Workers see the storm. They can't afford an umbrella.

Findings02 · five patterns

Workers know AI is coming. Most are doing nothing about it.

hourly worker scrolling AI news headlines on a phone during an unpaid break, uniform still on, fluorescent break room, editorial muted

Awareness vs. action across 93 respondents

Aware of AI advances

79 of 93

Expect job change in 3 years

51 of 93

Feel prepared for AI

27 of 93

Zero upskilling actions taken

38 of 93

What we saw

85% of survey respondents report at least moderate AI awareness, and 55% expect AI to change their job in 1-3 years. But 41% have taken zero upskilling actions in the past twelve months — not a course, not a workshop, not even an article.

Counter-signal

The $35-60K earners are the engagement sweet spot: 31% attended a workshop, 27% pursued a certification — roughly triple the rate of under-$35K workers.

Why it matters

Awareness-only campaigns reinforce anxiety without resolving it. Every dollar spent raising awareness is a dollar diverted from the actual barrier: economics.

Design implication

Stop awareness advertising — it's already at 85%. Redirect the budget into on-the-clock 10-minute microlearning modules embedded in existing paid shifts. The intervention must cost workers zero dollars and zero personal time.

I am pretty scared of Artificial Intelligence. Requires absolutely no influence whatsoever on the way I prepare for my career.

— P25 · aspiring lawyer, Prolific

The root barrier isn't time or motivation. It's money.

a $200 online course price tag next to a paystub showing $14/hour, calculator beside them on a worn kitchen table, overhead editorial light, muted

Top upskilling barriers (Q9, multi-select)

Too expensive

33 of 93

Not enough time

24 of 93

Employer doesn't support it

17 of 93

Too tired after work

11 of 93

What we saw

Cost leads upskilling barriers at 35%, time at 26%, employer non-support at 18%. Under the surface both time and cost trace to the same root: economic precarity that makes discretionary learning structurally impossible for the 46% of the sample earning under $35,000.

Counter-signal

Only 8% say upskilling is unnecessary. The gap between knowing and acting is arithmetic, not psychological.

Why it matters

Survey instruments that offer 'time' as a socially neutral option capture the surface. Policy built on the time-barrier label prescribes scheduling fixes for what is actually an economic problem.

Design implication

Remove cost and the 'time barrier' collapses with it — for hourly workers they are the same barrier. Fund 15-minute paid learning slots during shift time, or accept that the lowest-income 46% will stay frozen regardless of messaging.

Time. At work forty one to forty two hours a week at a main job. Maybe work four days instead of five and get paid more to compensate me. Basically financial independence.

— P21 · customer support, Moen

Workers want employer-led AI training. Employers are absent.

empty conference room with 'AI Training' written on a dusty whiteboard, chairs still stacked, late-afternoon light through blinds, editorial

Where workers actually learn about AI

Social media

34 of 93

YouTube / podcasts

32 of 93

Colleagues

31 of 93

Company communications

6 of 93

What we saw

42% name employer-provided AI training during work hours as the #1 support they want. The top three Q10 responses are all institutional: employer training (42%), clear company AI policy (34%), government retraining (29%).

Counter-signal

Self-employed and freelance workers (8% of the sample) have no employer to ask. Their parallel need for platform or community training is unmet by any current channel.

Why it matters

The 36-point gap between what workers want (42% employer training) and what they get (6% company communication) is the institutional failure of the AI transition reduced to a single statistic.

Design implication

Audit how your org communicates AI to frontline workers. If the answer is 'we don't,' you are the 6%. Launch a quarterly 30-minute 'AI at [Company]' peer-led briefing during paid time — a conversation, not a webinar.

They say, oh, you will always have a job here, and we're not gonna have full AI integration. But do they really deliver that.

— P21 · customer support, Moen

Hands on the tool predicts security. Hands off predicts dread.

nonprofit food-bank worker drafting an email with an AI chat assistant visible on the laptop screen, warm daylight, focused posture, editorial

Feel more secure than a year ago, by AI usage

Daily AI user (n=23)

13 of 23

Few times per week (n=14)

7 of 14

Rarely (n=18)

2 of 18

Never (n=29)

6 of 29

What we saw

Daily AI users report feeling more secure at 57%. Rare users: 11%. The 46-point gap dwarfs every other correlation in the study. The qual confirms the mechanism: P01 (daily user) is the sample's most optimistic voice; P21 (passive, imposed contact only) is its most anxious.

Counter-signal

Causality is ambiguous — confident workers may adopt AI faster rather than AI building confidence. But 22 of 29 never-users cluster in the 'about the same' security cell, suggesting delayed reckoning, not protection.

Why it matters

The 29 never-users are not safe. They are late. When AI arrives in their workplace, they'll encounter it without prior experience — the exact condition that produces P21-style fear.

Design implication

Get tools into hands. Start with the 29 never-users: one AI tool relevant to their role, 15 minutes of guided first-use during paid time, measure security perception 30 days later. Direct exposure beats any reassurance message.

I'm not too worried about it. I'm actually hoping AI cannot take over everything, but take over a majority of the tasks because it will make our life ten times easier.

— P01 · nonprofit food distribution

Mixed-role workers are the AI canary. Already singing.

retail worker split-composition — half at a POS screen showing an AI inventory suggestion, half on the warehouse floor stocking shelves — editorial muted

Feel less secure than a year ago, by role type

Desk/office (n=19)

4 of 19

Frontline/hands-on (n=39)

15 of 39

Mixed desk+hands (n=35)

19 of 35

What we saw

Desk workers report 63% increased security; mixed-role workers report 26% increased and 54% decreased — the largest cross-tab gap in the study. The mechanism: AI automates the desk half of a hybrid job while offering nothing for the physical half.

Counter-signal

The qualitative strand predicted knowledge workers would feel most threatened. Quant reversed it: 79% of desk/office workers (15 of 19) feel the same or more secure — they experience AI as augmentation, not amputation.

Why it matters

A mixed-role retail worker whose mornings are inventory and afternoons are shelves doesn't become a more valuable shelf-stocker when AI eats inventory. They become a less valuable worker — their full-time status was justified by the differentiated half.

Design implication

Prioritise mixed-role workers in transition support. Build AI integration pathways that target the desk-adjacent tasks specifically, so the desk half becomes AI-augmented rather than AI-replaced. Frame it as 'AI makes the desk part faster so you bring more to the physical part.'

A robot could definitely take it over and do it for you. The only reason humans are still doing it is because customers feel the most comfortable with a human.

— P21 · customer support, Moen

From awareness to frozen

The five-step trajectory most workers in our sample are stuck somewhere inside — measured in our data, voiced in the real interviews.

Hear about AI

85% report at least moderate awareness. The information is everywhere.

Sense the change coming

55% expect their job to change within three years. The clock starts in the head.

Run the numbers

A $200 online course plus eight hours of lost wages. The arithmetic breaks for 46% earning under $35K.

Wait for the employer

42% want employer-led training. Only 6% get any company communication at all.

Freeze in place

41% have done nothing in 12 months. Not a course, not an article, not a question to a colleague.

What to build differently.

Five design moves that would change the relationship between the user and the score.

Embed AI in paid work time

Cost is the #1 barrier (35%) for a sample where 46% earn under $35,000. Off-hours courses, paid certifications and self-directed learning will not reach the population that most needs them. The intervention must be free, on-the-clock, and 15 minutes or less.

Route through peer channels

Company communications reach 6%. Social media and colleagues reach 33-37%. Identify one AI peer champion per team, pay them, and call it 'sharing' rather than 'training.' P21 would attend a 20-minute chat with a coworker. He would not attend a webinar.

Serve the mixed-role canary first

Mixed-role workers report 54% decreased security — the study's highest threat signal. Build integration pathways that augment their desk-adjacent tasks specifically, so AI preserves differentiated value rather than amputating it.

Close the trust-utility gap

54% trust AI but only 28% say it makes work easier. Audit deployed tools against the jobs frontline workers actually do. Design 'draft and review' workflows, not 'generate and send.' Meet workers at their real workflow, not the one the tool designer imagined.

Build role-specific pathways

14% don't know where to start. 10% say available training isn't relevant to their role. Build three pilots — frontline, mixed, desk — each with 5-7 modules under 15 minutes, each teaching one role-specific AI skill. Measure security perception at 30 and 90 days.

They see the storm.
They can't afford umbrellas.

A narrated brief of the three load-bearing findings

The same axis, from fear-passive to hands-on

The Frozen Canary

The Fear-Passive

Economically Compelled

Hands-On Optimist

Workers know AI is coming. Most are doing nothing about it.

The root barrier isn't time or motivation. It's money.

Workers want employer-led AI training. Employers are absent.

Hands on the tool predicts security. Hands off predicts dread.

Mixed-role workers are the AI canary. Already singing.

From awareness to frozen

03 · In their words

What to build differently.

Embed AI in paid work time

Route through peer channels

Serve the mixed-role canary first

Close the trust-utility gap

Build role-specific pathways

Methodology