Don’t Build for What VCs Fund: Stephen Todres on the Great Agency Reset, Human Agents and Concentrated Venture Capital

Stephen Todres is on the investment team at American Family Ventures (AFV), the venture arm of American Family Insurance. He invests at the early stage across InsurTech, with a particular focus on proptech, homeowners and other personal lines. AFV got started over a decade ago, Stephen says, and is launching its fifth fund now.

Most of our conversation was about agencies. AFV’s thesis, which it calls the Great Agency Reset, starts from a conviction the firm has held for years: human agents are not going anywhere. What changes, Stephen says, is the software underneath them and how fast agencies are willing to buy it.

In Episode 175 of InsurTechTalk, Stephen and I covered the Great Agency Reset, why AI vendors that “sell the work” are finally getting agencies to adopt technology, what concentrated venture capital means for insurance startups, and the advice he gives founders who ask whether to build a tech company, an agency or an MGA.

About Stephen Todres

Stephen started his career at a startup focused on performance marketing. In his words it was “essentially a tech-enabled agency,” working with large gaming and entertainment clients. (I came from the mobile gaming world, so we had that in common.) He then went to business school and got his MBA at Columbia Business School.

During the MBA he interned at a couple of venture firms. The most relevant, he says, was MetaProp, one of the early proptech VC firms, where he quickly found a lot of overlap between proptech and InsurTech. That was his segue into American Family Ventures, which he joined coming out of the MBA program.

At AFV he focuses on early-stage investing across InsurTech, with proptech and personal lines as his particular areas. The firm is flexible in how it operates, he says, and his own work is a combination of diligence, sourcing and supporting existing portfolio companies.

What American Family Ventures Invests In

The firm goes by American Family Ventures or AFV, Stephen told me. It is backed by American Family Insurance, and its focus, in his words, is “anything shaping the future of the insurance industry.” That includes technology sold to carriers and brokerages, and broader fintech or proptech use cases that still apply to insurance.

The roots are still early stage: seed through Series A. With each new fund, Stephen says, AFV refreshes its thesis areas, and the mandate for the new fund is similar to the earlier ones.

Horizontal Platforms Count Too

Stephen added that in the current market, many of the most interesting solutions for carriers are horizontal platforms that aren’t insurance-specific. Think of how much carriers spend on Salesforce. A big part of AFV’s job, he says, is knowing that wider landscape well.

I told him that mapping it out is not easy. When a partner asks me for a quick hour on what’s going on in insurance, my first question is: an hour on which topic? Distribution alone, which in my opinion is one of the critical issues, could fill it. I’ve been in the industry for about ten years and I still learn something new almost every day.

Agencies: Three Convictions

When AFV builds a thesis, Stephen says, one framework is to ask what will stay the same in the industry for the next five to ten years and what will fundamentally change. Applied to agencies and distribution, the firm reached three main conclusions.

1. Human Agents Aren’t Going Anywhere

“The first is we don’t think human agents are going anywhere,” Stephen says. AFV held that conviction even during the direct-to-consumer craze about ten years ago. Human agents remain a key part of the process, especially for more complex coverage decisions, and he doesn’t expect AI to replace them in that time frame. So AFV stays bullish on solutions that help existing agents rather than trying to displace them.

2. AI Could Make Agency Management Systems Switchable

The second conclusion is where AI has the larger impact. Stephen points to agency management systems, where incumbents such as Vertafore and Applied have extremely sticky products with “greater than 90% retention.” He sees an opportunity for AI to make it easier to switch off legacy systems into tools that are more bespoke and user-friendly, for new agencies and for existing brokerages. It’s an area AFV is spending quite a bit of time on, he says.

3. Selling the Work, Not a Workflow Tool

The third is speed of adoption. Agencies have historically been “a very difficult market to crack for InsurTech startups,” Stephen says. “And I think because of the promise that a lot of these new AI vendors have in terms of selling the work as opposed to just selling a workflow tool, we’re seeing much faster adoption and scaling than we’ve seen in the past years.”

He named two AFV portfolio companies that have capitalized on that trend, Pathwork and Qumis, and called agency technology a space the firm is “super bullish on” heading into the new fund.

The Great Agency Reset

Before we got there, I admitted I was tempted to talk about AI and the frontier labs instead. Stephen offered to cover anything, as long as the episode aired “before the AI kills that stuff.” I said I’d rather talk about the Great Agency Reset than whether we’re in the Matrix.

Stephen describes the reset as a few factors coming together.

Consolidation

The first is growing consolidation. Without listing every transaction of the past year or two, Stephen says there is a lot of interest from private equity and other stakeholders in acquiring, rolling up and consolidating brokerages. That brings technology challenges in integration, and opportunities to make things more streamlined. He expects it to continue for the next few years.

AI-Native Agencies With a Different Margin Profile

The second is a growing number of startups that take legacy business models and use AI to create a completely different margin profile. Stephen’s example is the AI-native startup agency that needs relatively few human agents and can operate much more efficiently. He has also seen roll-ups in the space, and named Gyde on the Medicare brokerage side. The margin expansion from AI is still to be determined, he says, but there’s a more plausible story to tell than in the past, and that has made AFV more interested than in previous years.

What I See in the Deal Flow

I asked whether AFV had invested in COVU, Ali Safavi’s company, because its model is a good match for the trend: software for managing the agency, and the consolidation around it. Stephen knew the name, but it is not in AFV’s portfolio.

I told him what writing the weekly report has been like lately. There are so many deals, most of them small, that the report got very long, and I had to decide which deals are interesting enough to cover. We see the consolidation, with Trucordia and the other big brokers acquiring agencies year after year. There’s still more to be done there, and if you connect it to AI, the opportunity is mainly in operations and efficiency. But the agents and brokers are not going away. In commercial especially, for anything complicated, you need a person: it’s a hard risk to underwrite, or even to shop for.

Why Human Agents Are Here to Stay

Stephen sees the same thing in proptech. There has been years of talk about real estate agents being displaced by AI or by online platforms, and he says it simply hasn’t happened. When people make large financial decisions, they still want people to help them. That doesn’t rule out innovation or some compression in fees. “But to think that someone’s just going to use Claude to buy a house, or for a leader at a company to rely on that for buying a large commercial insurance policy, it seems far-fetched, at least in the near to medium term.”

Who Is Looking Out for the Policyholder

I asked why so many regulated intermediaries exist in the first place, and whether there’s more room for tech-enabled agencies. Stephen took it to the policyholder: who is looking out for the consumer? A very small percentage of agents are bad actors, he says, and he sees it in health care, in both ACA and Medicare. It’s important to have rules and the ability to crack down on them without handicapping the 99 percent of agents who do their jobs the right way. He calls it a difficult, iterative problem, and says he’s always interested in speaking to people using technology or AI to help solve it.

When Things Go Wrong, You Want a Person

My own view is simple: when you’re in a bad situation, you want to talk to a person, not a robot or someone reading a script. I told Stephen about an Airbnb I booked for my mother. The property manager didn’t show up. I booked her a hotel quickly, which wasn’t the problem. The problem was that every attempt to reach the platform led to a different machine, and a person would only call back 24 hours later, when what I needed was a bed for my mother within a few hours. It’s the same if your house burns down in the Palisades or anywhere else in California.

Stephen told a story from an AFV team offsite in Madison, Wisconsin, where American Family is headquartered. They heard from someone whose area had been hit by a recent storm or tornado, and about what human agents do in that moment: the empathy, the hand-holding, being there when people need them. A machine is not going to replicate that, he says. “I think that’s a big part of our thesis, is finding technology that can empower individuals to better provide that kind of service.”

I added my caveat: for now. Stephen’s reply: “I think we’re going to need a different level of technology to service our Jewish mothers.”

VC Trends: More Money, Fewer Firms

After a detour into babka and knishes on the Lower East Side (Stephen’s tip: Yonah Schimmel is near Katz’s Deli, so you can double up), I asked how he sees venture capital in insurance, ten years into InsurTech.

Stephen calls it a unique time. “There’s more money than ever in venture capital,” he says, which he didn’t expect to say after the COVID era. What makes it unique is concentration: capital flowing to a smaller number of VC firms and a smaller number of startups. Deal count over the past year is nowhere near the 2021 peak, he says, but the amount of capital is. He cited two thirds of recent capital going into Anthropic, OpenAI and xAI. InsurTech sees a similar dynamic, though not to that extreme.

The Upside, and the Risk

For a Series A or Series B startup, the growth markets have opened up, Stephen says, making growth equity rounds much smoother than in the past couple of years. Strong companies with high growth and good metrics will benefit.

The risk is investors who don’t know insurance. Generalist funds often have unrealistic expectations around growth and unit economics, he says, that don’t map well to the SaaS or AI segments they’re used to. He is still more bullish on generalists being in the space: they bring best practices for product development and hiring, and skills that translate from other industries. “But I start to get worried when I see complex insurance companies growing without true insurance expertise on the cap table or at the founder level.” His expectation: the market will be bifurcated into winners and losers, and it will be increasingly hard to tell which is which.

My Rule for Generalists

I agree that the insurance industry needs generalist investors, because it needs the capital. But in my opinion they always need to co-invest with an insurance specialist. Otherwise they lose. We saw it in what Stephen called a bubble: a lot of money went in, created hype and valuations, and didn’t hold, because that money didn’t respect the insurance KPIs. It applied valuations from SaaS and other industries that weren’t right for insurance.

Tech Company, Agency or MGA?

Early-stage founders ask me this a lot: should we be a tech company, use our technology to start an agency, or become an MGA? In my opinion, the MGA is on the rise again. Each model has its own catch-22.

“I think in some ways it’s the wrong question, not to call you out on your own podcast,” Stephen said. It’s totally founder and team dependent. His advice: build the business you want to build and are best suited to build. “Don’t try to build for what you think VCs are going to want to fund, because then you’re just the blind following the blind.”

Have core convictions, he says: your North Star, your mission, what you’re trying to solve. Then iterate and adapt the model on top. Pivots and business model changes are fine. Chasing VC money to tell a story is not.

Are Agencies Venture-Backable?

To be determined, Stephen says, but the short answer is yes. Agencies, not all of them VC-backed, have grown, reached the public markets and done well there; he gave Goosehead as an example. The burden of proof is high, though. A founder has to show that the agency can grow with enough capital efficiency to avoid “diluting the hell out of your early stage investors,” and that it can bring together the right mix of human expertise and technology, which he says is really hard. He sees it as a higher-risk, higher-reward story than more traditional SaaS or AI businesses.

I said that’s the question I keep hearing. Anyone can take the Kaplan course, get licensed and start an agency, which is why the industry has such a long-tail problem. Raising money for one means fitting a thesis and showing returns in a time frame that works for the investors.

Stephen sees MGAs the same way. An MGA adds the difficulty of securing capacity and proving it knows how to underwrite, but he puts about 80 percent of the challenges in common with traditional brokerages.

Reinsurance Capital and ILS

I told Stephen that Rick Zullo of Equal Ventures had sent me down a research rabbit hole on taking some of the burden off reinsurers. Stephen said AFV had talked about this with American Family’s reinsurance team. You can see reinsurance as very large, in the tens of billions of dollars, or as tiny next to other asset classes. There’s a lot of capital out there that could provide a very similar function, and bridging that gap, through ILS or other products, is an interesting area to explore. “I don’t often agree with Rick, but I’ll throw him a bone on this one.”

I’ll give Rick and Jonathan Crystal credit too: there are many investors who truly understand the industry.

What Insurance Should Talk About More: People

Stephen’s one-word answer was people. AFV puts a lot of emphasis on a startup’s impact on people, he says, and both VCs and founders tend to overlook two people challenges in insurance.

The first is change management. The industry needs AI implementation and process improvement, and that doesn’t happen without buy-in at every level, from top executives to the most junior employees. Where AFV has seen carriers succeed with AI, he says, they built processes that got input and buy-in from every level. He doesn’t see a straightforward technological fix: much of it is change management and closer to consulting.

The second is an aging workforce. Many agents are retiring and agencies need new talent, which Stephen sees as a potential tailwind for technology adoption. The same applies to claims adjusters. His example of a model that works is Goosehead again, which recruits undergraduates on campus and shows them a lucrative career in the agency world. He’d like to see versions of that across other segments of the industry.

How to Pitch AFV

Founders can find Stephen on LinkedIn or by email (his first name at amfamventures.com). Two things to know before pitching:

• Don’t wait until you’re fundraising. The biggest misconception, he says, is that founders should only talk to VCs when they’re actively raising. AFV likes building relationships at the earliest stages, even before the company exists.
• Make specific requests. Stephen says the firm prides itself on being among the best-connected VC firms in insurance, so founders should ask for specific introductions, such as two reinsurers or two brokerages they want to meet. “The more specific, the better.”

My take

AFV’s view that human agents aren’t going anywhere matches what I said on the recording. In commercial especially, for anything complicated, you need a person, because it’s a hard risk to underwrite and even to shop for. And when something goes wrong, as it did with my mother’s Airbnb, nobody wants a machine or a script. For now, technology can’t provide that.

Stephen’s point about concentration fits what I tracked in my Q4 2025 M&A report: 76 insurtech deals in Q3 2025, the lowest count since 2016, and only four investors made two or more insurtech investments that quarter. American Family Ventures was one of them.

On generalists, Stephen and I land close together. The industry needs their capital, but in my opinion they should always co-invest with an insurance specialist. The last cycle showed why: the money applied SaaS valuations and didn’t respect insurance KPIs, and the valuations didn’t hold.

Founders keep asking me whether to be a tech company, an agency or an MGA. My view on the recording was that the MGA is on the rise again, and that each model has its own catch-22. Stephen called it the wrong question: for him, the team decides the model, not the other way around.

Key Takeaways

• AFV’s agency thesis starts from what won’t change in five to ten years: human agents, especially for complex coverage decisions
• Stephen sees AI’s larger impact on agency management systems, where sticky incumbents could become switchable
• Agencies are adopting faster because AI vendors sell the work, not a workflow tool
• The Great Agency Reset combines brokerage consolidation with AI-native agencies built on a different margin profile
• There’s more venture money than ever, Stephen says, concentrated in fewer firms and fewer startups
• Generalist investors bring useful practices, but complex insurance companies need insurance expertise on the cap table or among the founders
• Build the business your team is best suited to build, not the one you think VCs want to fund
• Agencies are backable, Stephen says, with a high burden of proof on capital efficiency and on combining people with technology
• The industry’s under-discussed issue is people: AI change management and an aging agent and adjuster workforce

Stephen’s figures and market statements, including greater than 90 percent retention for agency management incumbents, two thirds of recent venture capital going to Anthropic, OpenAI and xAI, and 80 percent of challenges shared by MGAs and brokerages, are as he stated them and were not independently verified.

Related on InsurTech.me

• Only four investors made 2+ insurtech investments in Q3 2025, American Family Ventures among them (M&A quarterly report, Q4 2025)
• Carrier-backed venture arms remain active in deals (weekly report, March 8, 2026)
• Kate Johnson on American Family Ventures’ fourth fund (Episode 106, 2023)
• Ali Safavi, COVU: AI-native services instead of selling tools (Episode 135)
• Will Johnson, Gyde (Episode 153)
• Rick Zullo, Equal Ventures: investable underwriting and access to capacity (Episode 170)
• Jonathan Crystal, Crystal Venture Partners (Episode 163)

Connect

• Stephen Todres on LinkedIn
• American Family Ventures: amfamventures.com
• AFV’s article: The Great Agency Reset

Every insurance deal, decoded weekly: the Insurance & InsurTech Investment Intelligence Report.