Dearborn Labs Introduces AI Chiefs of Staff for Executives

No memos. No mandates. No code. Five AI agents running in our leadership layer — built entirely by people who'd never coded before. Here's how it happened. #insurance #AI #insurtech #DearbornLabs

Last year when I left Applied, I kept coming back to one thought: AI isn’t just another feature for insurance — it’s going to change how the whole thing works. It felt like one of those moments you don’t watch from the sidelines. Since then, we’ve spent months talking with brokerages, carriers, and MGAs — from small shops to Top 100s, department leaders, and vendor partners. Everyone says some version of the same three things: ▎ “It’s a lot of noise.” ▎ “We know we have to do something.” ▎ “The pieces don’t connect.” Out-of-the-box doesn’t work here. You can’t serve a brokerage without tying into the AMS, carriers, communication tools, SOPs — and modeling data the way it actually flows. The hard part is that every insurance business is different. That’s also where the value lives. That’s the bet behind Crux. Today, we’re officially launching Crux Studio — a platform for building branded AI agents for insurance. Agents that are always on. That sound like your brokerage. That handle real work — quotes, COIs, payments, servicing, and everything in between — across the channels your customers already use. All while connecting your systems behind the scenes. The future is clear: every business front will have an AI agent. We’re building the platform that powers it for insurance. Take a look at the video overview below ⬇️ — more at trycrux.com Also — picking the song for this took way longer than it should have. "Minimal Abstract Is." Corporate lo-fi banger. You're welcome. Excited to keep building — with Bo Brown, Michael Onyon, the rest of the Crux crew, and the customers who've shaped this with us.

A senior producer at an insurance agency built a near-perfect Standard Operating Procedure on her first AI-assisted attempt. She had barely used a large language model before that day. I want to name what actually made it work, because it was not the AI. She had been doing the quoting workflow for years. She knew the carrier eligibility rules cold. She knew which carrier to go to when Farmers declined, which to go to for high fire risk, which to go to when there had been prior losses, and when the California Fair Plan plus a companion policy was the last resort. She had never written any of it down. But she had clarity. We turned on voice dictation. She talked through the process in chunks. The AI produced a structured, complete SOP on the first pass. One correction needed. Here is the counterfactual most people skip past. If she had been fuzzy on the process, the AI would have produced a fuzzy SOP. Confident-sounding, structured, and wrong. And nobody would have caught it until it hit production. → The AI did not figure out the business → The AI amplified what was already clear in her head → Clarity in, clarity out. Garbage in, garbage out faster. This is the part of AI adoption that gets lied about constantly. Vendors tell you the tool will "learn your business" or "understand your processes." It will not. It will reflect what you give it, sharpened and faster. The companies that are winning with AI right now are not the ones who bought better tools. They are the ones whose people already knew their work well enough that AI could amplify it. If your team cannot describe a workflow out loud, AI cannot document it for you. It will just produce something that looks like documentation. That gap will cost you more than the subscription ever did. Before you roll out more tools, ask who on your team could voice-dictate their highest-leverage workflow this afternoon. That is the person you start with.

If you’re a small or mid-sized business owner paying under $10K yearly in premium. I can help you have a better conversation with your broker in under 2 minutes. For free. Here’s what I did I have been experimenting with this on different AI tools Copilot, Gemini, ChatGPT, and Claude. I uploaded my own commercial cleaning policy docs directly into all of them , my personal favorite is Claude. I carefully went through their answers and compared them to the original info on the docs. By the way, Hallucinations on all of them were very minor. I’m not saying you should depend blindly on this neither am I saying brokers should put clients sensitive or personal info on public AI spaces. To reiterate, this has been exclusively used on my own business docs. if you’re a business owner who wants to upload your own documents into Claude, ChatGPT, Gemini or Copilot to better understand your coverage before a broker conversation, that is entirely your call to make. Next time you get your renewal docs from your commercial broker. Copy & paste this prompt I created ⬇️ If you’re too lazy to write you can can attach the docs, and voice note your operations etc. “Act as an expert Risk Management Consultant specializing in commercial insurance. I will describe my business operations below and share my insurance policy documents. Please: 1. Analyze my policy documents thoroughly, including all endorsements, exclusions, and riders. 2. Summarize my key coverages in plain, simple language 3. Clearly outline what is included and what is excluded and flag where exclusions may interact or overlap in ways that could affect my coverage 4. Identify any gaps or exposures based on my specific business operations 5. Flag each risk as High, Medium, or Low priority 6. Flag any area where you are uncertain or where I should verify directly with my broker or the original policy document 7. Provide a list of questions I should bring to my broker Here is a description of my business: [Describe what your business does, number of employees, revenue range, and any specific operations, equipment, or services worth noting] Here are my policy documents: [Upload your full policy docs including any endorsements or renewal summaries] Please present your findings in a clear, structured report format that a business owner with no insurance background can easily understand. Where something is complex or ambiguous, say so plainly rather than presenting it with false confidence.” You welcome, SZ

Something shifted in insurance AI this week. Not the hype. The architecture. For 3 years, AI in insurance meant tools bolted on top of workflows. Copilots. Assistants. Chatbots that helped underwriters draft emails faster. Useful. But peripheral. Then Guidewire dropped ProNavigator. And something about it felt different. Not because of the feature list — but because of what it assumes about how AI should sit inside an insurance operation. It knows your internal processes. It understands role-specific context — UW, Claims, Service. It tracks decisions with audit trails baked in. It enforces governance by design — not as an afterthought. This isn't AI helping work happen. This is AI becoming part of how work happens. And the operational numbers from firms moving in this direction are hard to ignore: → Claims turnaround: 14 days → 2 days → Cost per case: down 3x → Underwriting productivity: up 50% → GWP per underwriter: up 30% → 90% of manual work automated in some deployments → 95% fewer policy review errors This is no longer incremental efficiency. This is operating model transformation. These aren't pilot results. These are production numbers from firms that crossed the line from experimenting to embedding. But here’s the part most teams are underestimating: 78% of insurance firms believe they couldn't pass an AI audit today. Only 22% have a fully developed AI strategy. We're deploying AI faster than we're governing it. And that gap — between adoption speed and governance maturity — is where the next wave of competitive advantage will be won or lost. What makes ProNavigator interesting isn't the speed it unlocks. It's that it's solving for trust, consistency, and compliance at scale — not just throughput. That's a different design philosophy. And it matters. The firms that win won't be the ones who used AI the most. They'll be the ones who governed it the best. Because once AI becomes infrastructure — not a feature, not a function, but actual operating infrastructure — it stops being optional to get this right. A misconfigured underwriting model doesn't just create a bad recommendation. It creates a regulatory event. The question to every insurance leader right now: Are you still adopting AI… or are you redesigning your operating model around it? Those are not the same journey. And the gap between them is widening every quarter.

Most insurance companies I talk to want to start with an "AI strategy." They want a roadmap. A committee. A 6-month assessment phase. Meanwhile, their underwriters are still copy-pasting vehicle data from Excel files into a quoting tool. Manually. Row by row. Here's what I've learned in the field: the companies that actually move forward with AI don't start with a strategy. They start with a problem. One process. One bottleneck. One workflow that everyone in the team knows is broken but nobody has time to fix. For most commercial insurers, that's the quoting process. It's manual, repetitive, error-prone, and it's the one workflow where speed directly impacts revenue ! So instead of spending 6 months mapping your "AI maturity," try this: Pick the workflow your team complains about the most. Automate it. Measure the before and after. Then decide what comes next. That's not a shortcut : that's how every successful AI deployment I've seen actually started. One workflow. Real results. Then expansion. The roadmap builds itself when the first win is undeniable. I'm Tanguy, co-founder of MincaAI. We build AI-powered automation for insurance operations: underwriting, issuance, endorsements, claims. Want to see what AI can do for your workflows? Book a discovery call (link in the first comment.)

18 months ago I would have told an insurance broker that AI is a technology worth exploring. Today, that advice would get you left behind. Kader Sakkaria, Global Head of Data Technology at Gallagher, said it better than I could: "The next three years will determine which firms shape the future of insurance and which are left behind." Cannot agree more. What the industry called modernization is now survival. A year and a half ago, AI tools were hard to set up and basically inaccessible to a traditional insurance brokerage. Today, non-technical people across any industry are building workflows quickly and easily. The tools caught up. The question now is when the brokers will. So where do you start? Start with what takes up the most of your time and isn't business critical. One workflow at a time. We're happy to help if you need support getting started. Link to the article below. https://lnkd.in/egM-CPVm

Most AI conversations in insurance are still stuck at proof of concept. BrokerTech Ventures just published a recap of a panel I moderated at BrokerTech Connect: Chicago this past summer with Garrett Droege, former SVP and Digital Practice Leader at IMA Financial Group. He's been inside one of the most complex brokerage operations in the country trying to actually build what everyone talks about. One line he said stuck with me. "We have a lot of documents that tell our people how to do what they do. We built an LLM so they could just chat with all those policies and procedures. And it worked incredibly well." No vendor pitch. No grand vision. Just a real problem, solved. What I loved about the whole conversation was how honest Garrett was about the limits. AI isn't ready for end to end workflows yet. It still needs a skilled human driving it. Hallucinations are real. Expert review isn't optional. But the firms leaning in aren't waiting for perfect. They're being intentional about what to pick off first. The other thing I kept coming back to: the adoption challenge isn't really a technology challenge. The people closest to the work know exactly where the friction is. They're the ones who should be driving this, not a top down mandate from the C-suite. Garrett put it better than I could. "It's not a CEO message saying we're going to use AI. It's seeing the person next to you take 30 percent of their day back." Thanks to BrokerTech Ventures for putting this together. Link to the full piece: https://lnkd.in/dEfp4bXs

Today's my first day at Fulcrum, and I want to tell you why I'm here. Insurance brokerages handle millions of policies every year. When a policy needs to be reviewed. Checking that the coverage, limits, endorsements, and details are all correct — it traditionally takes a human roughly a full day. Sometimes longer. Fulcrum does it in 5–10 minutes. But here's the part that really hit me: Because software scales infinitely, Fulcrum can run 100,000 policy reviews at the same time. All of them done in under 10 minutes. Simultaneously. Think about what that means. A task that used to require thousands of hours of manual work — done in the time it takes to make coffee. That's not a marginal improvement. That's a fundamentally different way to operate. Fulcrum builds AI agents that handle the operational work insurance brokerages have traditionally outsourced — policy checking, proposal generation, certificate issuance, coverage analysis. It plugs directly into the tools teams already use, so there's nothing new to learn. Nearly half of the top 50 U.S. insurance brokerages already use Fulcrum. The platform executes over 2,500 hours of servicing work every single day. I joined because this is one of those rare cases where the technology doesn't just make things faster — it changes what's possible. Excited to be here. More to come. 🦾

GCs may be relying on a risk allocation that no longer matches the insurance behind it. Most companies now use AI in some form. Whether their insurance still responds if those tools cause a problem is a question many may not have asked. In the US, the insurance market has started to move. From January 2026, Verisk’s ISO introduced optional endorsements allowing carriers to exclude certain generative AI exposures from standard commercial general liability policies. These are the kinds of policy changes that can slip in through a normal renewal cycle and materially change the position. The issue is not theoretical. As AI moves into live workflows, insurers are starting to narrow cover, add endorsements or push AI risk into separate products. The issue often gets missed because it sits across functions. AI deployment often sits with technology. Insurance sits with finance or risk. Legal may only see the issue once there is a claim, a dispute or a contract negotiation that assumes cover is there. For a GC, this is not just an insurance point. If cover no longer responds to an AI-related loss, that gap flows straight into your contracts, indemnities and liability positions. Two simple questions are worth asking: - Has the last renewal introduced any AI exclusions or sub-limits? - Do your key supplier and customer contracts still make sense if it has? If nobody knows, that already feels like useful information.