APIs are eating the transactional half of broking and leaving the judgement half more valuable than before. An argument for why the best brokers should be cheering.
Whenever I present embedded insurance to an industry audience, the same question arrives, usually politely and always meaning the same thing: so what happens to brokers? The questioner expects one of two answers — the diplomatic one (nothing changes) or the disruptive one (they disappear). I think both are wrong, and the honest answer is more interesting: APIs are splitting the broking profession in half, and only one half was ever the point.
The two jobs hiding inside one title
A broker does two fundamentally different things under one job title. The first is transactional: collecting client information, keying it into insurer systems, obtaining quotes, comparing them, issuing documents, processing renewals. The second is judgement: understanding a risk the proposal form cannot describe, structuring a programme, negotiating terms that do not exist on a rate card, and standing next to the client when a large claim goes wrong.
The transactional half is data plumbing performed by humans. It exists because insurers historically exposed their products through PDFs and portals rather than interfaces, so someone had to be the interface. That job is exactly what APIs abolish — not because anyone dislikes brokers, but because retyping known information between systems is not a profession, it is a workaround.
The judgement half is a different substance entirely. No API structures a contractor's liability programme across three jurisdictions, or knows that this insurer's claims team pays fire claims properly and that one litigates. That knowledge is earned case by case, and clients with complex risks will pay for it long after the last proposal form is retired.
The parts of broking that software replaces were the parts brokers themselves complained about. What remains is the profession.
Where the line actually falls
The split does not fall between personal and commercial lines, or small and large clients. It falls between standardisable and non-standardisable risk. A fleet renewal for identical vehicles, a shop package, a group medical scheme for an office of thirty — these are standard enough to quote from data, and channels that quote them from data at near-zero distribution cost will win them. A refinery, a construction consortium, a shipping line — no checkout will ever hold those, and the brokers who serve them are, if anything, undersupplied.
The uncomfortable middle is the broker whose book is mostly standard risks serviced manually — renewals that survive on inertia rather than advice. For that model, embedded distribution is a genuine threat, and I will not pretend otherwise. The margin that funded manual servicing of standard risks is precisely what automation competes away.
What the smart brokers are doing
The brokers we find most interesting are not resisting the API layer; they are using it. Some are becoming distribution platforms themselves — putting their licence and their client base behind digital journeys for their standard book, freeing their people for the accounts where advice moves the outcome. Others are specialising upward into the risks automation cannot touch, and letting the standard business go to channels that serve it more cheaply. Both moves treat technology as what it is: a reallocation of expensive human attention toward the work that deserves it.
There is also a role that becomes more important, not less. Embedded products are simple by design, and simplicity has edges. Someone needs to tell a growing business the moment its risks have outgrown checkout products — that its single package policy should become a structured programme. That referral judgement is broking, and platforms like ours are natural sources of exactly those referrals, because we see the customer outgrow the product in the data before anyone else does.
Where we stand
Yasmina was built to move standard insurance to where transactions happen. We think that is where standard insurance belongs, and we think the profession that spent a century carrying insurance information between parties by hand deserves better work than that. The broker of the next decade holds fewer relationships, knows them more deeply, and lets machines do the typing. The ones already moving that way are not our competitors. Increasingly, they are our partners.