Delphina Raises $7.5M to Scale Governed AI Analytics

Approving a definition inside Delphina can change answers for colleagues the expert has never met. The San Francisco AI analytics company is building its Context Factory around that recurring act of judgment, using agents to maintain business knowledge while people retain responsibility for important changes.

Delphina announced $7.5M in new funding on November 12, 2026, bringing total funding to $15M. Radical Ventures and Costanoa co-led the financing, with Tenacity Capital participating alongside individual angels including Fei-Fei Li and Guido Imbens. The capital will support Context Factory expansion and additional go-to-market and customer engineering staff, putting implementation work beside the product's broader ambitions.

What Delphina's new funding supports

Delphina describes Context Factory as a way to build and maintain the business context that AI analytics needs. A warehouse contains data, but the meaning of a customer, a transaction or an active account can depend on decisions held elsewhere. Delphina's proposition is to make that understanding available to the software answering questions, then keep it current as the organization changes.

The founders' announcement says agents establish an initial knowledge baseline from existing company material, investigate uncertain answers and conflicting definitions, and draft corrections. Important changes go to the appropriate expert for approval. The resulting knowledge is versioned and reused, so a correction can influence later analysis rather than remain inside an isolated conversation.

That gives customer engineering a substantial assignment. A new deployment encounters an enterprise's existing definitions, permissions and habits, all of which affect whether an answer is useful. Expanding that team alongside go-to-market staff suggests a commercial effort that includes helping customers make their knowledge usable, as well as finding more customers willing to try the software. Delphina has not disclosed a hiring target or a timetable for this expansion.

The founders have lived with widespread data access

Co-founder and CEO Jeremy Hermann built Uber's Michelangelo machine learning platform and co-founded Tecton. Co-founder and President Duncan Gilchrist led global pricing data science at Uber. Their official company story connects Delphina's direction to the advantages and engineering demands they encountered when data informed work across Uber.

The financing also has a history. Delphina's December 2023 seed announcement described a $7.5M round co-led by Costanoa Ventures and Radical Ventures for a data-science copilot. The November 2026 financing adds another $7.5M and takes cumulative capital to $15M. The current announcement does not identify the latest financing as a particular series or disclose a valuation.

The continuity is the founders' interest in getting more useful work from enterprise data. The current product puts greater emphasis on the knowledge that makes an answer dependable for a business user. An executive can ask a short question with a very long organizational history behind it. Someone still has to establish which definition applies, and a system serving more colleagues needs a way to carry that decision forward without asking the same expert to reconstruct it each time.

More questioners change the data team's work

Delphina reports that business teams now ask 73% of the questions on its platform and that CEOs at more than half of its customers are active users. These are company-reported adoption figures. They describe who is using the system, rather than independently establish its accuracy, financial return or competitive standing.

The Substack customer account provides a more specific view of the workflow. Delphina describes questions and answers shared through Slack, with query visibility and a way for the analytics team to correct the context. Leadership's questions become visible to the people responsible for the data, giving those people an opportunity to understand the direction of an inquiry and contribute before assumptions spread.

That changes the relationship around a request. An analyst can learn something from the question itself: which issue has become urgent, what another team is trying to understand, or where a familiar metric has acquired a new purpose. Faster access can therefore produce additional work worth doing. The expert has more opportunity to examine meaning and exceptions while colleagues explore the business directly.

Delphina names customers across publishing, education, healthcare, financial services and aviation. Those organizations bring different questions and operating requirements. The shared commercial problem is making company knowledge reusable while leaving someone responsible for its interpretation.

Governed context has to fit the existing enterprise

Delphina's integration catalog lists warehouses and databases including Snowflake, Databricks, BigQuery and PostgreSQL, alongside documentation, business intelligence and collaboration systems. Those connections matter because useful context already sits in several places. A definition may appear in a dashboard, a document or a discussion, with each supplying a different part of the explanation.

The company also describes controls on its security page, including role-scoped read-only warehouse access, audit logging and isolated code execution. Delphina says it supports deployment within a customer's VPC and does not use customer data to train shared models. Those are product and security representations from Delphina, which enterprise buyers can assess against their own requirements and diligence.

For a data leader, the purchase reaches into ownership as much as interface design. Who can approve a definition, where that approval is recorded and how later changes reach dependent answers are practical deployment questions. They become more consequential as executives and operators ask questions beyond the dashboards the data team originally prepared.

Delphina is funding a larger effort to bring that workflow into more enterprises. At each one, a person who understands an unusual business rule may find their next contribution becoming part of the context used by a much wider group of colleagues. The knowledge keeps acquiring new readers as the questions move around the company.