Set rules for sharing survey results
Your AI barrier survey is finished, and a manager wants employee survey results for every team, plus the comments. Before you prepare that briefing, decide what employees could work out from the whole release.
We think transparency should explain the findings and reporting decisions, not expose every breakdown. Detailed reporting has a place, but sharing survey results needs a review of who could recognise whom.
The goal is a briefing of employee survey results and a private release decision log. The workflow below is our editorial advice informed by official guidance; it doesn't guarantee anonymity, legal compliance or prevention of disclosure. Assemble these prerequisites before starting.
Sources: Privacy notice: Civil Service People Survey respondents - GOV.UK, How do we ensure anonymisation is effective? | ICO, Anonymisation for text data - UK Data Service
Step 1: Choose defensible reporting groups
The Cabinet Office's Civil Service People Survey privacy notice, updated September 19, 2025, sets a minimum of 10 people answering the same question for reporting, with separate arrangements for comments. That's one survey's policy, not a universal safe threshold or a confidentiality guarantee.
Set a documented local reporting rule, then assess context. ICO guidance supplied on October 18, 2026, says to consider both the information and who may access it; that guidance is under review following the Data (Use and Access) Act.
ONS guidance supplied on October 18, 2026, includes combining categories for births and deaths tables in England and Wales. The following decision table is our editorial application to employee reporting, not a transfer of ONS thresholds. Record what each choice preserves or loses.
| Situation | Suggested decision | Trade-off |
|---|---|---|
| Broad group passes contextual review | Publish the result | Retains group detail |
| Small team | Merge into a meaningful larger group | Loses team specificity |
| Distinctive demographic combination | Broaden categories | Loses subgroup detail |
| Breakdown remains revealing | Withhold it | Retains less detail |
Sources: Privacy notice: Civil Service People Survey respondents - GOV.UK, How do we ensure anonymisation is effective? | ICO, Policy on protecting confidentiality in tables of birth and death statistics - Office for National Statistics
Step 2: Review the whole release
Ask, "Could someone reconstruct the hidden result from the total and the other teams?" Employee Feedback's July 19, 2026, vendor article recommends withholding an additional figure when subtraction would reveal a suppressed group.
We'd review employee survey results as one release, including earlier reports, rather than approve each table separately. Use this suggested sequence before signing off.
- Gather totals, team tables, demographic cuts and previous reports.
- Check whether their combinations reveal withheld results.
- Broaden a breakdown or remove another revealing figure.
- Repeat the review and log the final decision.
Sources: Anonymous or confidential – what your employee survey actually promises
Step 3: Preserve meaning in paraphrases
UK Data Service guidance dated August 5, 2026, recommends reducing specificity while preserving analytical meaning, and recording text changes in an anonymisation log. Its guidance concerns qualitative data generally; applying it to AI survey comments is an editorial recommendation.
For a hypothetical comment about a unique client incident, replace the distinctive role, client and timing with a broader description: a respondent reported uncertainty about using client information with AI. Label it as a paraphrase and keep it outside quotation marks.
UK Data Service's same guidance says automated anonymisation needs careful review. Ask, "Would colleagues still recognise the event or person?" Generalise further or omit the comment if retaining its useful meaning would remain revealing.
Sources: Anonymisation for text data - UK Data Service
Step 4: Brief employees on approved themes
Use this suggested fill-in template only with your actual results and approved release decisions: During [survey period], respondents reported [supported barrier theme]. These are self-reported perceptions, not evidence of causal productivity effects or intervention effectiveness.
Continue: To reduce identification risk, this briefing combines [groups] and withholds [breakdowns]. Comment themes are paraphrased. [Owner] will review [proposed intervention], with an update on [date].
AI Adoption Score's AI Adoption Barriers survey identifies concrete obstacles and who can remove them. It has no minimum group size for department columns, so a department of one shows that person's answers. Review the output before choosing what to share.
Release an explanation you can defend
Your expected result is a briefing employees can understand and a private log explaining each publication, grouping, omission and paraphrase. If a manager asks for finer detail, offer the broader actionable theme and explain the reporting decision.
In our view, a useful briefing earns its detail through review, not through pressure to publish everything.
Frequently asked questions
Who should approve the employee briefing?
We suggest assigning one accountable release owner, supported by the survey lead and whoever owns employee confidentiality commitments. Ask a reviewer familiar with the workforce to challenge recognisable details. Keep the approval focused on the exact briefing and recipient group, with unresolved decisions recorded before publication.
Should managers receive the same version as employees?
Decide access by purpose rather than seniority alone. As editorial guidance, prepare a recipient-specific release decision and explain why any additional detail is needed. ICO guidance says identification assessments should consider who may access the information, so approval for one audience shouldn't automatically authorise wider circulation.
What should the private decision log contain?
Include the proposed content, intended audience, decision, reason, reviewer and approved version. For edited comments, record the changes separately with restricted access. This suggested structure extends UK Data Service's recommendation for an anonymisation log into a practical release record; it isn't a prescribed or validated template.
Sources: How do we ensure anonymisation is effective? | ICO, Anonymisation for text data - UK Data Service
