How to Build DEI Surveys That Drive Meaningful Change

DEI surveys are among the highest-stakes feedback programs any organization can run. Get them right, and you surface systemic issues that enable genuine improvement. Get them wrong, and you create cynicism, damage trust, and waste the emotional labor of the people sharing vulnerable feedback.

The fundamental challenge: the employees whose experiences matter most (those facing bias, exclusion, or inequity) are the least likely to trust a corporate survey with their honest feedback. Traditional DEI surveys suffer from social desirability bias at extreme levels. People give the "right" answers, not the real ones. And when marginalized employees do share honest feedback, they fear being identified, especially in small teams.

Koji addresses these challenges by providing an anonymous AI interviewer that employees consistently rate as safer to be candid with than any human-administered survey or focus group.

Core DEI Measurement Dimensions

Belonging

"Do I feel like I belong here? Can I be my authentic self at work?"

Inclusion

"Are diverse perspectives sought, valued, and incorporated into decisions?"

Equity

"Are opportunities, resources, and outcomes distributed fairly regardless of identity?"

Psychological Safety

"Can I speak up, disagree, or report concerns without fear of retaliation?"

Representation

"Do I see people like me in leadership? In decision-making roles?"

Systemic Barriers

"Are there policies, practices, or norms that disadvantage certain groups?"

Building DEI Studies with Koji

Study Structure

Q1: Belonging (Scale, 1-10) "On a scale of 1-10, how strongly do you feel you belong at this organization?"

• Labels: 1 = "Not at all", 10 = "Completely"
• Anchor probing: "What contributes most to that feeling?"
• AI instruction: "Be gentle and empathetic. Create space for both positive and negative experiences."

Q2: Authentic Self (Open-ended) "How comfortable do you feel being your authentic self at work?"

• Probing depth: 3
• AI instruction: "Explore specific dimensions: identity, communication style, work approach, personal life. Don't assume or project specific identities."

Q3: Inclusion in Decisions (Scale, 1-5) "How well are diverse perspectives included in team decisions?"

• Labels: 1 = "Not at all", 5 = "Consistently"
• Probing: "Can you think of a time when diverse perspectives were or weren't included?"

Q4: Fair Treatment (Open-ended) "Do you feel that everyone here has equal access to opportunities for growth and advancement?"

• Probing depth: 3
• AI instruction: "Explore specific systems: promotion, project assignment, mentorship, visibility. Look for patterns, not just individual incidents."

Q5: Psychological Safety (Scale, 1-5) "How safe do you feel speaking up about problems or disagreeing with leadership?"

• Labels: 1 = "Very unsafe", 5 = "Very safe"
• Probing: AI explores specific situations where they did or didn't speak up

Q6: Microaggressions (Open-ended) "Have you experienced or witnessed any behaviors that made you or others feel excluded or devalued?"

• Probing depth: 2
• AI instruction: "Handle with extreme sensitivity. Don't push if the person seems uncomfortable. Validate their experience."

Q7: Systemic Barriers (Open-ended) "Are there any policies, processes, or unwritten rules that you feel create disadvantages for certain groups?"

• Probing depth: 2
• This question surfaces institutional issues rather than individual behaviors

Q8: Leadership (Scale, 1-5) "How committed do you believe leadership is to DEI?"

• Probing: "What gives you that impression? What actions (or lack of actions) shape your view?"

Q9: Improvement (Open-ended) "What's the single most important thing this organization could do to be more inclusive?"

• Probing depth: 2

Q10: Progress (Single Choice) "Compared to a year ago, how has the DEI culture here changed?"

• Options: Significantly improved / Somewhat improved / About the same / Somewhat worse / Significantly worse
• Probing: "What specific changes (or lack of changes) lead you to that assessment?"

Demographic Collection (Optional, Handle with Care)

If collecting demographic data for intersectional analysis:

• Make it entirely optional
• Explain exactly why it's being collected and how it will be used
• Use broad categories to prevent identification in small teams
• Never cross-tab demographics with responses in teams smaller than 15

Why AI Interviews Are Essential for DEI Research

Safety

Employees from marginalized groups face real risks when sharing DEI feedback. With HR, they worry about being labeled "difficult." In focus groups, they perform for the room. With an AI interviewer, there's no one to judge, no one to gossip, no one to retaliate. The research consistently shows that people share more candidly about sensitive topics with AI than with human interviewers.

Anonymity

Koji's anonymous mode collects zero personally identifiable information. No IP addresses, no login required, no device tracking. This is the level of anonymity needed for employees to trust the process.

Consistency

Every employee gets the same thoughtful, patient, empathetic interview. No variation based on the interviewer's own identity, biases, or comfort with sensitive topics.

Scale

Organizations can't run 1,000 individual DEI focus groups. Koji can conduct 1,000 deep conversational interviews simultaneously, giving every employee an equal voice.

Emotional nuance

Voice interviews capture tone, hesitation, and emotion that text surveys miss entirely. For DEI topics, these signals are often more revealing than the words themselves.

Analysis and Action

What Koji Reports Generate

• Belonging index with distribution and trend tracking
• Dimension heatmap showing strongest and weakest DEI areas
• Theme analysis organized by DEI dimension with supporting quotes
• Barrier identification listing specific policies, practices, and norms flagged by employees
• Intersectional analysis (if demographic data is collected) showing how experiences differ across groups
• Progress tracking comparing results to previous waves

The DEI Action Framework

1. Share results transparently (aggregate only). Trust is built through transparency.
2. Acknowledge what you heard. Don't spin or minimize. If employees said belonging is low, say so.
3. Commit to 2-3 specific actions with timelines. Vague commitments destroy credibility.
4. Follow through publicly. Report progress at 30, 60, and 90 days.
5. Re-measure to confirm impact. Run a pulse survey 6 months later on the specific areas you addressed.

Best Practices

Don't survey without commitment to act

The single biggest mistake in DEI research is asking for feedback and then doing nothing. It's worse than not asking at all, because it signals that the organization doesn't take DEI seriously despite claiming to.

Include everyone, not just marginalized groups

DEI is everyone's experience. Understanding how different groups experience the same organization reveals the gaps. Majorities often rate inclusion higher than minorities. That gap is the data.

Use open-ended questions over scales

Scales are useful for tracking trends, but the real value in DEI research is in the stories, experiences, and specific barriers that only come through conversation.

Run it regularly

Annual is minimum. Quarterly pulses on specific dimensions keep the conversation alive and signal ongoing commitment.

Protect respondent confidentiality above all else

If even one employee feels their responses were traced back to them, you will never get honest DEI feedback again. Err on the side of less demographic granularity.

Why Koji Is the Best Tool for DEI Surveys

• AI reduces social desirability bias more than any other methodology
• True anonymity with zero PII collection
• Sensitive probing that creates space for vulnerable sharing
• Voice option captures emotional nuance
• Multi-language support for global and diverse workforces
• Consistent quality every employee gets the same empathetic interview
• Scale to give every employee a voice, not just a sample
• Automated analysis that themes sensitive responses without human readers seeing individual answers

DEI measurement deserves more than a Likert scale. It deserves a conversation.