Why One API Wasn’t Enough: Duncan Crystal on Strategic Placement at Scale

Duncan Crystal recorded this episode from a hotel in Chicago, where he was meeting current and future partners at the Broker Tech Connect event. It is a fitting setting. His company, Herald, exists to solve one of the least glamorous and most expensive problems in commercial insurance: getting a client’s risk to the right market, in the right format, without a broker losing a day to it.

Herald started in 2021 with a clean idea: a unified API for commercial insurance, so that nobody has to visit every carrier portal. Many carriers and brokers use it today. But Duncan is candid that the idea hit a ceiling, and that AI is what broke through it.

In Episode 173 of InsurTechTalk, Duncan and I covered why placement is still so manual, why the “we will replace the broker” era of insurtech was a mistake, and why he believes judgment about what to build is becoming the real bottleneck in a vibe-coding world.

About Duncan Crystal

Duncan Crystal is the Co-Founder and Chief Revenue Officer of Herald, a digital infrastructure company for commercial insurance. He started in insurance at At-Bay, the cyber insurance MGU, where he joined before the company sold its first policy and focused on distribution: getting brokers to use the portal, getting them to send submissions by email, and later plugging At-Bay’s API into broker technology initiatives at large partners. He co-founded Herald in early 2021 with his childhood friend Matt, now Herald’s CEO and previously At-Bay’s first product manager. Before insurance, Duncan studied at Johns Hopkins, where he captained the wrestling team, and then played fullback in rugby in Washington, DC.

What Herald Sells: Strategic Placement at Scale

Herald’s core system is called the intelligent placement system, and the vision behind it is what Duncan calls strategic placement at scale.

Placement Strategy vs. Placement Reality

Every brokerage has a placement strategy: how it wants its producers to connect clients to the markets. Then there is the placement reality. When every producer runs their own process, there is a gap between where the brokerage would like business placed and where it actually ends up. There are plenty of good reasons for that gap, and plenty of friction in the industry behind them. Herald uses technology to close it.

Why Commercial Placement Is Still a Gordian Knot

Duncan’s shorthand for explaining insurance to people outside it: it is one of the industries least touched by the internet, where faxes are still flying around. By his count, there are roughly 35,000 to 40,000 agencies in the US and about 2,000 P&C carriers, each with more than one product. Moving data from a client to those markets and back, in every format imaginable, still relies heavily on manual labor: spreadsheets, scratch pads next to the desk, a managed inbox. He puts the scale at more than a trillion dollars of insurance transactions a year in the US.

The Five Steps of Placement

Herald breaks placement into categories, each with its own set of niche problems that add up to a very big deal:

• Submission intake
• Choosing a marketing strategy, meaning which carriers to approach, which is where the appetite challenge fits
• The marketing itself: sending risk information to markets, collecting quotes, and going back and forth until the terms are acceptable
• Organizing the market results and presenting them to the client, which is part risk advisory and part sales
• Binding and issuing the policy

Who “The Market” Actually Is

When a broker “sends it to the market,” that market is MGAs, wholesalers, carriers directly, and even reinsurers on the reinsurance side. Each has its own channel. Some insist on email. Some run portals where brokers log in and get quotes back instantly. Some expose their underwriting capability through an API so developers can build on top of it. The broker has to figure out who is likely to take the risk and how that market wants to be approached, then keep leaving the client context to do it. Duncan calls this the swivel chair effect, and Herald tries to organize all of it in one place.

The Old Insurtech Mistake: “We Will Replace the Broker”

Duncan pushed back on the idea that the goal is to reduce the number of portals brokers use. He calls that an old way of thinking, and he sees it everywhere in the industry: the overzealous approach of “I am going to remove this entirely, you will use my solution because I know best, instead of everything you have done for the past 20 or 30 years.”

Companies like Next and Lemonade started with that premise: the broker or the agent would become unnecessary. Then came the pivot to selling through agents, and eventually the press release saying they had always loved agents. Everybody, as we joked, loves a partner. Duncan admits he has fallen into some version of this thinking himself, and says it took reflection to get out of it. It humbles you.

Herald’s Ceiling and “Radical Acceptance”

Herald began as a unified API for commercial insurance: the technological equivalent of not needing to visit all the portals. It has been genuinely useful, and many partners on both the carrier and broker side use it.

The Ceiling

Some risks, submissions or accounts cannot be handled through the APIs carriers offer. The broker has to add human touch: go to a portal, email an underwriter, or engage the market in some special way. Forcing all of insurance into a single API, Duncan says, was bursting at the seams.

What AI Changed

AI let Herald break through that ceiling. The company expanded its aperture to what Duncan calls radical acceptance: helping brokers work wherever they work, rather than forcing or implicitly nudging them toward only using Herald’s unified API.

From the Wrestling Mat to At-Bay

Duncan describes picking up rugby after college as a socially acceptable outlet for the violent energy he built up as a wrestling captain, and says many of his closest friends today came from that rugby team.

His route into insurance ran through his childhood friend Matt, whom he met at summer camp. Matt joined At-Bay first as its first product manager, built the policy administration system, and pulled Duncan in. Both were there before the first policy was sold. What opened Duncan’s mind to starting a company, with someone he trusts, was seeing how much industry-specific friction sits between the people who represent clients and the supply side of the market: the emails, handwritten notes, faxes, phone calls and spreadsheets it takes to place risk.

The Messy Middle Between the CRM and the AMS

Duncan describes a brokerage’s technology in three layers. A CRM such as Salesforce or HubSpot helps acquire customers and manage the sales process. An agency management system tracks the policies already placed and handles billing and accounting. In between, the broker and their team are left with an inbox and a series of point solutions. That is what Herald calls the messy middle of placement.

Where Brokerages Can Grow

When brokerages look for growth, Duncan sees them circling the same three levers inside that messy middle:

• Placing clients with the right carrier
• Building strategic carrier relationships, which often means better coverage and better commission
• Removing friction so throughput is faster

His illustration: if an employee can handle five accounts a day and, a year later, eight, that is a 60 percent throughput increase. Many attempts at that kind of gain have been frustrated.

How Brokerages Solve It Today, and What AI Changes

Duncan focuses on the largest brokerages. By his count, the top 50 control about 56 percent of commercial insurance premium, set against roughly 35,000 agencies overall. The techniques he sees large brokers use today:

• Task specialization, with teams that focus only on the placement or marketing step
• Carrier panels, meaning special relationships where more commission is promised in return for routing business
• Rainmaking producers who are encouraged to sell as much as possible

The problem, he says, is that in a technology-free world producers take the path of least resistance. They do the easy thing, not always the strategic thing. His view of AI’s role is a colleague that can look across all of a brokerage’s markets, understand who is likely to provide better coverage, better commission and better claims handling, and route the placement accordingly. That should be better for the brokerage, the carrier and the client.

Commission Isn’t the Whole Story

We talked about the incentive structures: the carrier incentive books, the president’s clubs, and the perks for keeping a clean book. Duncan agrees that every decision in insurance is shaped by incentives in some way. What surprised him when he entered the industry is how many decisions were not driven by commission at all.

Producers, he explains, are underwater from a work perspective. A producer might be able to work hard to earn three extra points on a placement, but has no time for it. The real incentive is getting the thing off their desk so they can finish the day or focus on higher-reward work. That is where technology comes in: making the strategically right thing the easiest thing to do mechanically. For a firm, three points would be a very big deal.

What Insurance Should Talk About More: Technology Spend

Asked the closing question, Duncan said the industry does not talk enough about technology, and specifically how much money it takes to do it well. He finds himself drawing on other industries’ benchmarks to explain how hard it is to build, and more importantly maintain, technology at scale that handles the complexity and use cases across insurance. The brokerage side has been technology-lean compared with other industries. He expects insurance broking to eventually look like investment banking and other financial services in how much it spends on technology, and to have to reorganize around technology being part of every step.

Vibe Coding Is a Double-Edged Sword

Duncan closed on a point that applies well beyond brokers. It is easier than ever for people to put ideas out there or to use Claude to manifest technology. But that also means the judgment about what to build, when, which problem you are solving, and how to do it sustainably over time becomes a bigger deal, because all the ideas have entered the room. Every broker and every employee can manifest their own vision. Having the discretion to choose is the new challenge.

Key Takeaways

• Every brokerage has a placement strategy and a placement reality, and the gap between them is where technology can create value
• Placement is a five-step process, and each step carries its own niche problems that compound into a very large inefficiency
• Forcing all of commercial insurance into one API hit a ceiling; accepting that brokers work in many places, and using AI to meet them there, is what Herald calls radical acceptance
• The “we will replace the broker” premise of early insurtech was a mistake, and even experienced founders can fall back into it
• Producers are often driven less by commission than by clearing their desk, so the strategic choice has to be the easy choice
• Insurance brokerage will likely need to invest in technology the way other financial services firms do
• In a vibe-coding world, judgment about what to build is the new bottleneck