, I want you to tell me."

Acknowledge her position. "I understand you work closely with both of them. That's not easy." Acknowledgement often reduces defensiveness more than reassurance does.

Explain what's asked of her. Honest answers about what she personally saw and heard. Not opinions, not conclusions, and not a verdict on anyone.

Get what she actually observed

Start open. "Tell me about the team meeting on the 14th. What do you remember?" Not "Did you hear him say X?" — which tells her what others have alleged and may shape her memory.

Separate observation from information. "Is that something you saw or heard yourself, or something someone told you?" Firsthand observations are evidence; secondhand accounts may lead to other witnesses.

Use documents neutrally. If there's an email showing she was present, show it without accusation: "This email suggests you were in the meeting. Does that fit with what you remember?"

Accept "I don't remember" — but test it gently. Ask about surrounding details: where she was sitting, who else was there, what happened next. Context often surfaces memory without pressure.

Don't share what others have said. Telling the witness what the complainant or respondent reported can contaminate her account and breach confidentiality.

And don't ask for opinions. "Do you think he harassed her?" isn't her question to answer and puts her in exactly the position she fears.

When the witness is loyal to one side

Some witnesses are close friends of the respondent or the complainant, and their account may lean accordingly. That doesn't make them unreliable — but it does call for care.

Don't challenge the loyalty. "I know you two are close" is enough; there's no need to question it.

Stick to observation. Specific questions about what she saw and heard produce useful information even from a partial witness.

Record the relationship. How the witness knows the parties is relevant context for the investigation's assessment — recorded neutrally, not as an accusation.

Close carefully

Ask if there's anything else she thinks is important.

Ask who else might have relevant information.

Remind her of confidentiality expectations — in the form the organization's legal guidance permits — and how to report retaliation.

Thank her. Participating was a burden she didn't choose.

Four ways it goes wrong

The overpromiser, who assures the witness complete confidentiality or anonymity to get her talking.

The leader, who tells the witness what was alleged and asks her to confirm it.

The accuser, who confronts the witness with evidence of her presence as if she's lying.

The opinion-seeker, who asks the witness whether she thinks misconduct occurred.

Why this isn't trained

Witness interviews are treated as routine. Investigators focus preparation on complainants and respondents; witnesses get less attention.

Leading questions are tempting. When the investigator knows the allegation, it's natural to ask the witness to confirm it.

Reluctance is hard to read. "I didn't see anything" may be true, or may be protective. Distinguishing requires patient technique.

And practice witnesses are cooperative. Colleagues playing witnesses rarely hold back, blend gossip with observation or ask for promises.

What witness-interview simulation can rehearse

A ten-minute simulation can put the investigator opposite a reluctant witness who asks for confidentiality, minimizes what she saw and mixes observation with office gossip — so the investigator practices honest assurances, open questions and separating firsthand from secondhand information. The AI agent in Foretell AI plays the witness consistently; policies, confidentiality instructions and legal guidance stay with the organization.

Four versions to build:

Design caution. Workplace investigations are governed by organizational policy and applicable employment law. Confidentiality instructions to employees, anonymity, anti-retaliation obligations and the use of witness statements vary by jurisdiction. Modules must use the organization's own policies and legal guidance. Nothing here is legal or HR advice.

Designing the module

Ten minutes, scored against a witness-interview rubric.

Pass one — the opening. Was the witness's role explained neutrally? Was the confidentiality question answered accurately? Was retaliation protection explained?

Pass two — the account. Were questions open and non-leading? Was firsthand observation separated from secondhand information? Were documents used neutrally?

Pass three — boundaries. Did the investigator avoid sharing others' accounts, asking for opinions or promising anonymity?

Rubric on observable behavior: Was the witness told she's not in trouble? Was complete confidentiality promised? Was anti-retaliation explained? Were questions open? Was any allegation disclosed in a question? Was observation separated from hearsay? Were others' accounts shared? Was an opinion requested? Were further witnesses asked about?

Allegations disclosed in questions is the measure. Every time an investigator tells a witness what's been alleged, the witness's independent account becomes less reliable.

For HR, employee relations and compliance teams

Witnesses often decide investigations. Independent observers can corroborate or undermine either side's account.

Leading questions weaken findings. Contaminated witness accounts are easy to challenge.

Honesty protects participation. Witnesses told the truth about confidentiality are more likely to cooperate in future investigations.

And the skill is universal. Every investigation involves witness interviews, and many investigators have never practiced one with a reluctant witness.

For law schools and HR programs, it completes the investigation-interview trio: complainant, respondent and witness.

Frequently asked questions

How do you interview a reluctant witness in a workplace investigation? Explain neutrally why they're being interviewed, answer confidentiality questions honestly, acknowledge their position, ask open questions about what they personally observed, and avoid leading questions or requests for opinions.

Can witnesses in an investigation remain anonymous? Often not completely. Investigators can limit disclosure to those who need to know, but anonymity may not be possible if a witness's account is relied on. Policies vary.

Should investigators tell witnesses what was alleged? Generally no more than necessary. Disclosing allegations in questions can shape the witness's memory and breach confidentiality.

What is the difference between firsthand and secondhand information in an investigation? Firsthand information is what the witness personally saw or heard. Secondhand information is what someone else told them — which may point to other witnesses but isn't direct evidence of the events.

The short version

"I don't want to get anyone in trouble. I didn't really see anything." Maybe. Or maybe she's deciding how much to say.

Tell her she's not in trouble. Tell her the truth about confidentiality and how retaliation is handled. Acknowledge that she works with both of them. Then ask open questions about what she saw herself, separate it from what she heard, and show her documents without accusation. Don't tell her what's been alleged. Don't ask what she thinks.

A reluctant witness becomes a useful one when the process gives her no reason to be afraid of the truth.

Foretell AI lets HR, employee relations and compliance teams build investigation simulations — including reluctant witnesses, confidentiality questions and non-leading interview practice like the one above — with configurable AI witnesses, recordings and rubric-based evaluation. If your investigators lead witnesses without realizing it, we're happy to walk through how other organizations have structured it.