OpenAI invests in Poetic to automate underwriting and compliance
▼ Summary
– AI startup Poetic emerged from stealth with $50M in funding and a $500M valuation, backed by OpenAI, Founders Fund, and Kleiner Perkins, to automate sensitive financial back-office work like underwriting and fraud checks.
– Founded by former Google and Waymo engineer Markie Wagner, Poetic’s system ingests a firm’s procedures and training videos, runs processes end-to-end, and self-repairs when software changes, using a custom programming language for AI direction.
– Named customers include SoFi, AIG, and Chime, with claimed results of 100% accuracy on automated fraud decisioning and $200M yearly savings for one Fortune 500 firm, though these figures are hard to verify due to limited disclosure on system operations.
– Poetic claims client data stays in the customer’s own environment with zero retention, and it complies with SOC 2 Type II, PCI, HIPAA, and GDPR regulations.
– OpenAI’s investment reflects its strategy to fund application-layer startups that extend its technology, as Poetic enters the competitive enterprise workflow automation field where the shift to agentic AI in finance is underway.
A stealth-mode AI startup has stepped into the spotlight with a hefty $50 million in funding and a valuation of $500 million, aiming to transform the back-office operations of the financial sector. Poetic, the company in question, is targeting sensitive tasks like insurance underwriting, compliance, and fraud detection, automating them with a system that learns from a firm’s existing procedures. Its investors are a heavyweight trio: OpenAI, Peter Thiel’s Founders Fund, and Kleiner Perkins.
The company was founded by Markie Wagner, a former machine-learning engineer at Google and Waymo who previously led the AI consultancy Delphi Labs. Poetic’s core promise is straightforward: “turn your business into software.” The platform ingests a company’s procedures, training videos, and expert feedback, then runs processes from start to finish. When the underlying software changes, the system reportedly repairs itself, a capability made feasible at scale through a custom-built programming language designed to direct AI.
The client list is equally impressive. Named users include SoFi, AIG, and Chime. Poetic claims remarkable results: 100% accuracy on automated fraud decisioning, over 99% quality in processing insurance broker quotes for AIG, and $200 million saved annually on fraud detection for a Fortune 500 financial firm. “What Poetic has built is genuinely different,” said Leigh Marie Braswell, a partner at Kleiner Perkins, noting that it executes complex processes “with accuracy that exceeds what human teams can deliver.”
These figures are the company’s own, and they are substantial. However, Poetic has shared little about how its system actually works, making those claims difficult to verify. This opacity is significant when the pitch involves automating regulated, high-stakes decisions. The company insists that client data remains within the customer’s own environment with zero retention, and it lists compliance with SOC 2 Type II, PCI, HIPAA, and GDPR.
The involvement of OpenAI may be the most telling detail. OpenAI has evolved into a platform company with a dedicated investment arm, seeding application-layer startups that run on its models. From agent-swarm builder Isara to its own $10 billion private-equity deployment vehicle, each investment deepens a network of companies that both extend and depend on OpenAI’s technology.
Poetic is entering a rapidly growing category: enterprise workflow automation. It is far from alone, and the shift from generative to agentic AI in finance is already underway. The central question is whether “100% accuracy” can survive the scrutiny of regulators, auditors, and the messy edge cases that make underwriting and compliance inherently difficult. For now, the money, and OpenAI’s name, are betting that it can.
(Source: The Next Web)