What your workforce actually thinks about AI
A sample Alignment Index baseline report for Halford & Wren, a made-up firm. The people and figures are made up to show what your report would tell you.
Headline findings
Five findings from the baseline, before any training.
Alignment Index: baseline reportSample report- Prepared forHalford & Wren Consulting
- Baseline window15 Apr to 12 May 2026
- Cohort780 UK employees
- Took part642 (82%)
-
62%
More anxious than enthusiastic about AI
62% feel anxious about AI or fear being left behind in their role. Only 6% are actively enthusiastic.
-
34%
A third use AI for work without telling their manager
34% say they use AI their firm hasn’t sanctioned. 28% haven’t told their manager. Nothing at the firm yet makes AI use feel safe to admit.
-
73%
Most can’t yet spot a hallucination
Shown a work scenario with a hallucination risk, 73% chose wrongly or weren’t sure.
-
38%
Confidence is running ahead of demonstrated skill
38% think AI is more accurate than it is, and a similar share rate their own judgement above what they can show. Completion data can’t see this, because the training looks done.
-
9%
A small group are already confident and capable
9% use AI safely and well, with attitude and capability both high. They aren’t yet visible or connected across teams.
Sample report with a made-up organisation and figures, to show the format.
Measure 1 of 5
Attitude
How people feel about AI: trust, security and willingness to engage.
The lower of the two scores, held down by fear of being left behind. The dominant feeling is fear of not being good enough, not resistance. People aren’t pushing back on AI. They’re worried about being judged for not yet using it well.
So communications should avoid words that feed the worry (“transformation”, “the future of work”) and say that training starts where each person is.
How people feel about AI
Share of people who chose each. One person can choose more than one. Illustrative figures.
Measure 2 of 5
Capability
Whether people understand AI, can use it well for their work, can spot when it’s wrong, and know what’s safe.
Capability sits above attitude: people are more able than they are yet willing or disciplined. These are the gaps by topic.
- 73%
Can’t identify a hallucination risk
Shown three short scenarios and asked which holds a hallucination risk, 73% chose wrongly or said “I’m not sure”.
- 68%
Don’t know the firm has an AI policy
Of those who did know, fewer than half could say what it says.
- 54%
Unclear what data they can share with AI tools
The lines around client data, internal data and personal data aren’t clear.
- 47%
Can’t say what the EU AI Act asks of them
People know the Act exists, but not what it means for them personally.
- 38%
Believe AI is more accurate than it is
Over-trust is as dangerous as under-trust.
The gap, team by team
Short scored scenarios measure what people can show. Set against how capable they say they are, that gives the gap. Where belief runs ahead of what people can show, that over-confidence is where the risk sits, and it shows up as hidden AI use and acting on AI without checking it.
Here, confidence runs well ahead of tested capability in Sales and Finance, which is where training should start. Legal is the reverse: more capable than it feels.
Alignment Index: by teamSample reportSample report with made-up teams and figures, to show the format.
Measure 3 of 5
Perceived alignment
Whether people believe leadership is handling AI in their interest.
Asked separately from how people feel about AI itself, and kept apart from the attitude score. Set it beside what leadership intends.
Illustrative figure.
Measure 4 of 5
AI usage
What people are actually using AI for at work, and how.
The headline. Use is mostly writing and summarising: the low-risk, everyday tasks where confidence grows. Encourage these as ways in, and put guardrails around the higher-risk ones, client work and data analysis.
The unseen risk. 14% of AI users draft client deliverables with AI, and few have told their project leads. This is the highest-exposure pattern in the data.
What AI is being used for
Share of people who use AI, not of the whole workforce. Illustrative figures.
Measure 5 of 5
Honest disclosure
Whether people have misused AI. Anonymous, so they answer truthfully. These are the behaviours people wouldn’t tell their manager or HR about. Share of everyone who took part (642).
| Behaviour | Share |
|---|---|
| Used a public AI tool with client-identifiable information in the last 60 days | 34% |
| Hasn’t told their manager about their AI use | 28% |
| Acted on AI output without checking it | 19% |
| Pasted internal documents into ChatGPT or a similar public tool | 11% |
| Used AI to produce client deliverables without saying so | 7% |
Illustrative figures.
What to do with this
These five aren’t edge cases. They come from the same place: people doing more with AI than their trust, discipline and the firm’s signals have caught up with. They carry on unless the conditions behind them change. Measure again after training, and what’s left is for policy and process to fix.
Recommendations
Five moves for the training, and two for Halford & Wren beyond it.
Lead with confidence, not skills
Fear of being left behind is the strongest feeling in the data (62%). Tell people the training starts where they are and builds confidence before skills.
Make the AI policy known
68% don’t know the firm has an AI policy. Put it where people can ask about it, at the moment they need it.
Hold a manager-led conversation
28% haven’t told their manager about their AI use. That’s a safety problem, not a knowledge problem. A 15-minute structured conversation between each manager and their team would bring it into the open.
Refer three risk patterns to legal
Three patterns in the data need a policy decision. We flag them so legal and compliance can judge whether the current policy covers them.
Make your confident, capable people visible
The 9% aren’t yet known or connected. An opt-in way to recognise them, and a cross-team channel, would let them help everyone else.
Beyond the training
Look at the signals around trying AI
The 28% who haven’t told their manager won’t change through training alone. That changes when people believe trying AI is rewarded, not penalised. Review internal messages, performance frameworks and unwritten norms over the next quarter.
Treat EU AI Act readiness as policy and process too
Training closes most of the knowledge gap on the Act. It doesn’t close the policy and process gap. Take the gaps in this report to your compliance team.
What happens next
Attitude and capability are scored separately and measured again on the same questions. The aim is both rising, with attitude catching up so growing capability is matched by trust and discipline.
Week 8
In-flight report
Who’s taking part, and themes that need attention part way through.
Week 12
Outcome report
How both scores and the gap have moved, the risk patterns that remain, and what to do in the next 90 days.
Quarterly
Check-ins
Short follow-ups track how behaviour drifts over time.
On request
Specific questions
How your people see a new AI policy, tool or incident.
How the data was collected
Anonymous by design, and reported at group level only.
See your own report. Test 50 people, free
Run a free Alignment Index for up to 50 people and see how your workforce compares with the 1,000 UK workers we tested. Results in a fortnight.