Anonymous Employee Research with AI Interviews: Get the Honest Feedback Surveys Miss

Anonymous employee research, in one sentence

Anonymous employee research is a study where employees give candid feedback through interviews or surveys without their name, email, manager or team being attached to what they say. With Koji you share one interview link with no pre-interview form, turn on Remove personal details so the names people mention in their answers are replaced with labels before anyone reads them, and the AI moderator runs a 5–15 minute voice or text conversation with the depth of a 1:1, without the scheduling or the fear of being overheard.

If you have ever sent an "anonymous" survey through a tool that asks for an email at the start, you already know the problem: people do not believe it. Response rates drop, and the people who do respond water down their answers. An AI-moderated interview changes the dynamic: employees talk to a neutral interviewer that probes their answers gently, the way a skilled qualitative researcher would. What makes the promise believable is the setup, so this guide is precise about what Koji does and does not do.

Why anonymous matters more than ever

Gallup's State of the Global Workplace 2025 report found that global employee engagement fell from 23% to 21% in 2024. When engagement is that low, leaders need to hear the reasons, and those reasons are exactly what people hold back when they think they can be identified.

The traditional fix has been engagement surveys. They are easy to send, but they have three structural problems:

• Multiple choice cannot capture nuance. "Somewhat agree" tells you nothing about why.
• Open-text fields get one-line answers. No follow-up means no depth.
• People assume the data is not actually anonymous. Email links, small team filters and named comments all leak identity.

An AI interview addresses the first two directly: it holds a real conversation and asks follow-up questions automatically. The third depends on how you set the study up and how you report it. Koji gives you the pieces: a shared link with no form in front of it, an optional Anonymous badge, and a setting that removes personal details from answers. It is the same study builder, interviewer and report Koji uses for customer, churn and market research, so the People team works in the same tool as the rest of the company.

Anonymous or confidential: what Koji actually does

There is no single "anonymous mode" in Koji. Anonymity comes from three choices you make in the study's Branding & Settings, and it is worth knowing what each one does before you make a promise to employees.

1. No pre-interview form. Leave the Pre-Interview Form switched off. Participants go straight from the link to the conversation, and no name, email or department field is asked.
2. One shared link, not personal links. Everyone gets the same link. A personal link (one per person, from a CSV import) ties each response to a row in your list, which is right for a day-30 onboarding check-in and wrong for anything sensitive.
3. Remove personal details (Privacy section). Employees name their manager, their colleagues and sometimes themselves in their own words. With this setting on, Koji replaces people's names, email addresses, phone numbers, street addresses and the companies people work for with labels such as [NAME] and [COMPANY] when each interview finishes, before any analysis. Summaries, report quotes and exports are built from the cleaned text, and the original is not kept. Products and tools people mention stay, as do job titles, cities and countries, so findings keep their meaning.

The Anonymous badge (Branding & Settings, badges) is a label on the welcome page, with text you can edit. It tells participants what you promise; it does not change what is stored. Show it only when your setup keeps that promise.

Know the limits before you write the invitation:

• Remove personal details works only on interviews that finish after you turn it on. Turn it on before you share the link.
• It is automatic, so it can miss some details. Read responses before you share quotes beyond the People team.
• It removes direct identifiers, not every clue. A story only one person could tell, or a job title combined with a small team, can still point to someone. See Quasi-Identifiers in Research Data.
• Koji does not keep the voice recording of an interview where details were removed.

In practice that makes the study confidential by design and anonymous in reporting: nobody is asked who they are, names they say are taken out, and you report patterns, not individuals. Say that in your invitation in plain words; it is a promise employees can check.

A 6-step playbook for running anonymous employee interviews

Step 1: Pick the right research question

Anonymous interviews suit sensitive topics:

• "What would make you stay another two years?" (stay interviews)
• "What is the one thing leadership keeps getting wrong?" (culture audit)
• "Why do you think people on this team are burning out?" (retention risk)
• "If you could change one part of your manager relationship, what would it be?" (manager feedback)
• "What is your honest take on our DEI commitments: promises kept or just slogans?" (DEI pulse)

Write one sharp research question first. Koji's research assistant turns it into a full interview brief you can edit.

Step 2: Build the study with structured questions

Koji's structured question types let you mix open exploration with quantitative tracking in the same interview:

• Open-ended for the deep "why" questions
• Scale for engagement ratings (0–10 or 1–5)
• Single choice for forced trade-offs ("If we could only fix one thing this quarter...")
• Multiple choice for blockers ("Which of these gets in the way most?")
• Ranking for priority ("Rank what matters most about your job")
• Yes/no for clear binary signals ("Would you recommend this team to a friend?")

You get an engagement number and the quotes that explain it from one study. See the Structured Questions in AI Interviews guide for the full pattern library.

Step 3: Set up the study for confidentiality

Open the study's Branding & Settings:

1. Under Pre-Interview Form, keep the form off.
2. Under Privacy, turn on Remove personal details.
3. Under badges, turn on Anonymous and edit the label if you want different words, such as "Confidential". Only do this once steps 1 and 2 are in place.

Then share the study's single link. Avoid adding questions that ask for team, office or start date unless the study needs them; each one narrows the group a quote could come from.

Step 4: Distribute through low-friction channels

Share the single interview link via:

• An all-hands email from the CEO (not HR: leadership endorsement matters)
• A Slack #general post pinned for two weeks
• A link in your HR platform next to the regular pulse
• A QR code on break room posters for hourly workforces

Koji interview links never require a login. Employees click, talk to the AI for 5–15 minutes, and leave.

Step 5: Watch insights build in real time

As interviews come in, the study's report updates. You see emerging themes, quotes, sentiment and quality scores without waiting for someone to code transcripts. See Real-Time Research Insights for what to look for in the first 24 hours.

Step 6: Report up without breaking anonymity

Koji's report brings 30, 100 or 500 interviews together into a narrative with verbatim quotes, themes, scale distributions and recommendations. Before you share it, apply one rule by hand: if a theme comes from fewer than 5 people, leave it out of what you publish or fold it into a broader theme, so a single voice cannot be traced. Minimum Base Sizes for Segment Reporting explains why.

Share the report through Koji's public link, or export the data. Your CHRO sees the patterns; individual employees stay out of it. See Publishing & Sharing Reports.

Where anonymous AI interviews beat traditional employee surveys

Koji automates the parts that traditional engagement tools leave to people: follow-up questions, theme clustering, quote extraction and report writing. A People team of one can run a full anonymous engagement study end to end.

Designing questions that produce honest answers

Three principles, drawn from The Mom Test and qualitative interviewing practice:

1. Ask about the past, not the future. "Tell me about the last time you considered leaving" beats "Are you a flight risk?". The second is a yes/no nobody answers truthfully.
2. Ground every question in a specific story. "Walk me through your last 1:1 with your manager" gets you far more than "How is your manager?"
3. Let the AI probe. Turn on follow-ups for important open-ended questions. Koji's interviewer digs into the surface answer to find the real story. See the AI Follow-Up Probing guide.

Common pitfalls to avoid

• Don't fake anonymity. If you use personal links or a form that asks for an email, do not call the study anonymous. Employees will find out and trust collapses.
• Don't over-segment. Filtering to "engineers in Berlin who joined after 2023" can re-identify people in small samples. Set a minimum group size (for example 5) before showing breakdowns.
• Don't share raw transcripts widely. Even with names removed, a detailed story can identify its teller. Share themes and checked quotes.
• Don't skip the close-the-loop step. The quickest way to kill future participation is to collect feedback and never report back. Share what you heard and what will change.
• Don't rely on one annual study. AI interviews are quick to run quarterly, and continuous listening shows trends an annual snapshot misses.

When to escalate beyond anonymous interviews

Anonymous interviews are the right starting point, but some topics (harassment, discrimination, ethics violations) need a formal investigation channel with legal protections. Use Koji to see patterns and prevalence, then route specific allegations to your ethics hotline or legal team. See the Research Debrief Guide.

Pricing and access

The settings in this guide (no pre-interview form, the Anonymous badge and Remove personal details) are available on every account, Free included. A Free account starts with 10 credits. Each text interview costs 1 credit and each voice interview 3 credits, so a department-level study of about 30 text interviews uses 30 credits. Only interviews that pass Koji's quality gate use credits; see How the Quality Gate Works. Plans and credit packs are on the pricing page.

Related Resources

• Quasi-Identifiers in Research Data: why removing names is not the whole job
• AI Customer Research for HR Tech and People Ops Software
• Structured Questions in AI Interviews: the question types that mix qualitative depth with quantitative tracking
• Voice of the Employee Program: building a continuous employee listening program
• Stay Interviews at Scale: running retention interviews with AI moderation
• How the Quality Gate Works: why you only pay for high-quality responses
• Avoiding Bias in Research Interviews: keeping anonymous studies methodologically sound
• Koji for HR and People Teams: the full HR research playbook
• Your Consent Form Is Part of the Instrument: matching the promise to what the setup actually delivers

Further reading on the blog

• The Death of Static Surveys: Why AI Interviews Are the Future of User Research — Static surveys are suffering from <2% response rates. Learn why "Survey Fatigue" is killing your data and how AI-moderated interviews are ac
• Best Delighted Alternatives in 2026: AI Research Interviews Beat NPS Surveys — Delighted is shutting down June 30, 2026. Here are the 6 best alternatives — and why the smartest migration isn't a replacement, it's an upg
• Focus Groups vs Interviews: Which Research Method Gets You Better Data? (2026) — Should you run focus groups or one-on-one interviews? This guide compares both methods on depth, cost, bias, and speed — and shows why AI-mo

<!-- further-reading:blog -->