Anxious about the next harassment complaint that lands in your inbox. Good. The fastest way to burn cash is to stumble through an investigation, let chat messages disappear, and hope no one asks for your process. Once a lawsuit is on the horizon, courts can penalize you for evidence you fail to keep. The better way is simple. Lock down the evidence, run a clean, credible investigation, and use modern tools without letting them drive the bus. In other words,even if you use AI to sort and summarize, keep people in charge of the actual judgement calls. That is how legal risk management can support better business decisions and protect your bottom line.

What changed, and why you should care

Federal AI policy has whipsawed. Agencies removed several AI-specific employment resources in 2025, which means fewer training wheels and more responsibility on you to get your own house in order. Title VII and other applicable employment laws still apply, including when employers use AI-assisted tools in employment decisions. Translation for leaders and HR: you cannot point to a tool and shrug. You still own the decisions.

The EEOC rescinded its 2024 Enforcement Guidance on Harassment in the Workplace in January 2026. That did not erase federal laws prohibiting unlawful workplace harassment or the Supreme Court precedent governing employer liability. Employers still need effective complaint procedures and prompt, appropriate responses when harassment concerns are raised. Those two steps are the heart of the employer defense the Supreme Court recognized in 1998. What is gone is the detailed federal playbook so your own process has to carry more weight.

The move that can save you money

Preserve the chats now. Once a complaint makes a lawsuit foreseeable, put the people involved on legal hold and pause any automatic deletion settings. Microsoft Teams content can be discoverable, and Microsoft provides tools for preserving, collecting, reviewing, and exporting Teams content through Purview eDiscovery. Without a hold, you can lose earlier versions of edited messages and the history of reactions.If your team cannot find the Teams chat, opposing counsel may. The employee on the other end of the chat may have kept a copy.

Know the spoliation rule. If electronically stored information that should have been preserved in anticipation or conduct of litigation is lost because reasonable preservation steps were not taken, and it cannot be restored or replaced through additional discovery, Federal Rule of Civil Procedure 37(e) provides potential remedies if the loss hurt the other side and, when there is an intent to deprive another party of the information’s use, more serious sanctions. Your preservation process may need to account for chat, meeting recordings, reactions, and side channels depending on the matter. State courts, including Texas, apply their own spoliation rules.

Use AI like an associate, not an oracle

AI is useful for discrete investigation tasks, like drafting interview summaries, creating evidence tables, and reorganizing notes. It works best with a disciplined prompt. One popular method is RISEN, which stands for Role, Instructions, Steps, End Goal, and Narrowing. That sequence keeps outputs usable and manageable for real investigations.One rule comes first: use an enterprise tool your company controls, not a free public chatbot. Investigation files are full of sensitive personal information.

Here is a classic risk. An AI summary flipped a witness timeline, stating an event happened three months after a hire when it was actually three months before the interview. That is not a harmless glitch. A date error can poison credibility findings, which then poisons your outcome. The cure is the boring one. Verify before you insert.Also treat your prompts and AI drafts like any other investigation notes. They may be discoverable

Want a concrete model? Ask your AI tool to group evidence by complainant, respondent, witnesses, and documents, then separate agreement from conflict. Use that structure, but confirm every line against the source record before it lands in your findings. An associate’s work gets checked. So does AI’s.

Harassment liability in one paragraph you can use

The Supreme Court’s Faragher and Ellerth decisions established important rules governing employer liability for supervisor harassment. A supervisor, for this purpose, is  someone the employer empowers to take tangible employment actions like hiring, firing, failing to promote, or significantly changing duties or benefits.When supervisor harassment results in a tangible employment action, the employer may face vicarious liability. In certain hostile-work-environment cases without a tangible employment action, an employer may have an affirmative defense based on reasonable efforts to prevent and promptly correct harassment and the employee’s unreasonable failure to use available preventive or corrective opportunities.

The practical point has not changed: maintain effective complaint procedures, respond promptly, investigate appropriately, and take corrective action when warranted. Also, be aware that state law can be even stricter. For example in Texas employers of any size can be liable for sexual harassment if they or their supervisors knew or should have known about it and failed to take immediate and appropriate corrective action. Supervisors may also face personal liability. 

Technical takeaways you can apply this week

1) Build a repeatable prompt kit. RISEN is a practical scaffold. Assign the role, give clear instructions, list steps, state the end goal, and narrow for tone, length, and audience. Then break work into small tasks. Discrete tasks produce better outputs than vague, mega prompts. Save prompts that work so your team can reuse them.

2) Treat “summarize, verify, insert” as a rule. Use that order for chat threads, meetings, and interview notes. It keeps reports readable and helps reduce avoidable errors. Verify means checking every name, date, and quote against the source.

3) Expect hallucinations, and design around them. AI tools can produce wrong details and confident nonsense. Build verification steps into your template, and never let AI write credibility findings or legal conclusions on its own.

4) Thread the chats. Conversation reconstruction matters.One message pulled out of context can look worse, or better, than it was.  Use available eDiscovery tools so you see the chat in context, not as a jumble of exported lines.

5) Keep your AI governance lightweight and real. NIST’s AI Risk Management Framework and its profile for generative AI provide a voluntary framework for identifying and managing AI risk. NIST is revising the framework, so check for the current version. Use those concepts to help document how HR and Legal review data sources, assess risks, and maintain appropriate human oversight.

6) Remember AI does not erase civil rights law. Even as agencies adjust public guidance, discrimination laws did not vanish. Hiring tools, monitoring tools, and analytics still need to be used consistently with applicable federal and state employment laws Some states go further. Illinois now regulates AI in employment decisions, and Texas bars using AI with intent to unlawfully discriminate.. Keep that guardrail in view.

Monday morning checklist for HR and operations

• Determine whether a preservation obligation exists and, when appropriate, issue a written hold that covers relevant Teams chats, meeting recordings, channel posts, direct messages, attachments, and shared drives tied to the matter. Coordinate preservation with IT. Include personal phones and texts if employees use them for work, and document when and to whom you sent the hold.
• Use an intake template that identifies everyone who saw or heard the conduct, and every system that might contain related messages.Note when the company first learned of the conduct.
• Draft interview questions with the RISEN method, but verify quotes and dates directly against your notes or transcripts before you finalize summaries.
• Split your draft report into the four buckets : complainant, respondent, witnesses, and documents. Close with agreement versus conflict. Then have counsel test the legal analysis against the applicable framework.
• Remind everyone involved, in writing, that retaliation is prohibited. 
• If the complaint involves national origin, review the EEOC’s national origin discrimination resources and confirm your policy and process against the applicable legal requirements. National origin protections cover everyone, including American workers.

The business case, without the fluff

Every hour you can responsibly shave off the investigation is an hour your managers get back to run the business. Clean evidence handling can reduce discovery fights and spoliation issues later. A tight report grounded in the evidence gives counsel a better record to work with if the matter becomes a charge or lawsuit. That is how good investigation practices k support the business in an environment that punishes sloppiness.

What we do for employer clients

Treaty Oak Employers’ Law Group is a Texas law firm with attorneys licensed in Texas, North Carolina, Wyoming, and Colorado that represents employers. We design practical policies, run complex investigations, help employers address preservation and eDiscovery workflows, and train HR and managers to use AI responsibly without outsourcing judgment. When it gets messy, we defend.

If AI is already finding its way into your investigation process, make sure your process is keeping up with it. We can help you build an investigation workflow that uses the technology without outsourcing the judgment.