- Poetic raises $50M Series A at a $500M valuation led by Kleiner Perkins, with OpenAI, Founders Fund, and First Harmonic also investing
- Founder Markie Wagner — a Thiel Fellow and former ML engineer at Google and Waymo — built Poetic after years of watching enterprise AI fail the moment it met a real workflow
- The San Francisco startup hit an eight-figure run rate in 2025 with just four employees, posting a 100% pilot-to-production conversion rate at SoFi and AIG
Markie Wagner had a front-row seat to one of enterprise tech’s most persistent failures. As an ML engineer at Google and then Waymo, she watched AI systems that looked impressive in controlled environments fall apart the moment they hit the complexity of real operations — multi-hour processes, thousands of unwritten rules, workflows where being wrong 5% of the time is not an option.
She left, founded Delphi Labs, an ML consultancy deploying AI for some of the world’s largest organisations, and kept watching the same thing happen. The problem was not the models. It was the architecture. Autonomous agents, reasoning freely through high-stakes workflows, were simply not reliable enough for the work enterprises actually needed done.
So she built something different. That company is now called Poetic — formerly Forge — and this week it closed a $50M Series A at a $500M valuation, led by Kleiner Perkins, with participation from OpenAI, Founders Fund, and First Harmonic.
The fix: software that learns like AI but runs like code
Poetic’s core departure from the rest of the enterprise AI market is architectural. Rather than deploying a model that reasons autonomously through a workflow, Poetic built a proprietary programming language that lets operations teams describe their most complex processes in natural language. The platform then converts that logic into deterministic, near-tokenless execution — controllable, auditable, and repeatable at scale.
The target is the work that has defeated everything else: fraud investigations, transaction monitoring, compliance checks, insurance reviews — multi-hour processes that run thousands of times a day and carry near-zero tolerance for error. As experts have noted, the investment opportunity in AI is shifting almost exclusively to vertical, workflow-native solutions — precisely the category Poetic is building into. Wagner’s argument is that these workflows are not just technically hard. They are full of institutional knowledge that nobody ever wrote down, accumulated over decades, and impossible to capture in a prompt.
“Right now in AI, there’s too much attention on quick demos and shiny objects, and not enough on outcomes. We built a new kind of software that learns like AI but runs like code, so the hardest work in your business finally gets done reliably.” — Markie Wagner, CEO, Poetic
What production actually looks like
The numbers Poetic is posting in live deployments are the kind that tend to end investor debates. At SoFi, the company reached 99%+ quality executing fraud investigations end-to-end in five weeks. AIG is a named customer on complex, multi-hour insurance processes that previously required significant manual effort. The company reports double-digit millions in savings for Fortune 500 clients — and a 100% pilot-to-production conversion rate across every engagement it has run.
That last figure matters more than it might appear. The global AI agents market is projected to grow from $7.84 billion in 2025 to $52.62 billion by 2030 — but the enterprise AI graveyard is already full of pilots that impressed in demo conditions and quietly stalled before reaching production. A perfect conversion rate, if it holds as Poetic scales, is the clearest possible signal that the reliability problem Wagner identified has actually been solved.
The company reached an eight-figure run rate in 2025 with four employees — a ratio that says as much about product leverage as it does about capital efficiency.
Why Kleiner Perkins backed it twice
This is not Kleiner Perkins’ first bet on Poetic. The firm recently raised $3.5B across two new funds, cementing its position as one of the most active AI investors of the current cycle. It led Poetic’s seed round too, and partner Leigh Marie Braswell backed the company personally before it had a product to show. Doubling down at a nine-figure valuation, with OpenAI and Founders Fund alongside, is a statement of conviction that goes beyond a standard follow-on.
“Markie is one of the most prescient founders I’ve encountered on AI, and I’ve had a front row seat since the beginning. What Poetic has built is genuinely different — a platform that can execute the complex, high-stakes processes that large enterprises actually run, with accuracy that exceeds what human teams can deliver.” — Leigh Marie Braswell, Partner, Kleiner Perkins
OpenAI’s participation adds a different dimension. As a strategic investor, its presence signals that Poetic’s deterministic, low-token execution model is seen as complementary to — rather than competitive with — the broader LLM ecosystem. Wagner herself is a Thiel Fellow, meaning Founders Fund’s participation closes a circle that began before Poetic existed.
What the funding is for
Poetic will use the capital to expand its forward-deployed engineering team, enter new industries beyond financial services, and scale within its existing customer base. The global intelligent process automation market — the broader category Poetic is competing in — is forecast to reach $35.8 billion by 2030, with financial services, healthcare, and government as the three largest verticals. Healthcare authorisations, government processing, and supply chain decisions are the kinds of workflows the company is eyeing next.
Wagner has been direct about the bar she is holding herself to: “The enterprise AI landscape is littered with pilots that never made it to production. For us, we’ve had a 100% pilot-to-production conversion rate. Our technology works, and we hire the best of the best — because we’re not here to run pilots. We’re here to transform businesses.”
The company that started with one founder watching AI fail in production, again and again, now has $50M to prove the fix works at scale.
