Sandy Kaul, Franklin Templeton's Head of Digital Assets and Innovation, published a paper on July 22 arguing that agentic AI — software that transacts on its own, without a human clicking approve each time — is crypto's "killer use case." Franklin Templeton manages roughly $1.8 trillion. When a firm that size makes the case in writing, it's worth reading closely, and worth separating what's actually demonstrated from what's being pitched ([Decrypt](https://decrypt.co/374055/franklin-templeton-agentic-ai-crypto-killer-use-case)).
There are two claims in the paper. One is an infrastructure argument. The other is an investment thesis. They get bundled together, and only one of them is about plumbing.
The speed argument
Kaul's infrastructure case is straightforward: an agent making thousands of small payments an hour can't wait on rails built for human-paced commerce. Card networks like Visa typically settle in one to three business days. Newer blockchains process 3,000–13,000 transactions per second with immediate finality. For a human buying groceries, a multi-day settlement lag is invisible. For an agent negotiating and paying for API calls, compute, or data in real time, it's the difference between working and not.
That's the part of the argument that's actually about architecture, and it's the same logic that's been building all year: Google shipped an AI agent payment protocol in 2025 backed by the Ethereum Foundation, and the x402 Foundation launched July 14 with roughly 40 organizations attached, including Visa, Mastercard, and AWS. None of that is Franklin Templeton's claim — it's the observable trend the paper is reacting to, and it's the same trend SatoHub's directory tracks as payment-rail resources builders actually reach for.
The projections
The paper also puts numbers on where this goes: AI agents driving 15–25% of U.S. e-commerce by 2030, and agentic commerce as a $3–5 trillion market by the same year. Those are Franklin Templeton's projections, not measured outcomes — nobody has shipped 2030 yet, and forecasts this size are routinely wrong in both directions. Treat them as the firm's stated thesis, not as settled fact, and watch for who else runs the same numbers independently.
Where the argument quietly changes lanes
Here's the sentence that does the real work in the paper: to capture the value of these decentralized networks, the argument goes, investors need to hold the cryptocurrencies issued by those networks. That's not an infrastructure claim anymore — it's a portfolio-allocation pitch, and it rests on a much shakier premise: that a network being useful and a network's native token capturing that usefulness are the same fact. They aren't automatically. Plenty of infrastructure gets used constantly without its associated token tracking that usage in price. "Agents need fast settlement" and "therefore buy this asset" are two different claims stapled together, and the first one being true doesn't make the second one true. That's not a call on any token, up or down — it's just naming the seam where evidence-based architecture talk turns into an investment story wearing the same paragraph.
Why this lands where it does
The interesting part for builders isn't the forecast — it's that a firm this size is now writing publicly about agent-native rails at all, which is a signal that the underlying tooling (x402, agent wallets, onchain settlement) is far enough along to argue about. That's the layer SatoHub's library exists to map: what an agent needs to actually hold funds and pay for things — wallet controls, payment-rail skills, MCPs — tracked with what's live and what's still a pitch deck, not blended together.
What to watch
Watch whether the ~40 organizations behind x402 ship real integrations or just logos on a launch page. Watch whether legacy rails respond with their own real-time settlement layers, which would undercut the speed argument entirely. And watch for actual transaction volume from agentic commerce, not projected volume — that's the number that turns this from a thesis into a track record.
Sources - [Franklin Templeton Says Agentic AI Is Crypto's 'Killer Use Case' — Decrypt](https://decrypt.co/374055/franklin-templeton-agentic-ai-crypto-killer-use-case)