The Big Firms Have Chosen Their Model. Both of Them Chose Claude.
In the span of three weeks, two of the world’s largest professional services firms announced global alliances with Anthropic — and the scale of both commitments is worth reading carefully.
PwC is deploying Claude across three areas of its business: agentic software delivery for major enterprise clients, AI-native deal execution including M&A and due diligence, and the redesign of core enterprise functions such as finance, supply chain, and HR. The firm will train and certify 30,000 professionals on Claude, with a joint Center of Excellence to drive deployment across client engagements. PwC’s announcement included one number that belongs in every insurance leader’s peripheral vision: underwriting cycles compressed from ten weeks to ten days. That is not a pilot. That is a production result being used to sell AI transformation services to insurance carriers.
KPMG followed with its own announcement: a global alliance embedding Claude into KPMG Digital Gateway, its client delivery platform, with access extended to all 276,000 KPMG professionals worldwide. The initial focus is tax and private equity, but the alliance also covers cybersecurity — KPMG teams are using Claude to find and fix vulnerabilities in client systems. Anthropic named KPMG as a preferred consultant for private equity clients, which adds a distribution channel that will touch portfolio companies across financial services.
Taken together, these are not technology vendor partnerships. They are commitments by two of the most credentialed professional services organizations in the world to deliver AI through a specific model. When KPMG or PwC arrives at a carrier’s offices, they are increasingly bringing Claude with them.
Why This Matters for Insurance:
The professional services channel is how AI gets deployed at carriers, not through direct vendor relationships. When PwC’s transformation teams quote a ten-week-to-ten-day underwriting cycle improvement as a client result, that becomes the expectation against which carrier CIOs will be measured. And when KPMG’s tax and audit teams rebuild their workflows on agentic AI, the compliance and audit functions at the carriers they serve will face the same pressure to keep pace. The choice of AI model is no longer just a technology decision — it is increasingly embedded in which advisors you work with and what those advisors bring into the room.
Sources:
- PwC and Anthropic Expand Alliance — PwC trains 30,000 professionals on Claude
- KPMG and Anthropic Sign Global Alliance and Launch Digital Gateway Powered by Claude
Anthropic Files for an IPO. What That Means for the Industry That Relies on It.
On June 1, Anthropic filed confidentially for an initial public offering. The filing followed a $65 billion Series H round that pushed the company’s valuation to just under $1 trillion. No pricing or share count has been set; the actual offering will depend on market conditions.
The timing lands in a genuinely unusual IPO environment. SpaceX is pursuing a listing that targets a $2 trillion valuation. OpenAI, which raised $122 billion in March at an $852 billion valuation, is also expected to file. The result is an AI-sector public market debut season unlike anything in recent memory — one that will test whether institutional investors assign durable enterprise value to frontier AI companies or treat them as the latest version of the dot-com moment.
For insurance leaders, the IPO matters for reasons that go beyond financial markets. Anthropic’s governance structure has emphasized safety and responsible deployment in ways that have shaped both its model behavior and its enterprise relationships. A public company faces different pressures: quarterly earnings, shareholder expectations about growth, and the constant tension between revenue velocity and the kind of careful model development that Anthropic has staked its brand on. How that balance shifts — or doesn’t — will affect every organization using Claude in production.
Why This Matters for Insurance:
Carriers that have built workflows around Claude, or are evaluating it, are now choosing a vendor in the process of going public. That changes the strategic calculus. A public Anthropic is subject to earnings pressure that a private one was not. The S-1, when it becomes public, will contain detailed information about revenue concentration, enterprise contract structures, and the cost economics of frontier model development. For procurement and technology leadership at carriers, it is worth reading closely. The same applies to OpenAI when that filing arrives.
Source:
Bain’s Survey of 951 Companies Just Described the AI Problem Your Board Is About to Ask You About.
Bain & Company’s Automation and AI Pathfinder Survey, published this week, is the kind of document that travels well in board packages. Its central finding: AI programs are consistently underdelivering on the savings they promised, budgets are increasing anyway, and the gap between the two is widening in ways that most leadership teams have not yet acknowledged.
The specific numbers are sobering. While 37% of companies targeted cost reductions of 11% to 20%, nearly 40% of those that measured outcomes landed in the 0% to 10% range instead. The technology worked. The savings didn’t arrive. And 90% of those same companies are now increasing their budgets for the next wave — AI agents — funded in large part by the savings from the prior wave that fell short of projections.
Bain identifies three structural causes. First, only 7% of companies are running fully autonomous agents in production; the dominant operating model requires human approval for most agent actions, which changes the economics of any business case built on full automation. Second, data access and integration remains the number one barrier to AI progress, cited by 41% of respondents — above compliance concerns, budget, skills, and executive buy-in. Third, companies are deploying AI into broken processes rather than redesigning the processes first, which locks in workflow debt rather than eliminating it.
The companies that are actually delivering on their AI targets, Bain finds, made a small number of specific organizational decisions: they treated data access as a CEO-level problem rather than an IT problem, they validated investment cases against actual prior returns before approving new spending, and they designated explicit ownership of AI agent governance before deployment rather than improvising accountability when something went wrong.
Why This Matters for Insurance:
Insurance organizations have all three of the structural problems Bain identifies, and probably in more acute form than most industries. Claims, underwriting, and policy administration data lives in siloed systems that predate modern data architecture. Core workflows have accumulated decades of regulatory compliance layers that make process redesign politically difficult. And the business cases for AI programs in insurance have frequently been built on full-automation economics applied to environments where human-in-the-loop requirements — some regulatory, some prudential — are not optional. The Bain framework is a useful diagnostic tool for any carrier trying to understand why prior AI investments underdelivered and what to change before approving the next round.
Source:
73% of Insurance CEOs Have Named AI Their Top Investment Priority. Only 32% Have Seen Meaningful Returns.
New KPMG data published this week puts a precise number on the gap between insurance leadership’s AI ambitions and actual results: three-quarters of insurance CEOs now name AI as their top investment priority, but only about one in three has generated returns of any meaningful measure.
The survey data, drawn from KPMG’s annual insurance CEO study, shows that the industry has moved decisively past the question of whether to invest in AI. KPMG describes boards and C-suite leadership as having made “significant progress in understanding AI” — now treating it as a strategic business priority rather than an IT project. Two-thirds of respondents expect AI-derived returns within one to three years, up from 21% when the same question was asked two years ago. The same proportion plan to allocate 10% to 20% of their budgets toward AI. The separate finding that 92% of financial services companies broadly have generated some profit from AI use helps explain the optimism, even as specific insurance results remain uneven.
The concerns holding leaders back are worth noting. Compliance and security risks are the most commonly cited obstacles. A growing number of executives are worried about dependency on a small number of large technology providers — a legitimate strategic concern as carriers build core workflows on platforms controlled by vendors whose priorities may not align with the insurance market’s regulatory constraints or long-term stability requirements.
KPMG’s Riccardo Altenburg, Tech, Data and AI Lead for Insurance, put the path forward plainly: organizations that build a scalable AI foundation with quality data are best positioned to realize sustained value. That is consistent with what Bain found in its broader automation survey — data infrastructure is not a precondition you wait to meet, but a problem you solve as you deploy.
Why This Matters for Insurance:
The gap between CEO-level AI commitment and measurable returns is not a sign that insurance is behind — it is a sign that the industry is in the same place that every other sector finds itself roughly 18 to 24 months after declaring AI a strategic priority. The companies that closed this gap in other industries did so not by finding better technology but by solving the organizational and data problems that technology alone cannot fix. For insurance carriers, the practical implication is that AI governance, data integration, and workflow redesign are now C-suite priorities, not IT department projects.
Source:
Honeycomb Raises $40 Million on a Simple Bet: AI Can Underwrite Apartments Without Looking at Them.
Honeycomb Insurance, an AI-native specialty insurer focused on apartment buildings and condo associations, raised a $40 million Series B this week led by Zeev Ventures, bringing total funding to $95 million. The company ended 2025 with $275 million in gross written premium, covers more than $100 billion in total insured value across 22 states, and is already profitable.
The underlying model is worth understanding in some detail. Honeycomb’s automated underwriting engine ingests hundreds of data points per property — geospatial datasets, aerial imagery, building history — and prices each risk individually without a physical inspection. For well-maintained buildings that traditional carriers either overcharge or decline entirely, that can produce coverage priced up to 40% cheaper. The addressable market is not small: the U.S. multifamily insurance segment alone generates more than $34 billion in annual premium, and about 30% of Americans live in apartment buildings or condos.
CEO Itai Ben-Zaken, a Wharton MBA and Israeli military intelligence veteran who co-founded the company in 2019, is deliberate about keeping the raise modest. His benchmark is Neptune, the flood insurer that went public last year at roughly a $2.8 billion valuation on approximately $400 million in gross written premiums. At its current trajectory — Ben-Zaken projects crossing $500 million in GWP within a few years — Honeycomb is building toward a similar exit profile while remaining cash flow positive.
The competitive headwind is real. A quieter 2025 hurricane season and new capital flowing into reinsurance markets have pushed commercial property rates down 5% to 15%. Honeycomb’s response is operational: the company argues that an agent can sell five of its policies in the same time it takes to sell one elsewhere, which maintains its value proposition even as the pricing environment softens.
Why This Matters for Insurance:
Honeycomb represents a specific challenge to incumbent property underwriters: a company that has made the inspection-free underwriting model work at scale and achieved profitability doing it. The technology it is using — satellite imagery, geospatial data, and machine learning applied to individual property risk — is not new. What Honeycomb has done is prove the economics at meaningful premium volume. For large carriers still running physical inspection workflows on commercial property, the question is whether that model survives a decade of direct competition from companies that never needed the inspection in the first place.
Source:
The White House Signed an AI Executive Order This Week. Here Is What It Actually Says.
President Trump signed an executive order on June 2 titled “Promoting Advanced Artificial Intelligence Innovation and Security.” The order’s central focus is cybersecurity, not regulation — and that distinction matters.
The order does not create new AI regulation, and it explicitly prohibits construing any provision as authorization for “a mandatory governmental licensing, preclearance, or permitting requirement for the development, publication, release, or distribution of new AI models.” That language is deliberate. It signals that the administration’s posture toward frontier AI development remains permissive.
What the order does create is a set of cybersecurity-focused mandates. Within 30 days, CISA must issue guidance to prioritize the cyber defense of civilian federal systems and expand AI-enabled defensive tools. The Treasury Department must establish a voluntary AI cybersecurity clearinghouse to coordinate vulnerability scanning and patch distribution, with participation from the private sector and critical infrastructure operators — including rural hospitals, community banks, and local utilities. Within 60 days, a classified benchmarking process for frontier AI models must be developed, along with a voluntary framework for AI developers to provide pre-release government access to high-capability models.
The order also directs the Attorney General to prioritize enforcement of existing computer fraud and wire fraud statutes against anyone using AI to conduct unauthorized system access or facilitate other crimes.
Why This Matters for Insurance:
Insurance is listed as critical infrastructure in the broader federal framework, and the voluntary AI cybersecurity clearinghouse created by this order has direct relevance for carriers running AI systems in production. The explicit prohibition on mandatory AI licensing is also significant for insurtech companies and carriers building on frontier models — the regulatory risk of a federal pre-approval requirement for new AI deployments has been reduced, at least for the duration of this administration. State-level AI regulation remains a separate and more complicated picture, as the variation in health insurance AI statutes across Nebraska, Colorado, and Georgia demonstrated last week.
Sources:
From the AI World: Gemini 3.5 Arrives, and Agent Speed Just Changed.
Google released Gemini 3.5 Flash at Google I/O on May 19, positioning it as the company’s strongest agentic and coding model to date. The headline performance claim: output speed four times faster than other frontier models at comparable intelligence levels, with benchmark results that Google says surpass its own prior Gemini 3.1 Pro on complex coding and long-horizon agentic tasks.
The practical significance for enterprise AI users is the combination of speed and cost. Google states that 3.5 Flash completes complex agentic tasks at less than half the cost of other frontier models, and it is already deployed in the Gemini app, AI Mode in Google Search, and enterprise platforms. Macquarie Bank is piloting the model for customer onboarding, using it to reason over complex 100-plus page documents and produce recommendations with low latency. Salesforce is integrating it into Agentforce. Xero is using it to manage multi-week autonomous workflows like 1099 tax form preparation.
The broader competitive context: this release, combined with the Anthropic IPO filing and the PwC and KPMG alliance announcements, marks a week in which the enterprise AI market shifted noticeably. Model capability is advancing fast enough that the constraint on AI deployment is no longer primarily the model — it is the organizational readiness to use it, which is exactly what Bain’s survey found.
Why This Matters for Insurance:
Speed and cost economics matter for insurance AI applications in ways they do not for general productivity tools. Underwriting workflows that require rapid reasoning over large documents — policy applications, loss runs, submission packages — benefit directly from lower latency at lower token cost. The 3.5 Flash release raises the practical floor of what an agentic insurance AI deployment can deliver. It also intensifies the competitive pressure on carriers that are still in the planning and evaluation phase: the capability gap between what early movers have in production and what evaluators are considering is widening every quarter.
Source:
By James W. Moore
If You Found This Useful
This newsletter exists because insurance executives need trusted, sourced analysis on AI — not vendor marketing dressed up as news. If a colleague would benefit from receiving it, forward this issue or share the link.
And if you haven’t downloaded The Independent Insurance Agent’s AI Playbook yet, it is available free at insuranceindustry.ai/ai-playbook. More than 33,000 words of practical guidance, no cost, no catch.
Sources
- PwC and Anthropic Expanded Partnership — Anthropic
- KPMG and Anthropic Sign Global Alliance and Launch Digital Gateway Powered by Claude — KPMG
- Anthropic Files to Go Public — TechCrunch
- Your AI Budget Is Growing. Your Returns Aren’t. Here’s Why. — Bain & Company
- AI Is the Top Investment Area for 3 Out of 4 Insurance CEOs — TechRadar
- Exclusive: Investors Bet $40 Million on Honeycomb’s No-Inspection Home Insurance AI — Fortune
- Promoting Advanced Artificial Intelligence Innovation and Security — The White House
- Gemini 3.5: Frontier Intelligence with Action — Google
AI Disclaimer: This content was created with assistance from artificial intelligence technology. While content is based on factual information from the source material, readers should verify all details directly with the respective sources before making business decisions.
