Karsten Wenzlaff, Advisor
August 26th, 2025
Karsten Wenzlaff is a Crowdfunding and Fintech Expert from Germany. He is Secretary-General of Digit[...]
AI Usage Data Shows Early Labour Market Strain
Mar 5, 2026 | NCFA Market Insight | AI And The Future Of Work
Image: Labour market impacts of AI: New measure, early evidence (Anthropic Mar 2026)
Early Signs Of AI Changing Knowledge Work Hiring
On March 5 2026, new Anthropic artificial intelligence labour market impact research | download 17 page PDF offers a more grounded way to think about AI and jobs. Most AI labour studies ask what large language models could do in theory. This one asks what workers actually do with AI now. That sounds like a small change, but it matters. Financial institutions, fintechs, and policymakers need to know where real workflow change starts, not just where capability headlines point.
The paper introduces a new measure called observed exposure. It combines occupational task data, Claude usage data, and earlier research on which tasks language models can theoretically handle. The approach tries to separate technical possibility from real use. That makes this study potentially more useful than other broad forecasts about AI replacing jobs. To be clear, at this point the research doesn't claim mass displacement has arrived. It looks instead for early pressure points that may show up before unemployment spikes.
Why This Research Matters
This report matters because it focuses on real work patterns. The authors use the US O*NET task database, Anthropic Economic Index usage data, and earlier task level exposure estimates to build their measure. They also give more weight to automated and work related uses than to lighter forms of assistance. That helps the study track where AI starts to move from drafting help into actual production work.
See: Why AI Investment Is Missing What Workers Actually Want
That said, the research has real limits, and those limits matter. The usage data comes from Claude, not from the full AI economy. A bank using Microsoft, Google, OpenAI, or internal tools may show a different pattern. The measure also tracks task exposure, not direct job loss. And the authors make clear that adoption still lags capability because of legal controls, software requirements, human review, and workflow friction. That caution strengthens the paper rather than weakening it. It keeps the analysis honest.
AI Can Do More Than Workers Use Today
The most striking findings is the size of the gap between capability and real usage. In STEM (ie. computer and math) occupations, earlier work suggests language models could theoretically touch 94% of tasks. Anthropic’s observed measure puts current Claude coverage in that category at just 33%. In other words, the technology frontier sits much further ahead than the adoption frontier.
That gap matters for fintech and finance because it points to where the real bottleneck sits. The next productivity jump might not come from a bigger, better, faster or smarter model alone. It might come from firms that redesign workflows, tighten controls, connect tools to internal systems, and build trust in output quality. Which operators can close the adoption gap faster than peers.
Programming At The Front Of The Curve
The report finds that computer programmers rank as the most exposed occupation, with 74.5% coverage. Customer service representatives follow at 70.1%, and data entry keyers come next at 67.1%. Digital, repeatable, text heavy work moves first.
See: Fintech Fridays EP55: Global Hiring Trends: How Gen Z Talent Thrives
What stands out for NCFA readers is that this isn't just software engineers. Financial analysts are also among the most exposed occupations in the paper’s wider discussion. That should get the attention of banks, insurers, wealth platforms, lenders, and compliance teams. A lot of value in finance comes from summarizing information, checking documents, explaining options, processing forms, and moving data from one system to another. Those are exactly the kinds of tasks that AI enters early.
Higher Paid Knowledge Work Feels The Pressure First
Workers in the most exposed occupations earn 47% more on average than workers in the unexposed group. People with graduate degrees make up 17.4% of the most exposed group, compared with 4.5% of the unexposed group.
Early pressure is showing up in better paid, more educated, office based roles. In practice, that means AI may alter the work of analysts, developers, researchers, service teams, and operations professionals before it touches many frontline physical jobs. For financial institutions, this points to a management challenge as much as a technology one. Firms need to rethink role design, supervision, training, and quality control inside knowledge work functions.
Younger Workers Show The Earliest Cracks
The research does not find a growing rise in unemployment for workers in the most exposed occupations since late 2022. That is an important result because it pushes back on louder claims that AI already causes wide labour market damage.
But the report does find an early warning sign. For workers aged 22 to 25, job finding rates into highly exposed occupations fall by about 14% compared with 2022 levels. The authors stress that this result is only barely statistically significant and open to other interpretations. So something to watch and track.
This is how labour market change often starts. Not with mass layoffs. Not with an obvious cliff. It starts when firms stop hiring as many junior people because AI now handles part of the routine work those hires once learned on. That's important for banking, fintech, consulting, and other white collar sectors that depend on apprenticeship style career ladders. If entry points narrow, firms may save money in the short run while weakening their future talent pipeline.
What Fintech And Financial Institutions Should Watch
Firms should watch three things closely to better anticipate where the real gap may open between leaders and laggards:
- Where routine document and analysis work starts to compress
- Whether junior hiring weakens in functions that already rely on digital knowledge work
- Which institutions move beyond pilots and actually rebuild operating processes around AI
See: Why Fintechs Are Rebuilding How They Measure Work
For fintech builders, the market needs tools that help institutions safely plug AI into live work. Audit trails, review controls, workflow routing, structured data extraction, policy checking, and role based permissions. Companies that solve those problems help close the distance between AI capability and actual business use and adoption.
In Conclusion
Anthropic’s research is quite useful because it shows where real adoption starts, where the biggest gaps remain, and where the cracks may start to appear and proliferate.
The National Crowdfunding & Fintech Association (NCFA Canada) is a financial innovation ecosystem that provides education, market intelligence, industry stewardship, networking and funding opportunities and services to thousands of community members and works closely with industry, government, partners and affiliates to create a vibrant and innovative fintech and funding industry in Canada. Decentralized and distributed, NCFA is engaged with global stakeholders and helps incubate projects and investment in fintech, alternative finance, crowdfunding, peer-to-peer finance, payments, digital assets and tokens, artificial intelligence, blockchain, cryptocurrency, regtech, and insurtech sectors. Join Canada's Fintech & Funding Community today FREE! Or become a contributing member and get perks. For more information, please visit: www.ncfacanada.org
 |  |  |
|---|---|---|
 |  |  |
Leave a Reply