The Underwriting Gap
Why Now
Two of 2026's hottest narratives — AI agents and RWA — only become one product at the underwriting layer. This is the moment that layer has to exist.
Two narratives converge#
AI agents can now read messy real-world documents and reason about them at a cost and volume no human team can match. Separately, RWA tokenization has crossed from pilot into live infrastructure. On their own, each is a trend. Together, at the point where an agent underwrites an asset and pays its own way to do it, they become a single product.
- AI agents are the only thing that can underwrite at the volume and cost the long tail demands.
- x402 is the only rail that lets an agent pay for its own compute and data per-deal.
- RWA is the demand — a booming market with no neutral verification layer underneath it.
The long-tail moment#
Today’s RWA market is small and institutional, so manual underwriting still just about works. But the growth is in the long tail — SME invoices, receivables, real estate in emerging markets — where deal sizes are small and volume is enormous. Manual underwriting cannot reach it. Autonomous underwriting can.
The subprime risk#
RWA does not have a subprime problem yet, precisely because it is small and gatekept. The moment tokenization opens up, the missing piece is not more issuers — it is a neutral, verifiable underwriting layer that no single institution controls.
“Everyone is building the vault. Nobody is building the appraiser.”
The window is open now
The appraisal layer has to be built before the long tail arrives, not after the first wave of bad tokenized assets teaches the market why it was needed.