In financial services, each generational commerce shift creates new rails and new giants. The postwar boom in consumer credit and air travel gave us Visa and Mastercard. E-commerce created PayPal and Stripe. Now, as agents become economic actors, they’ll need their own infrastructure to transact. This opens the door for new rails and developer tools purpose-built for this new class of buyers.

The Shift is Already Underway

The 2025 holiday season marked a turning point in how consumers discover and choose brands, with AI-driven traffic surging in retail and travel. By next holiday season, millions of consumers may use AI agents not just to research purchases but to complete them.

On the enterprise side, autonomous procurement and workflow automation are already moving from pilots to production. The Openclaw/Moltbook saga is an early, messy demo of this shift in the wild. Over the past few weeks, people have been spinning up agents that could discover tools, chain actions, and trigger real usage off simple prompts and loose permissions, in an exhilarating demonstration of what happens when software stops being a tool and starts behaving like an economic actor.

From Agent-Assisted to Agent-Led

Over the past year, a new set of protocols has emerged to support agent-assisted commerce. Google launched the Universal Commerce Protocol with Shopify, Walmart, Visa, Mastercard, Stripe, and others to standardize the journey from discovery through post-purchase. OpenAI and Stripe co-developed the Agentic Commerce Protocol to power instant checkout in ChatGPT. Visa introduced the Trusted Agent Protocol to help merchants distinguish legitimate agents from bots.

These are early versions, and they will almost certainly evolve toward more autonomy. However, today, they mostly inherit a consumer checkout mental model. The agent helps a person decide and then completes a transaction. The roadmaps naturally overweight discovery, authentication, and settlement at the point of sale.

Something more fundamental is happening on the enterprise side. Agents aren't being positioned as a UX layer sitting in front of a human decision. They're becoming the buying authority. They initiate and scale real work across third-party services and internal systems, incurring spend and creating obligations on the company's behalf, continuously, under policy. They don't assist a human buyer. They are the buyer.

Therefore while the emerging e-commerce systems focus on “how the money moves”, enterprise systems must decide “whether money should move at all”.

The Gap Sapiom Fills

Sapiom is built for enterprise agent autonomy. It treats spend as runtime control, embedding financial policy directly into agent workflows so every action is evaluated before cost is incurred.

In practice, that means you can give an agent an account or wallet, define explicit limits and allow lists, and enforce them in real time.

What does this look like in practice? Imagine you’ve built an AI agent that manages your team’s cloud infrastructure. Using Sapiom, you could allocate this agent a wallet (funded in stablecoin or fiat), set a rule that it can spend at most $100 per day on AWS, $50 per day on API calls to external services, and zero on anything outside a predefined list. Every time the agent tries to spend or commit to a service, the Sapiom runtime checks the rules in real-time and either approves, limits, or blocks the transaction. All transactions are logged and visible on a central dashboard. If the agent approaches a limit or does something unusual, you get an alert automatically. In effect, Sapiom acts as a smart financial gatekeeper for AI behavior.

Trust as the Missing Layer

The spend controls are a necessary primitive, but the deeper problem Sapiom is solving is trust.

Imagine a vibe coder, building a new app on their favorite platform. They want to build a richer experience and ship fast, so they start stitching in paid APIs for SMS, data enrichment, image generation, etc. Today that quickly turns into a mess. You have to create accounts across a dozen vendors, negotiate contracts for anything that isn’t pure PLG, manage keys, set limits, and then somehow decide which provider is best. That workflow does not scale when agents are composing software dynamically.

Sapiom sits between buyers and paid API vendors as the clearing layer.

For the buyer (i.e. the agents procuring external APIs), the challenge is vendor selection under autonomy. When an agent needs to call an external API or provision a resource, how does it know which provider to choose? Today that logic is largely hardcoded and/or the result of existing human relationships. Sapiom can act as a routing and trust layer that helps agents choose vendors based on reliability, performance, cost, and policy constraints, informed by network level signals no single organization could generate alone.

For sellers (i.e. paid API providers), the challenge is access control without killing growth. The best PLG businesses want users and now agents to experiment freely, but one bad actor or one runaway agent can create real regulatory and reputational exposure. The question is not only “can you enforce granular policy,” it’s “how do you decide what safe policy should be for this category of agent behavior.” Sapiom pairs enforcement with intelligence so access can be opened up safely.

If agents become the primary integrators, a unified wallet plus routing plus policy becomes a distribution channel. The long tail of niche API providers should want to be inside that alliance because it is how they get pulled into agent workflows by default, without building their own billing, risk, onboarding, and trust stack.

Why We Invested

We first partnered with Ilan during Sapiom's pre-seed. He was a solo founder just weeks out of Shopify but he had a crystal-clear vision that the next big distribution channel has no eyeballs, only APIs. He'd spent years in the weeds of traditional payments, but what struck us was his ability to take that knowledge and project it forward, painting a picture of agentic commerce that gets anyone he talks to genuinely excited.

Since then, he's assembled a strong team, shipped the first product, and brought together an exceptional alliance of partners (more on this soon). Today we're excited to announce Sapiom's seed round led by Accel, and we're proud to double down on our partnership.

When AI agents truly become economic actors, they'll need an entire stack equivalent to what humans and businesses have: identity, wallets, credit, compliance, all optimized for non-humans. It's a once-in-a-generation greenfield opportunity. We believe Sapiom has a real shot at defining those standards early.