Saphenia comes out of stealth building decision logic for enterprise agents

For immediate release

London, 31 August 2026. Saphenia today came out of stealth. The company builds decision logic for enterprise agents: the layer that teaches an AI agent how a specific organisation decides, and enforces those rules as answers are written.

Charlotte Werger spent a decade running data and AI inside large enterprises; AI leadership at BlackRock, Nike and Richemont, then Chief Data and AI Officer at H&M. That gave her a clear view of how large organisations actually decide, and of what agents were missing when they arrived. So she left H&M to build it.

Enterprises have deployed agents widely for drafting and summarising, and almost nowhere for decisions that carry consequences. No foundation model knows a company's thresholds, which of its sources wins when two disagree, or which cases always go to a person. That knowledge sits in policy documents, in senior people's heads, and in years of precedent, and it does not exist outside the company.

Saphenia captures it one recurring decision at a time. The result is a neurosymbolic architecture that combines knowledge graphs with learned logical constraints, forming compounding institutional knowledge owned by the customer, and portable across any model. Every value an agent produces comes back one of four ways: fulfilled with proof, refused by a rule it names, flagged for what was missing, or handed to a person because that call was always theirs. An answer that would break a business rule is never produced.

Putting AI agents to work in the enterprise remains difficult, despite ever improving model capabilities, as businesses require high precision, accuracy, and control rather than AI slop. We are scaling these capabilities so business users stay in control and leverage agents to improve their work without the added risk or overhead.

— Charlotte Werger, Founder, Saphenia

Saphenia is working with clients in private equity, financial services and pharma. The company was founded in November 2025 by Charlotte Werger as a solo founder. She has since built a founding team across product, sales, go-to-market and engineering, drawn from Google DeepMind, Databricks, Artefact and HSBC.

Backers include respected European AI founders, operators and angel investors who backed early DeepMind, Faculty, n8n and Lovable, and scouts from Sequoia and a16z.

About Saphenia

Saphenia builds decision logic for enterprise agents, so that organisations can hand their most complex decisions to an agent and know how each answer was reached. The company is based in London with a distributed team across Europe and North America.