Sapiom’s cover photo
Sapiom

Sapiom

Financial Services

San Francisco, California 1,465 followers

The foundation of how agents interact with the real economy

See jobs Follow

About us

The agentic economy is here, but it cannot progress until AI agents can autonomously access and operate across real-world systems and services. Sapiom gives AI agents trusted access to the API economy. We provide the essential infrastructure (KYA, wallets, spending and usage controls, and multi-rail payments) to unlock secure, programmatic commerce for the AI era. Built for the next era of commerce where agents buy, sell, and negotiate on our behalf. We're backed by world-class investors and are hiring exceptional engineers in SF to define the financial infrastructure for the agentic era. Learn more & join us: https://www.sapiom.ai

Website
https://www.sapiom.ai
Industry
Financial Services
Company size
2-10 employees
Headquarters
San Francisco, California
Type
Privately Held
Specialties
AI Payments, Agent Commerce, Financial Infrastructure, Stablecoin Settlement, Payment APIs, Developer Tools, KYA (Know Your Agent), Programmable Wallets, Autonomous Agents, B2B Payments, Fintech, x402 Protocol, and Micropayments

Locations

Employees at Sapiom

See all employees

Updates

  • View organization page for Sapiom
    Sapiom

    1,465 followers

    Pitchbook put a number on the machine economy. With a big, fat capital T. ~$20,000,000,000,000. ~$20 trillion of the world's knowledge work is already "AI-exposed." Agents handle about 1% of it today; GDP is already running at ~$36B a year, and that's at one percent. The crazy thing is that the market doesn't cap out. The cheaper thinking gets, the more of it the world buys, and nobody's found the ceiling. Translation: the machine economy is real, and it's 99% unbuilt. Part 1 of 2.

    View profile for Ilan Zerbib
    Ilan Zerbib

    PitchBook's Rudy Yang dropped a fascinating report yesterday and put a number on a market most people don't know exists yet because it's so new. For starters, they call it the machine economy: a market where AI agents discover, transact, and generate value with little to no human involvement. The number: ~$20 trillion of the world's knowledge work is already "AI-exposed" with AI agents handling about 1% of it today. Read that again. Not 1% of some speculative future. 1% of work agents could be doing right now. Every TAM slide I've ever seen tops out somewhere. This one doesn't. The cheaper thinking gets, the more of it people buy, and there's no real limit to how much thinking the world wants done. As QED's Amias Gerety put it, the biggest phase of any technology wave is never the cost savings. It's much bigger than that. It's the "completely new forms of economic activity" that only become possible once the thing gets cheap enough. For cognition (the raw thinking and reasoning itself) we are at the very beginning of the phase. The actual starting line. Agent GDP today: ~$36B a year and compounding (PitchBook). That's at 1% penetration. 1%!! The machine economy isn't a prediction anymore. It's a market that's 99% unbuilt. And that 99% is why I wake up, go to Equinox in the morning, and spend the rest of my day building with the crew at Sapiom HQ. Check out the full thing. I'll break it down even more in my next post.

  • Sapiom reposted this

    View profile for Ilan Zerbib
    Ilan Zerbib

    Thursday night, we opened our hackathon to outside builders for the first time. Which, if I'm being honest, is invigorating and a bit anxiety-inducing at the same time, even when you know what you've built genuinely works. It's like launching your first-ever art gallery and hoping people 1. show up 2. like the art and 3. buy the pieces. All that said, it ended up being one of the clearest proofs of why Sapiom exists. There were builders at all different stages, able to see what their agents were doing and what they cost. - One founder who joined us used Sapiom to build a hotel-enrichment flow, pulling ownership, tech stack, and signals from reviews for each property, at a few cents per hotel. - Someone else built a book recommendation agent that learns your taste over time. - Another team member, a communicator by nature, had never built anything outside of an LLM before and ended up building and deploying a daily 7am report that scans the web for AI failure stories to help inform his content strategy. It went off this morning without a hitch. Everyone can get an "AGI moment" in a demo. When you put real users on it, though, you can't see what failed, where it failed, or what it cost. You hope and pray the bill you get at the end won't cause heart palpitations. That dread is everywhere right now. Gartner expects that over 40% of agentic AI projects will be canceled by the end of 2027, citing escalating costs, unclear business value, and inadequate risk controls. 👆 That's the gap Sapiom was built to close. The hackathon was equal parts energetic, constructive, and community-first. My job was mostly to stay out of the way. David, Jordi, and Evelyn ran the night, and they were phenomenal. We're thinking about running these monthly. If you want in on the next one, let me know below. 

    • No alternative text description for this image
    • No alternative text description for this image
    • No alternative text description for this image
  • View organization page for Sapiom
    Sapiom

    1,465 followers

    Our team spans between France and San Francisco, so when a community that spent it's first year building bridges between those two cities threw its one-year anniversary, it was an easy room to be in. ✨ Sapiom sponsored Intertech Talents' 4th edition in Paris, with Nikita Dmitrieff from our technical team helping bring our story to builders passionate about the future of autonomous agents. Want to build the execution layer, and not just another brain? We're hiring. 👇 https://lnkd.in/g2q7_zFr

    • No alternative text description for this image
    • No alternative text description for this image
    • No alternative text description for this image

  • Sapiom reposted this

    View profile for Ilan Zerbib
    Ilan Zerbib

    One of my favorite parts of building a company where complex agents actually run outside a sandbox is our hackathon nights. The team has the simple task to “build something,” all on Sapiom. This week, one of the team members built an agent that scrapes trending stories off X and re-illustrates every single one as a 19th century French engraving. Old ink-and-watercolor style. It reminds me of something you would have seen in Le Petit Journal 150 years ago, except it's rendering whatever’s blowing up on X in the last 24 hours. He had an idea, described it in one prompt, and by the end of the night it was live and filterable by category, time window, and edition size. Other projects included: - A fully autonomous property manager, a “brain” workflow running on a cron that spins up listings, reacts to incoming events, and manages a fleet of smaller workflows underneath it, removing a human from needing to manage it day to day.  - A lead-gen engine where one loop finds and verifies leads, while a second loop emails them on its own schedule. On the surface, these three projects have nothing in common. Dogfooding is what ties them together. An assigned ticket gets you exactly what you expected to test for. A hackathon idea someone actually came up with has an emotional resonance to it, a true human-first interest in something they thought of themselves, and that's the version that drags the product into corners nobody would think to check on purpose. A woodcut-news generator leans on search, image gen, and caching in ways a property manager never will, and a cron-driven brain workflow managing other workflows breaks completely different things. Running all of it in the same eight hours means we find our own walls before a customer does.  Infrastructure is highly underrated. I love it.

  • Sapiom reposted this

    View profile for Ilan Zerbib
    Ilan Zerbib

    Karp’s right that enterprises are “paying for tokens that create no value.” IMO, he’s partially wrong about the fix. His answer is discipline. Data sovereignty, more careful spend, tighter governance. But that puts the burden on the enterprise to manage something they can’t actually see. You can’t ask a team to exercise judgement over a system they have no visibility into. It's like sending your 4 year old off for the day with no guidance, no boundaries, no idea where they even went, and expecting them back with no scrapes, no bruises, and definitely no new pet snake in their pocket (swap to a cheaper babysitter and I'm sure the kid may still come home with a snake). And watching from the front porch doesn’t fix it either. A camera pointed at the front door tells you the kid left. It doesn’t tell you where they went or why they came back with a snake. Watching a black box fail more slowly is still watching it fail. What actually changes the outcome is infrastructure that gives you real access at the task level: what ran, what it cost, whether it succeeded, how the routing decision got made. Not a dashboard. Control. That infra? It’s been built. And I’m damn proud of the Sapiom team for building it. 

  • View organization page for Sapiom
    Sapiom

    1,465 followers

    79% of agent projects never make it to production. This is why.

    View profile for Ilan Zerbib
    Ilan Zerbib

    In 1886, after centuries of horse-drawn carriages, Karl Benz introduced the first gas-powered automobile. It was a genuine disruption. It was also, if you look at one today, just another carriage built with the same shape, same bench seat, and a small engine replacing the horse that used to propel it forward. For fifteen years, the most disruptive machine of the century wore the body of the thing it replaced. That's where most agent infrastructure is today. Most of what agents run on right now was built for a different world, one that assumes a human is present at every step. I spent years inside payments at Shopify watching that assumption get built into everything, and rightly so for the world it was built in. Every API key has a human who provisioned it. Every billing relationship has a human who approved it. Every permission has a human somewhere in the loop who signed off. Agents break that assumption at every single step. The industry's answer so far has been to put 'agent' in front of the old tools. It doesn't change what they are. The engine is new…but the carriage is the same. So here's a test I'd offer anyone evaluating agent infrastructure right now, their own or a vendor's. Call it the horseless carriage test: 1. When your agent needs a new capability, does that mean a new vendor, a new key, a new billing relationship, or another sprint building it yourself? Or is it one step? 2. Was the governance designed for software that acts on its own, or inherited from identity systems built for people? 3. Can you see what an execution actually did, step by step, or only what everything cost in aggregate? If the honest answers are "new vendor every time," "inherited," and "aggregate," you're driving a horseless carriage. Building it all yourself just means you're the carriage maker. We started Sapiom by refusing to inherit it.

    • No alternative text description for this image
  • View organization page for Sapiom
    Sapiom

    1,465 followers

    Friends don't let friends 𝘩𝘺𝘱𝘰𝘵𝘩𝘦𝘵𝘪𝘤𝘢𝘭𝘭𝘺 duct tape their systems together. 😳 Stay with us for a minute. Right now, agents are in production, but the infrastructure around them is virtually taped together. Any team shipping agents at scale hits the same wall: a massive model-provider bill at the end of the month with no breakdown by agent, no attribution by execution, and no reliable way to tell which agents are working, which are failing, and which are quietly burning money. So, what happens? Teams rebuild the same plumbing over and over: sandboxes, state, spend controls, capability access, observability, payments, governance. It's necessary work, but it's not the thing that makes their product special, nor is it probably what they want to be focused on. The cost ceiling is becoming a real headache, and plenty of teams are keeping people on these workflows, not as a fallback, but because it's the right call while running agents at volume is still too expensive and too hard to reason about. Yesterday, Sapiom handled 1.5M+ agent transactions across ~40K tenant sandboxes for ~40K active agents, including 662K governance checks and 654K paid transactions. We have built the end-to-end platform where you can create the agent, run it on managed infrastructure, connect the capabilities it needs without stitching together API keys and vendor relationships, and see what it did, what it cost, and where it failed, at every step. Sapiom is the platform where agents actually run, allowing the folks building to....build. All without duct tape. 😎

    • No alternative text description for this image
  • View organization page for Sapiom
    Sapiom

    1,465 followers

    Most agents look great in a demo, but when you need to scale that’s when the real test starts. Can it keep running at 2am when a provider drops mid-task? Does it hold onto what it was doing through the 10,000th execution? And when something fails three steps deep, does anyone find out before the customer does? Plenty of agent stacks survive the demo, fewer survive a month in production.

Similar pages