AI Workplace Trust, Global Knowledge Workers 2026

Methodology

This study draws on a simulated panel of 54 US and UK knowledge workers in analyst, marketing, consulting, technical and operations roles who use generative AI tools at least weekly at work. Each respondent is a Minds persona calibrated against historical workforce data, AI-tool adoption signals and role-specific output expectations. Accuracy against held-out human responses validates at 80-95% on the underlying behavioural prompts.

The full unlocked study includes 14 cross-tab statistics by role, country and primary AI tool, downloadable charts, the raw response CSV, and unrestricted follow-up question access to the panel.

Undisclosed AI use is now the default, not the exception

96% of the panel acknowledge submitting AI-generated or substantially AI-assisted work to a manager or client without explicit disclosure. Only 4% say no. Disclosure norms have not kept pace with adoption; the workplace contract has been quietly rewritten without a single corporate policy change to mark the transition.

Open-text reasoning is unambiguous. Respondents do not frame the omission as deception. They frame it as a rational reading of what their employer rewards: output quality, deadline compression and judgement on the final draft. Disclosure of the tooling is, in their view, the equivalent of declaring that they used a search engine or a calculator. The asymmetry is that the employer assumes a craft-mode workflow that the worker has already replaced.

Job-replacement fear is low and bounded

When asked to score, on a 0-10 scale, their fear that AI will replace their role inside five years, the panel averages 3.1/10. 2% sit at 7 or above; 80% sit at 3 or below. The headline is not stoicism, it is judgement. Respondents distinguish between the parts of their job AI handles competently today (drafting, summarising, code-completion) and the parts they read as durably human (stakeholder management, judgement calls, accountability for the outcome).

The high-fear minority cluster in roles where the AI-handled fraction is already above 60%, copywriters working on commodity content, junior analysts producing routine reports. The low-fear majority report that AI raised their floor and freed them to do the work they were hired to do.

Self-reported productivity gains are almost universal

98% of the panel report that AI made them more productive at work, with the remainder reporting no change. Nobody in the sample reports that AI made them less productive. The headline understates the variance in what "more productive" means: some respondents report 2x throughput on the same scope, others report flat throughput but a meaningful increase in the cognitive surplus they spend on harder work.

The productivity gain is real but it is decoupled from disclosure. Respondents who do not disclose AI use are the same respondents claiming the productivity uplift. The employer captures the throughput gain and never sees the tooling that produced it. The next workplace negotiation, on training, on tooling budgets, on how this productivity is shared, will be conducted in the dark unless the disclosure norm shifts.

What this means for executives and people leaders

• The disclosure question is the policy question. 94% non-disclosure is a governance gap that audit, compliance and customer-trust will eventually surface. A clear, low-friction disclosure norm, ideally one that respects worker judgement rather than penalising tool use, closes the gap before the first lawsuit does.
• Train at the AI-handled boundary, not around it. Productivity gains land where workers can clearly demarcate which fraction of the job is now AI-handled and which is not. Role-by-role training that names that boundary is the highest-leverage intervention.
• Buy back the disclosure with shared productivity. Workers will disclose tool use when disclosure is rewarded rather than punished. Sharing the productivity gain, in time, in tooling budget, in promotion criteria, converts a hidden subsidy into a visible negotiation.

The full study includes the role-by-role disclosure breakdown, the productivity-gain distribution by tenure, the "would you switch jobs to a more AI-friendly employer" follow-up, and the open-text corpus. Sign up free to unlock and ask the panel your own follow-up questions.