WTW launched its AI Workforce Transformation solution on June 2, 2026: two diagnostics that score automation potential at the level of individual jobs rather than industries, built on a benchmark database of more than 30 million data points.

Its analysis of 900 occupations puts professional and judgment-intensive roles at 20% to 35% task automation. That band is where actuarial work sits, and what falls inside it is more specific than the number suggests.

Key Takeaways

  • 20% to 35% of tasks in professional and judgment-intensive roles are automatable on WTW's analysis, against 60% to 70% for operations and clerical work and up to 75% for frontline repeatable tasks.
  • 30 million benchmark data points across 900 occupations back WorkVue Agent's role-level scores, with ChangeVue identifying which organizational areas are ready rather than which are ambitious.
  • Underwriting active handling time falls from 45 minutes to 15 per submission on BCG's numbers, a 30% to 40% reduction, with quote turnaround down as much as 60%.
  • 54% of advanced AI adopters plan headcount cuts, five times the 11% rate among less mature peers.
  • 400,000 insurance professionals are projected to retire between 2021 and 2026, against only 214,000 employees aged 20 to 24.

What WTW Launched

Both tools sit on WTW's Reinventing Jobs methodology, which deconstructs a job into discrete tasks, evaluates each for optimization across technology, employees and non-traditional talent, then reassembles the remainder into a redesigned role.

WorkVue Agent produces automation potential for every job in an organization, drawn from the benchmark database rather than from industry averages. ChangeVue answers a different question: where implementation should start, based on readiness. A claims unit with clean data pipelines and documented processes may be a better first target than an underwriting desk with higher theoretical automation potential sitting on fragmented legacy workflows.

Role Category Task Automation Potential Typical Insurance Functions
Operations, administrative, clerical 60-70% of tasks Policy admin, data entry, document processing
Industrial and frontline Up to 75% of repeatable tasks Claims triage, FNOL intake, billing workflows
Professional and judgment-intensive 20-35% of tasks Actuarial analysis, underwriting, compliance

Shai Ganu, who leads WTW's executive compensation and board advisory practice, framed the sale to boards rather than to operations: "Boards don't need more theory on AI. They need precision. As their mandates expand to cover human capital governance, fiduciary duty now means knowing exactly where AI creates value and how work must be redesigned to capture it."

What Actually Falls Inside the 20 to 35% Band

The band is not a headcount reduction estimate. It is a statement about task composition, and the SOA's April 2026 analysis names the tasks precisely enough to check.

Agents preparing data for valuation runs and flagging inconsistencies. Digital assistants executing model test suites overnight and drafting summary reports. Monitoring agents identifying mismatched policy counts and automatically rerunning valuation cases. Every one is data manipulation or pattern detection, which is what puts them inside the band. The remaining 65% to 80% is judgment, communication and stakeholder management, and none of those tasks touch it.

Underwriting shows the same shape with harder numbers. BCG puts active handling time at 45 minutes falling to 15 per submission under full AI redesign, a 30% to 40% reduction, with quote turnaround compressing up to 60%. Mike McGavick, speaking at the CAS Seminar on Reinsurance, sized the wider version at $32 billion of annual sector inefficiency, or 12 to 14 cents per premium dollar, with underwriters spending roughly 40% of their time on non-core administrative work. The underwriter's judgment role does not shrink with the minutes; it concentrates into them. What survives is risk selection, portfolio-level assessment and the exception cases falling outside the model's confidence interval.

For an actuarial function the consequence is compositional rather than numerical. Assumption setting, model validation and exception review grow as a share of the work; data preparation, routine calculation and standard reporting shrink. The department does not contract the way operations or administration does, but the skill mix it hires for changes, which is what the SOA and CAS competence ladder is built around.

The expense modelling is where this stops being an HR question. A carrier restructuring roles on these scores books severance and retraining before the efficiency arrives, so a rate filing that projects expense ratio on a static staffing basis will misplace both the level and the timing. The J-curve is the shape, and its depth depends on how many roles move at once.

One split complicates the retraining assumption. The SOA's Summer 2025 survey found roughly 60% of late-career actuaries using AI for learning and idea generation against about 50% of early-career actuaries, while code generation ran the other way at roughly 48% early-career against 29% late-career. The adoption pattern is not uniform even inside one function.

The Demand Signal Is Running Ahead of the Readiness

The product launched the same week the demand evidence did, and the two do not fit together neatly.

Covenir's 2026 Insurance Operations Leaders Trends Report, from 152 U.S. operations decision-makers, found 54% of advanced AI adopters planning headcount cuts against 11% among less mature peers. The same survey found 70% with AI live in operations, up from 58%, while 20% cut training budgets and 7% protected them, ninety-one percent of executives reported maximum team strain, and forty-two percent named First Notice of Loss as where brand promise breaks down most often.

The demographics reframe those cuts. Bureau of Labor Statistics data projects 400,000 insurance professionals retiring between 2021 and 2026, with 1.37 million workers aged 55 or older, nearly one in four, against only 214,000 aged 20 to 24, a six-to-one ratio. Seventy-nine percent of Gen Z say they have never considered working in insurance. For a meaningful share of carriers, automation is not a cost play at all; it is a response to roles they cannot fill, and the disclosed workforce actions read differently under that reading.

What the diagnostic cannot supply is the capacity to act on it. Grant Thornton's 2026 survey of 950 executives found 44% saying governance and compliance challenges contributed to AI projects failing or underperforming, and only 24% very confident their organization could pass an independent AI governance review within 90 days.

That is the same imbalance every framework in this market keeps measuring. BCG's own 10-20-70 split assigns 10% of the scaling challenge to algorithms, 20% to technology and data and 70% to people and process, while Capgemini finds P&C insurers spending 72% of AI investment on technology against 28% on change management. A role-level automation score measures the gap with more precision than anything previously available. It does not close it, and the carriers whose training budgets are being cut are the ones buying the measurement.

Further Reading on actuary.info