The overnight shift at the metal finishing shop was not where most enterprise software ideas get born. But Sam Goldman and Taira Fujioka weren't there to pitch anything. They slept at the factory. They watched workers wrestle with purchase orders arriving in every conceivable format—scanned PDFs one minute, faxed forms the next, the occasional handwritten note that somehow made it through a supply chain built for the 21st century but operating like it was 1997.
The orders stacked up. Workers typed them, one agonizing line item at a time, into ERP systems old enough to have earnestly used the term "client-server architecture."
That immersion birthed Smartbase, what Goldman and Fujioka are calling an AI-native ERP. The wedge? Automating the purchase order chaos they watched unfold across those night shifts. Six weeks post-launch, the Y Combinator Spring 2026 startup reports $171,000 in contracted annual recurring revenue—a vendor-stated figure that appears on the company's YC page but hasn't been independently verified. One customer alone is paying close to $100,000 a year, plus a $150,000 implementation fee on top of that.
The Unglamorous Middle
What Smartbase does doesn't exactly scream "hot startup." It extracts data from incoming purchase orders and writes it into a manufacturer's existing ERP. That's it, at least initially.
But spend time in a metal finishing shop running high-mix custom work, and you realize manual order entry isn't just tedious—it devours hours daily. Email, scans, faxes (yes, still faxes): the system ingests them all, uses AI to pull line items while applying customer-specific pricing rules and whatever Byzantine business logic has accumulated over the years. Everything routes through a human review step before posting to the ERP. Optional automatic printing of work travelers completes the loop.
The company's internal ROI calculator suggests order entry time reductions of 90 percent, though this claim lacks independent validation.
The integration list is extensive—Acumatica, Epicor, Infor, Microsoft Dynamics 365, NetSuite, QuickBooks, SAP, SYSPRO, Sage X3. But Goldman and Fujioka position Smartbase as more than middleware plugging into legacy systems. They're building, in their words, an "AI-native ERP" that will eventually become the system of record itself. The PO automation is just the wedge, the way in.
For manufacturers in aerospace and defense—where regulations pile up like those overnight purchase orders—Smartbase is developing a US-hosted enclave to handle ITAR, CMMC, and DFARS requirements. FedRAMP compliance is part of the company's future plans but has not yet been achieved.
The Numbers, and Who's Behind Them
The founding team brings credible technical backgrounds. Fujioka was the youngest SDE 2 at AWS SageMaker Inference, building AI infrastructure eventually used by the likes of Salesforce. Goldman previously ran a seed-stage DevOps platform serving teams managing hundreds of millions in assets—though perhaps not the glamorous pedigree of some enterprise software founders, it's relevant operational experience.
Their first paying customer, a metal shop the founders haven't publicly named, signed a contract worth nearly $100,000 annually. Then there's that $150,000 implementation fee, which tells you something about the complexity involved. These aren't simple plug-and-play deployments. For metal finishing shops running three shifts, though, the math is straightforward: calculate what manual order entry costs in labor and errors, and suddenly six figures doesn't seem unreasonable.
Smartbase raised a $500,000 convertible note from Y Combinator, per CB Insights data from earlier this year. The two-person team describes their target as a "$6 billion vertical where people want to adopt a new ERP." That market sizing comes from the founders themselves, not third-party research firms, so the usual caveats apply.
Crowded Already
Here's where the story gets interesting, or perhaps concerning depending on your perspective.
Smartbase is entering a market where the incumbents aren't sitting still. Oracle announced 22 "Agentic Applications" for Fusion back in March. SAP unveiled its Joule AI assistant at Sapphire in May, positioning ERP as both system of record and "AI brain" (their term, not ours). Infor expanded industry-specific AI agents with AWS at Hannover Messe in April. Epicor debuted an agentic platform at its Insights conference in mid-May, promising 90-day cloud ERP deployments—a claim that should be taken with the appropriate grain of salt.
And competing startups? They're flooding in. YC batchmates Comena and Pollinate both launched AI agents for order processing and ERP data entry around February. Horizontal players like Workist and Orderful offer AI-powered order automation without Smartbase's vertical positioning. Even within metal finishing specifically, niche ERPs like Finishing Force already serve the industry with production and compliance tools purpose-built for the vertical.
The central question: does an AI-native architecture built from scratch offer enough advantage over retrofitted incumbents? And can Smartbase move quickly enough in this narrow vertical to establish real defensibility before larger players fully deploy their agent strategies?
It's not obvious either way.
What Comes Next
Smartbase's factory floor research continues shaping product direction, which feels appropriate given where this started. The compliance roadmap for aerospace and defense manufacturers isn't bolted on as an afterthought—it's a deliberate expansion into segments where switching costs run high and regulatory requirements create something approaching natural moats. Or at least speed bumps.
The $171,000 in contracted ARR represents early validation. But the real test arrives in the next six months. Can a two-person team support enterprise implementations while simultaneously building the broader ERP modules needed to become a true system of record? That first customer's $150,000 implementation fee suggests deployments complex enough that they won't scale without significant engineering and services capacity.
For now, Smartbase has what every early-stage vertical SaaS founder wants: a specific problem, customers willing to pay real money, and a wedge into a market the incumbents have arguably underserved. Whether sleeping at factories translates to category dominance depends entirely on execution speed.
In a field that just got suddenly, uncomfortably crowded.
