Zywave's study this week, The Broker of the Future, prices something carriers have felt for two years: AI is set to unlock more than $100 billion in annual value for carriers and brokers within five to seven years, not by replacing the broker, but by rewriting the economics of how risk gets priced, serviced, selected, and covered. Its sharpest figure is the gap already separating winners from laggards: a traditional agency runs on roughly $175,000 of revenue per employee, while an AI-native broker is proving capable of running at closer to $1,000,000 — nearly six times as much. It lands against a sobering workforce fact: roughly 400,000 insurance professionals will have retired since 2021 by the end of this year, with not enough new talent behind them.
Vendor caveat attached: Zywave sells distribution technology, and the methodology behind the revenue-per-employee gap was not disclosed. But the argument survives it — it does not depend on the exact multiple, but on a distinction carriers keep getting wrong.
What it means for carriers
First, the $100 billion is a rebuild number, not a layoff number. Zywave's breakdown: $30–40 billion a year from software absorbing the back office, $10–20 billion from carriers pricing on live data instead of a once-a-year snapshot, $20–30 billion from better risk selection and fraud detection, and $15–20 billion from new AI risks becoming insurable coverage. The AI-native firms are not winning by cutting staff; they are growing business faster than headcount. This is Nº 009 in distribution clothes: delete steps and rebuild the journey, not a cheaper pair of hands for the same process.
Second, the talent gap changes what the broker is for. More than half the workforce still sits in service and back-office work — exactly the work AI absorbs best. The 400,000 retirements are an experience problem, not just a hiring problem: the judgment that leaves with a retiring producer is the judgment that knows what a client said they needed versus what they are actually exposed to. That is what Zywave says AI cannot replace, and why its answer to "is AI coming for the broker's seat?" is no. Insurance is still a product that must be sold, not bought. Winners will move their remaining humans up to judgment and let software carry the paperwork.
Third, the study draws the accountability line this series keeps returning to. Zywave notes that no insurer has emerged to underwrite an AI system's mistake at the point of bind — the licensed human still owns the decision that commits capital. That is not a temporary gap waiting for a vendor to fill. It is the operating model: agents pre-qualify, a licensed person confirms, the ledger keeps the tape. Any redesign that lets software bind on its own authority is not efficiency. It is unpriced, unowned risk, and no carrier in the study will insure it.
Fourth, your distribution partners are sorting themselves while you watch: the largest brokerages consolidating premium and talent, small and mid-sized agencies becoming acquisition targets if they do not modernize, and specialists pulling away into defensible niches. If your smallest partners cannot modernize, their books and judgment will move with them. If your largest consolidate, they will arrive at your next renewal with more data and leverage than you have on them.
From the operator's chair
One: stop budgeting AI in distribution as headcount saved, and start budgeting it as capacity released. Pick one servicing journey — endorsements, renewals, certificates — and measure what matters: handoffs, approvals, and waiting days between a request and a decision. That is the coordination tax Nº 009 named, and the only number a rebuild can move.
Two: protect the bind seat explicitly, in writing, before any agent touches it. One document, in your words: what agents may prepare, what a licensed person must confirm, what evidence the ledger keeps, and how the mandate is suspended. Zywave's finding is your cover: nobody will underwrite an AI system's mistake at bind, so no efficiency case justifies letting one bind alone.
Three: reprice your distribution relationships around the four shifts, not last year's commission grid. Which partners get live-data pricing first? Which earn higher contingent commission for a better-selected book, not just a bigger one? Which small and mid-sized partners will you help modernize before someone else acquires them? Zywave's chief product officer said the winners are not the biggest AI spenders but the ones who get their own data in order first. Fix the foundation underneath distribution before buying the next tool on top of it.
The broker is not being replaced. The broker's economics are. The carriers that understand the difference will keep their distribution, their judgment, and their margin. The rest will keep paying for the paperwork and wondering where the $100 billion went.
— Krishna