What New AI Labour Market Research Means for Compliance and Safety Teams

A significant new research paper published this month by Anthropic economists Maxim Massenkoff and Peter McCrory offers the most detailed picture yet of how AI is — and critically, how it is not — reshaping the labour market. For safety managers, operations directors, HR leaders, and CFOs working across Australia's compliance-heavy industries, the findings deserve careful attention.

This article unpacks what the research says, what it means for the industries ComplyFlow serves, and how platforms like ours are positioned to support teams navigating this transition confidently.

What the Research Actually Found

The Anthropic paper, "Labor Market Impacts of AI: A New Measure and Early Evidence" (Massenkoff & McCrory, 2026), introduces a new way of measuring AI's real-world impact on jobs — one that distinguishes between what AI could theoretically do and what it is actually being used to do in professional settings.

The researchers call this new metric Observed Exposure — a measure that looks at how much of a job's tasks AI is actually handling in professional settings today, not just what it could theoretically do. It factors in real usage data from Anthropic's platform, academic task-level assessments, and gives greater weight to fully automated AI use over situations where AI is simply assisting a human.

The key finding is nuanced but important: there is a very large gap between what AI could theoretically handle and what it is actually automating today. Current AI coverage remains well below its theoretical ceiling across nearly every occupational category.

Specifically regarding unemployment, the researchers found no statistically significant increase in unemployment rates among workers in the most AI-exposed occupations since ChatGPT's release in late 2022. There is, however, early suggestive evidence that new hiring of younger workers (aged 22–25) has slowed in the most exposed roles — a signal worth watching, though one the authors treat with appropriate caution.

Where Is AI Actually Changing Compliance and Safety Work Right Now?

The Anthropic research includes a radar chart (left/above) plotting theoretical AI capability against observed AI coverage across occupational categories. Two areas stand out:

The blue area represents theoretical AI capability — the share of job tasks within each occupational category that AI could theoretically perform or accelerate, based on academic task-level assessments.

The red area represents observed AI coverage — the share of tasks in that category that AI is actually being used for in real professional contexts today, as measured through genuine platform usage data.

The gap between blue and red is stark. Across almost every occupational category — management, business and finance, legal, office and administration, computer and mathematics — the observed (red) coverage is a fraction of the theoretical (blue) potential.

This is not a story about AI falling short. It is a story about deployment being early-stage. As the authors note, the red area will continue to grow toward the blue as capabilities improve, adoption spreads, and organisations build the workflows to operationalise AI effectively.

For safety managers, operations directors, HR managers, procurement leaders, and CFOs working across Australia's compliance-heavy industries, this chart is reassuring in the short term and clarifying for the medium term. The roles most central to compliance, safety, and workforce management — the operational judgement calls, the contractor risk assessments, the site-specific decisions — are not being displaced today. But the trajectory is clear, and organisations that build AI-ready processes now will hold a meaningful advantage.

What This Means for the Industries We Serve

Construction

Construction sits in a category where physical, on-site tasks remain firmly beyond AI's reach. Supervising site works, managing subcontractor relationships, and making real-time safety calls in dynamic environments are not tasks AI automates. However, the administrative and documentation layer of construction compliance — contractor prequalification, induction management, permit processing, and incident reporting — falls squarely within the categories showing growing AI coverage.

For construction businesses managing across multiple projects and sites, the implication is that AI tools will increasingly handle document review and verification tasks. Organisations that have already digitised these workflows will be far better positioned to layer AI capabilities on top of them.

Mining

Mining operations face some of the most demanding compliance requirements in Australia, spanning WHS obligations, environmental reporting, contractor onboarding, and audit readiness. The research confirms that document-intensive, data-processing, and reporting tasks — which represent a significant share of the administrative burden in mining compliance — are among the areas seeing the fastest growth in real-world AI coverage.

For safety and compliance managers in mining, this creates both an opportunity and a risk. Those who adopt AI-assisted contractor compliance and safety management tools early will see genuine reductions in administrative overhead. Those who delay risk falling behind as competitors streamline compliance processes and redeploy that capacity toward higher-value oversight work.

Energy and Utilities

Energy and utilities organisations operate within complex, multi-site regulatory environments. The Anthropic research identifies business and finance, legal, and management occupational categories as having high theoretical AI exposure — all of which are well represented in energy sector compliance teams. As AI increasingly handles regulatory documentation, reporting, and change management tasks, organisations in this sector that have centralised their compliance data will be able to extract far greater value from AI tooling.

Logistics, Warehousing, and Transport

The research specifically identifies customer service representatives and data entry roles as among the most observed-AI-exposed occupations today — and logistics and warehousing operations employ large numbers of workers in these categories. For HR managers and safety coordinators in this sector, the practical implication is that AI is already handling routine communication, data entry, and record-keeping tasks that have historically consumed compliance teams' time.

The strategic move for logistics organisations is to consolidate onto a unified workforce compliance and safety management platform now, so that AI-assisted processes can be introduced on a clean, structured data foundation rather than across fragmented spreadsheets and legacy systems.

Property, Facilities Management, and Aged Care

Facilities managers and HR leaders in property and aged care oversee large contractor ecosystems and internal workforces with complex induction, certification, and document requirements. These roles sit at the intersection of administrative AI exposure and physical operational responsibility — precisely the combination that the research suggests will see significant change over the coming years.

The administrative burden of managing contractor and employee compliance across multiple sites is already substantial. AI-powered document review and automated compliance workflows are not a future concept for this sector — they are available today.

The Broader Implication: The Gap Is an Opportunity

Perhaps the most important strategic takeaway from the Massenkoff and McCrory research is this: the gap between theoretical AI capability and observed usage is not primarily a technology problem — it is an adoption and workflow problem.

The researchers note that many theoretically feasible AI tasks remain unused in practice due to "legal constraints, specific software requirements, human verification steps, or other hurdles." In compliance-intensive industries, those hurdles are exactly what the right platform infrastructure is designed to remove.

Organisations that already have their contractor data centralised, their documents digitised, their induction workflows structured, and their audit trails clean will be the first to realise genuine productivity gains as AI capabilities are layered in. Those still operating on spreadsheets, email threads, and paper-based processes face a compounding disadvantage.

How ComplyFlow Supports AI Adoption in Compliance Teams

ComplyFlow has been helping organisations manage contractor compliance and workforce safety since 2009 — long before AI entered the picture. Over the years, we have evolved our platform to embrace intelligent automation across our product suite, precisely because we believe the gap shown in that Anthropic chart represents the next frontier of operational efficiency for compliance teams.

Here is how our platform supports the transition:

AI-Powered Document Review Our AI solutions include intelligent document review that can process contractor compliance documentation, flag expiry dates, identify gaps, and surface risk indicators — removing the manual scanning burden from compliance teams and allowing them to focus on decisions rather than data entry.

No-Code AI Agent Builder ComplyFlow's no-code AI agent builder allows safety and operations managers to configure automated compliance workflows without requiring technical expertise. This directly addresses the adoption barrier the Anthropic research identifies — making AI operationally accessible for teams whose primary skillset is safety, not software.

Structured Data Foundation Every AI capability is only as good as the data it operates on. ComplyFlow's contractor management and safety management modules ensure that contractor records, certifications, inductions, and incident data are held in a structured, centralised, and audit-ready format — the foundation that makes meaningful AI application possible.

Digital Inductions and Site Access Our site induction software and site access management tools replace paper-based and fragmented onboarding processes with digital, trackable workflows. As AI increasingly assists in tailoring induction content and verifying competency documentation, having these processes on a structured platform becomes a prerequisite for realising that value.

Real-Time Incident Reporting and Visibility Incident management that generates clean, structured data in real time is increasingly important as organisations look to use AI to identify patterns, predict risk, and improve safety outcomes. ComplyFlow's real-time incident reporting tools ensure that the data exists in a form that AI can act on.

A Note on What the Research Does Not Suggest

It is worth being clear about what the Massenkoff and McCrory findings do not support. The research does not suggest that safety professionals, compliance managers, facilities directors, or HR leaders face imminent displacement. The occupations most exposed to observed AI usage today are concentrated in programming, customer service, and data entry — roles characterised by highly repetitive, well-defined tasks that can be fully automated.

The judgement-intensive, relationship-dependent, site-specific, and legally accountable work that defines compliance and safety leadership remains firmly in the blue-not-yet-red territory of the Anthropic chart. What AI is changing is the administrative scaffolding around that work — reducing the time spent on document processing, data entry, and routine reporting so that professionals can do more of the high-value work that genuinely requires human expertise.

Conclusion

The Anthropic labour market research provides the clearest evidence yet that AI's impact on professional roles is real but early-stage, measurable but not yet dramatic. For organisations in construction, mining, energy, logistics, property, and aged care, the strategic window to build AI-ready compliance infrastructure is open now — before the gap between theoretical capability and observed deployment narrows further.

ComplyFlow exists to help organisations make that transition confidently: with the right data structure, the right workflows, and the right AI tools to support safety and compliance teams as their operating environment evolves.

Request a Demo | Start Free | Explore AI Solutions

Source: Massenkoff, M. & McCrory, P. (2026). "Labor Market Impacts of AI: A New Measure and Early Evidence." Anthropic. Published 5 March 2026. Available at: anthropic.com/research/labor-market-impacts