Acrisure Cuts 2,250 Jobs as Broker AI Automation Splits the Distribution Channel

Acrisure chief executive Greg Williams told staff on May 20, 2026 that the firm would eliminate approximately 2,250 positions, roughly 11% of its global workforce, in phases running into 2027. The letter, first reported by Insurance Journal, named "advances in technology, AI, and digital platforms" as the reason. It is the largest known AI-attributed headcount cut in insurance distribution, and two comparable brokers deployed AI on the same timeline without cutting anyone.

Key Takeaways

• 2,250 roles, phased into 2027, at a broker running roughly $5 billion of 2025 revenue and 19,000 to 20,000 staff, carrying a $32 billion valuation from Bain Capital, Blackstone and Abu Dhabi Investment Authority money.
• 85% productivity gains and 2.5 hours saved per employee per week is what HUB International disclosed from its Claude deployment to more than 20,000 employees, with no associated layoffs.
• 20% of its global workforce, 8,500 to 9,000 of about 43,000 roles, is Chubb's parallel plan, targeting automation of 85% of major underwriting and claims processes for 1.5 combined ratio points of expense.
• Insurance sales agents grow 4% through 2034 in BLS projections while underwriters fall 3% and claims adjusters 5%, which is where the cuts are landing.

What the Letter Actually Commits To

The phrasing carries the commitment. "We will be reducing our headcount by approximately 2,250 roles," Williams wrote. "This process will begin today and continue in phases into 2027." That is a sustained programme tied to capability arriving over time, not a restructuring charge taken once.

The stated target is manual work: leveraging "AI, data, and automation to reduce manual work and create faster, more consistent outcomes." North America Insurance will organise "more intentionally around our lines of business." No department was named, but manual work in a broker means back-office processing, data entry, submission preparation and routine administration rather than client-facing advisory.

Scale sets the context. Acrisure is the sixth-largest broker globally, at roughly $4.8 billion of revenue in 2024 and approaching $5 billion in 2025, built by acquisition, 155 firms in 2021 alone, on private equity capital including Bain Capital's $2.1 billion in May 2025 and Abu Dhabi Investment Authority's $725 million in June 2022. An 11% reduction moves revenue per employee immediately, which is the metric a $32 billion valuation is defended with.

The Same Technology, Two Opposite Bets

Metric	Acrisure	HUB International	Baldwin Group
Revenue (est. 2025)	~$5B	~$5B	~$1.6B
Total employees	~19,000	20,000+	~5,000
AI workforce strategy	Substitution (2,250 cut)	Augmentation (no cuts disclosed)	Augmentation (no cuts disclosed)
AI partner disclosed	Not named	Anthropic (Claude)	Anthropic (Claude)
Productivity metric	Not disclosed	85% gains; 2.5 hrs/week saved	Not quantified publicly
Ownership	Private (PE-backed)	Private (PE-backed)	Public (NASDAQ: BWIN)
Announcement date	May 20, 2026	February 25, 2026	May 4, 2026

The disclosure pattern is the informative part. HUB named its vendor, named its product, and put three numbers on the table: 85% productivity gains in targeted use cases, 2.5 hours saved per employee per week, and satisfaction above 90% in early implementations. Baldwin named the same vendor. Acrisure cited AI as a category and disclosed neither a platform nor a productivity metric.

Chubb sits with Acrisure on the substitution side and is more explicit about it, planning to cut 20% of about 43,000 employees over three to four years while automating 85% of major underwriting and claims processes for 1.5 combined ratio points. The difference is absorption: at 43,000 people over four years, natural attrition covers much of a 20% target. An 11% cut announced in one letter does not wait for turnover.

Where this reaches actuarial work is the submission. Broker back-office output is the upstream input to every carrier pricing decision, and eliminating 2,250 processing roles raises the question of what performs that work now. Human processors introduced errors and also caught anomalies and flagged unusual risk characteristics, which is contextual judgment an automated pipeline has to be built to reproduce rather than inheriting.

The timing is what makes it a reserving problem rather than an operational one. A reduction phased into 2027 means the workforce producing accident year 2027 submissions is structurally different from the one that produced the experience in current loss triangles, and the change arrives gradually across the diagonal rather than at a single point. For commercial lines placed through a mix of brokers on opposite strategies, a single submission quality assumption across the channel is doing more work than it can support.

The Rules Were Written for the Roles That Survive

Broker regulation assumes a licensed human in the recommendation. State licensing and suitability obligations attach to the individual who compares coverage, assesses risk for recommendation purposes, and communicates with the policyholder, on the premise that a person exercises that judgment.

AI governance rules, meanwhile, were built pointing at carriers. The NAIC Model Bulletin, adopted in late 2023 and implemented in nearly half the states, requires written governance programmes, documented decision-making and anti-discrimination compliance from insurers. Colorado's SB24-205 took effect on February 1, 2026 and reaches "consequential decisions," defined around a consumer's access to or eligibility for insurance and differentiated price or material terms, with SB26-189 revising provisions in May 2026.

Neither frame lands cleanly on a broker whose AI decides which carriers to approach, which coverage forms to put forward, and how a risk is presented to an underwriter. Those functions touch suitability even when a human signs the placement, and whether they are consequential decisions is untested.

The BLS projections show why that gap matters more each year. Insurance sales agents are projected to grow 4% through 2034 while underwriters decline 3% and claims adjusters, appraisers, examiners and investigators decline 5%. The roles being automated out are the processing roles, and the licensing regime that would ask questions about judgment sits on the advisory roles that are growing. The work moving to machines is the work no broker-side rule was written to supervise.