Your AI strategy was probably built on assumptions about what employees and customers want from artificial intelligence. Anthropic just surveyed 80,508 people across 159 countries — and the results suggest those assumptions are wrong. The biggest finding isn’t about productivity or automation. It is about professional excellence, personal growth, and time freedom. If your AI rollout focuses purely on efficiency metrics, you are solving for the wrong outcome.
The study that rewrites the playbook
In December 2025, Anthropic ran what it describes as the largest multilingual qualitative study on AI aspirations ever conducted. Open-ended interviews with 80,508 Claude users across 70 languages, processed by AI classifiers to identify patterns at scale.
Strategic Reality: This is not a satisfaction survey or a product feedback form. It asked people what they hope AI will do for their lives — a question most enterprise AI strategies never bother to ask.
The methodology matters because it captures something unusual: genuine aspiration rather than feature requests. And the patterns that emerged challenge the standard boardroom narrative about AI adoption.
What people actually want
| Priority | Share of respondents | What it means |
|---|---|---|
| Professional excellence | 18.8% | Offload routine work to focus on strategy |
| Personal transformation | 13.7% | Growth, emotional wellbeing, life change |
| Life management | 13.5% | Organisational support, cognitive scaffolding |
| Time freedom | 11.1% | Reclaim time for relationships and leisure |
| Financial independence | 9.7% | Economic security through AI-enabled work |
| Societal transformation | 9.4% | Address systemic problems (poverty, climate, health) |
| Entrepreneurship | 8.7% | Build and scale businesses with AI as co-founder |
| Learning and growth | 8.4% | Accelerated knowledge acquisition |
| Creative expression | 5.6% | Bring artistic visions to life |
The top three priorities — professional excellence, personal transformation, and life management — account for 46% of all responses. None of them map cleanly to “increase output by X%.”
Critical Context: Only 18.8% of respondents prioritised professional excellence as their primary AI aspiration. The remaining 81.2% want something more personal — growth, time, independence, meaning. Most enterprise AI strategies address less than a fifth of what people actually care about.
Where AI already delivers — and where it falls short
The gap between aspiration and reality is where strategic opportunity lives. Anthropic’s data shows AI already working well in specific areas:
- Productivity gains: 32% report significant acceleration in their work
- Cognitive partnership: 17.2% use AI as a thinking collaborator, not just a task executor
- Learning support: 9.9% experienced measurable skill development
- Technical accessibility: 8.7% built projects they previously lacked the skills to attempt
- Research synthesis: 7.2% process complex information more effectively
- Emotional support: 6.1% found value in judgement-free conversation
Success Factor: The 17.2% using AI as a cognitive partner — rather than a productivity tool — report deeper satisfaction with the technology. This is a fundamentally different use case from automation, and one that most deployment strategies overlook entirely.
The productivity number (32%) is strong. But look at the distribution: cognitive partnership, learning, and technical accessibility together represent 35.8% of where AI delivers value. These are capability-building use cases, not efficiency ones.
For UK organisations, this has a direct implication. If your AI deployment is measured solely on time saved or tasks automated, you are missing the majority of the value your people are already finding.
The five tensions your workforce is living with
Anthropic identified a pattern it calls “light and shade” — five recurring tensions where benefits and risks coexist within the same individuals. Not across different groups. Within the same person.
| Tension | The benefit | The fear |
|---|---|---|
| Learning vs. cognitive atrophy | AI accelerates skill acquisition | Over-reliance erodes existing abilities |
| Decision support vs. unreliability | AI improves analysis quality | Hallucinations create verification burden |
| Emotional support vs. dependency | Judgement-free conversation space | Reduced human connection |
| Time saving vs. illusory productivity | More output in less time | Busy work expands to fill freed capacity |
| Economic empowerment vs. displacement | New income opportunities | Existing roles made redundant |
Strategic Insight: These tensions are not bugs in AI adoption — they are the lived experience of people using the technology daily. Strategies that acknowledge only the benefit side (or only the risk side) will fail to earn trust from the people expected to use these tools.
One respondent captured it precisely: “I use AI to review contracts, save time… and at the same time I fear: am I losing my ability to read by myself?”
This is the reality of AI adoption in 2026. Not enthusiastic champions or fearful resisters, but conflicted practitioners holding both positions simultaneously. Any change management programme that treats adoption as a binary — resisters vs. adopters — misreads the room.
What keeps people awake at night
The concerns data is where things get uncomfortable for AI vendors and deploying organisations alike.
| Concern | Frequency | Business implication |
|---|---|---|
| Unreliability | 26.7% | Trust deficit slows adoption and increases oversight costs |
| Jobs and economy | 22.3% | Workforce anxiety undermines engagement with AI tools |
| Autonomy and agency | 21.9% | People fear losing control over their own decisions |
| Cognitive atrophy | 16.3% | Skills degradation risk from over-reliance |
| Governance gaps | 14.7% | Insufficient frameworks create compliance uncertainty |
| Misinformation | 13.6% | Deepfakes and truth erosion affect organisational trust |
| Surveillance and privacy | 13.1% | Data exploitation concerns limit willing participation |
Reality Check: Unreliability is the top concern at 26.7% — not job loss, not privacy, not governance. People’s primary frustration is that AI gets things wrong and they have to check its work. This is a solvable problem, and solving it delivers disproportionate trust returns.
The unreliability concern (26.7%) outranks job fears (22.3%). That is counterintuitive and important. People are not primarily afraid of being replaced — they are frustrated by having to verify everything AI produces. The verification burden is the adoption bottleneck, not existential anxiety.
For UK businesses subject to regulatory scrutiny, the governance gaps concern (14.7%) maps directly to the EU AI Act’s influence on UK regulatory direction and the ongoing development of the UK’s own AI governance framework. Employees already sense the gap. They want guardrails, not because they oppose AI, but because they want to use it confidently.
How to actually use this data
The strategic response splits into three maturity levels, each building on the last.
For organisations just starting AI adoption
Reframe the value proposition. Stop selling AI internally as “efficiency gains” and start positioning it around professional excellence — the top aspiration at 18.8%. Frame AI tools as instruments that handle routine cognitive work so people can focus on judgement, strategy, and relationship-building.
Implementation Note: Run a simple internal survey modelled on Anthropic’s open-ended approach: “What would you most want AI to do for you?” The answers will almost certainly diverge from what your technology team assumed. Use the gap as your adoption strategy foundation.
Address unreliability head-on. Build verification workflows into every AI deployment. Make “checking AI’s work” a legitimate, valued activity rather than an embarrassing admission. The 26.7% unreliability concern is your biggest barrier — treat it as such.
For organisations with active AI programmes
Measure capability building, not just efficiency. Track how AI tools develop skills (9.9% of users report this), enable new types of work (8.7%), and improve decision quality (17.2%). These metrics capture value that time-saved measurements miss entirely.
Acknowledge the tensions explicitly. Use Anthropic’s “light and shade” framework in your change management. Name the five tensions in workshops and team discussions. People who feel their ambivalence is understood — rather than dismissed — engage more productively with new tools.
SME Advantage: Smaller organisations can move faster here. A 50-person company can run an open conversation about AI tensions in a single all-hands meeting. A 5,000-person enterprise needs months of cascaded workshops. Speed of cultural adaptation is a genuine competitive edge.
For organisations leading AI integration
Design for the whole person. The data shows people want AI for personal transformation (13.7%), life management (13.5%), and time freedom (11.1%) — not just work output. Organisations that allow AI tools for personal development and life management (within appropriate boundaries) will see higher voluntary adoption and satisfaction.
Build governance before you are forced to. The 14.7% governance gap concern is a leading indicator. Establish clear AI use policies, data handling standards, and escalation paths now. When UK regulation catches up — and it will — you want to be ahead of it, not scrambling.
Four risks hiding in the data
1. The measurement trap
If you measure AI success purely through productivity metrics, you will optimise for the wrong outcome. The data shows 81.2% of aspirations sit outside professional excellence. Narrow measurement creates narrow adoption — and leaves most of the value on the table.
2. The regional assumption error
Lower and middle-income countries show 67%+ positive sentiment toward AI, viewing it as an opportunity ladder. Wealthier regions — including the UK — express more concern about governance, surveillance, and displacement. If your organisation operates across markets, a single global AI strategy will land differently in Lagos and London. The same tool positioned as “empowering” in one context reads as “threatening” in another.
Hidden Cost: East Asian respondents show distinctly high concern about cognitive atrophy (18%) and meaning loss (13%). For UK organisations with Asia-Pacific operations or customers, this cultural dimension affects adoption, not just sentiment surveys.
3. The dependency spiral
Cognitive atrophy (16.3%) is the sleeper risk in this dataset. As AI handles more cognitive work, the skills to do that work without AI may degrade. This is not hypothetical — it mirrors what happened with GPS navigation and spatial awareness, or spell-checkers and spelling ability. Organisations need active skill maintenance programmes alongside AI deployment.
4. The trust asymmetry
People trust AI for information processing (research, synthesis, analysis) but distrust it for judgement calls. The verification burden from unreliability (26.7%) compounds this: every hallucination reinforces the view that AI cannot be trusted for anything important. Trust builds slowly and breaks fast. A single high-profile AI error in your organisation can set adoption back months.
What this means for UK business leaders
Three facts from this study should shape your AI strategy for the rest of 2026:
People want professional excellence, not just productivity. The 18.8% who prioritise professional excellence want to do better work, not just more work. Frame AI around quality and capability, not volume and speed.
Unreliability is the primary barrier. At 26.7%, it outranks every other concern. Invest in verification workflows, human review processes, and honest communication about AI limitations. The organisations that solve the trust problem first will capture the most value.
Your workforce holds contradictory views — and that is normal. The “light and shade” pattern means most people simultaneously value and fear AI. Strategies built for a world of pure enthusiasts or pure sceptics will fail. Build for ambivalence.
Next steps checklist
- Run an internal aspirations survey using open-ended questions (not multiple choice)
- Audit current AI metrics — do they capture capability building or just efficiency?
- Review your AI governance framework against the top seven concerns
- Brief leadership on the “light and shade” tensions framework
- Assess whether your AI positioning emphasises professional excellence or just productivity
Take Action: Download or bookmark the full Anthropic study — it contains regional breakdowns and demographic detail worth reviewing with your leadership team. The raw data tells a more nuanced story than any summary can capture.
Source: What 81,000 People Want from AI, Anthropic, 18 March 2026. Analysis by Resultsense.
Resultsense provides independent analysis of AI developments for UK businesses. Read more of our strategic insights or explore our AI glossary to build your team’s AI literacy.