Employers Holdings (EIG) AI: 94% Staff Adoption, AI-Built Product

Employers Holdings (EIG) reported a 94% AI staff adoption rate on its 2026-07-30 earnings call, the first time it has put that figure on the record, and the CEO confirmed that the company's new excess workers' compensation product was built with AI [1]. On the 2025-10-31 call, management had said only that it was using agentic AI to build the underwriting platform for that new product line [4]. Between those two calls, the line moved from a plan to written business.

EIG is a small insurance company that writes only workers' compensation, earning premium by underwriting employers' workers' compensation risk. It is running two AI applications at once. The first treats AI as a build method: the company's own build teams used generative and agentic AI to create the underwriting platform, the customer relationship management platform and the product tool itself for a new excess workers' compensation product. Excess workers' compensation is a policy that covers the portion of workers' compensation losses above an employer's own retention. This work serves internal product construction, which makes it a support function. The second application hands a general-purpose AI assistant to every employee for daily work. The first disclosure named Anthropic's Claude; the most recent one refers only to "our AI tool." Its scope is the whole company.

How the two AI applications evolved

The starting point was 2025-10-31. Management said it was spinning the excess workers' compensation line up in a very efficient way, hiring a team of underwriters while using agentic AI to build out the underwriting platform and the CRM platform [4]. The product did not exist yet, and the word "efficient" tied hiring and AI together in the same sentence, so no time saving can be assigned to AI on its own.

The 2026-02-20 call first explained the build method. Voice transcription was ingested by Claude to build the tool, it iterated daily for about four weeks, and the company said it was ready to launch months earlier than initially expected [3]. That is the only realized benefit management explicitly attributed to AI anywhere in this history, but it is not quantified, and the internal schedule it is measured against was never disclosed, so an outside reader cannot reconstruct it.

The same call opened the second line. The company said it was rolling Anthropic's Claude out across the organization, phrased as something in progress, with no seat count, no timetable and no cost. By 2026-04-30 the rollout had become organized: the company presented a company-wide AI implementation strategy and, one month before that call, brought approximately 400 employees from across the country together at an event to introduce it [2]. The only result available at that point was internal enthusiasm.

Where each application stands now

By 2026-07-30 both lines had reached a position that can be observed but has not entered the financial statements.

On the company-wide tool side, EIG reported a 94% AI staff adoption rate for the quarter and said it had implemented several AI-assisted use cases with meaningful, tangible ROIs [1]. That figure measures how many employees use the tool. It says nothing about the handling time, throughput or headcount of any specific process, and the company gave no employee total, no definition of what counts as "use" and no prior-period value.

On the build-method side, the CEO said he fully expects to build the large deductible product in exactly the same way the excess workers' compensation product was built [1]. The finished product wrote its first policy in June; the large deductible product is still only expected.

The financial metrics these two lines should eventually show up in are the underwriting expense ratio (underwriting expense divided by earned premium) and, for the new product line, gross written premium. The first measures how much expense each dollar of premium carries; the second measures how fast the new line takes in business. But the company itself has booked no money to AI. The CFO attributed this quarter's decline in underwriting expense item by item to expense management, lower personnel costs, policyholder dividends and bad debt expense. The link between AI and the financials is not yet proven.

What to watch next

What this history supports today is that AI at EIG is a lever on capacity and speed to market. It lowered the barrier for a small single-line insurer to stand up a new product line, and management has made it the established build method for the next product.

It is not a proven cost lever. Underwriting expense fell 8% year over year in the most recent quarter, but the company itself assigned that decline item by item to expense management, lower personnel costs, policyholder dividends and bad debt expense. Over the same period earned premium fell 12% and gross written premium fell 20%, so the expense carried by each dollar of premium actually got heavier. The purchase cost of AI itself is moving from license-based to usage-based pricing, and the amount has not been disclosed.

The single piece of evidence most likely to change that reading is the company breaking out the ROI of one AI-assisted use case: which role, which process, how much handling time or headcount was saved, and by what method it was calculated.

Application assessment

• Generative/agentic AI used by Employers' own teams as the build method for new workers' compensation product infrastructure - the excess workers' compensation underwriting platform, its CRM platform and the product tool itself - with the same method committed to for the planned large deductible product. | Business position: support function | Adoption stage: limited production | Scope: single business unit | Value type: not yet clear
• Company-wide rollout of a general-purpose AI assistant to every employee - named as Anthropic's Claude when the rollout was first reported and later referred to only as 'our AI tool' - carried by a formal company-wide AI implementation strategy and measured by a staff adoption rate. | Business position: shared services | Adoption stage: limited production | Scope: company-wide | Value type: cost reduction

Sources

[1] Drillr - Employers Holdings, Inc. (EIG) - 2026-07-30 - earnings call

"During the quarter, we achieved a 94% AI staff adoption rate and implemented several AI-assisted use cases with meaningful, tangible ROIs."

[2] Drillr - Employers Holdings, Inc. (EIG) - 2026-04-30 - earnings call

"Last month, we brought together approximately 400 employees from across the country to introduce our strategy for implementing AI throughout the organization."

[3] Drillr - Employers Holdings, Inc. (EIG) - 2026-02-20 - earnings call

"We used voice transcription that was ingested by Claude to build the tool, and it iterated daily for about four weeks, and we were ready to launch months earlier than we initially expected."

[4] Drillr - Employers Holdings, Inc. (EIG) - 2025-10-31 - earnings call

"we're spinning this up in a very efficient way by hiring a team of underwriters and then we're utilizing Agentic AI to build out the underwriting platform and the CRM platform"