Finster AI

Finster has won the validation that matters most to banks. FactSet invested in March 2026 and uses Finster as a core engine of FactSet AI for Banking, and UBS Investment Bank joined as a strategic investor in the Series B in August. The product combines event-driven agents, PowerPoint automation and cell-level citations, built by a founder from Google DeepMind and a COO who ran research operations at Morgan Stanley.

The enterprise design is strong; the public proof is thin. Finster describes single-tenant VPC deployment inside the client's infrastructure, bring-your-own model, SAML SSO and SCIM, and a contractual no-training clause — one of the most complete deployment stories in our coverage. We did not locate a named customer, a case study, published pricing, a SOC 2 report date or a trust centre, and the company had no open roles at our review.

We lower the rating to Emerging, at 6.26, with an Improving outlook. This Rating Review applies the full sub-criteria standard used for every company covered since October and replaces the 6.83 published on 20 October. Under that standard, controls and evidence not located publicly score low until documented. The largest moves are in Enterprise Readiness, now 6.43, where we did not locate an MNPI or information-barrier control for a product sold to investment banks, and in AI & Technology, now 6.50, where no evaluation has been published. Deployment and identity controls remain among the strongest in our coverage. We located no material news since the August Series B; the position stays in the Emerging quadrant.

Strengths

• Incumbent validation. FactSet invested and uses Finster as a core engine of FactSet AI for Banking; UBS Investment Bank joined as a strategic investor in the Series B.
• Distribution through FactSet. FactSet AI for Banking entered preview with select clients in March 2026 and is rolling out through 2026, giving Finster potential reach without a large sales force.
• Deployment flexibility. Single-tenant containerised VPC inside the client's infrastructure, bring-your-own model and US or EU regions — confirmed in the FactSet release.
• Identity and governance controls. SAML SSO, SCIM, MFA, Entra ID and Google Workspace sync, role-based access, a full audit trail and a contractual no-training clause.
• Event-driven agents. Triggers run multi-step workflows when events such as an earnings transcript arrive and deliver the result by email.
• AI and finance leadership. The founder spent seven years at Google DeepMind and worked on Gopher; the COO was Morgan Stanley's Global COO of Research.
• Licensed content. PitchBook, Preqin, Third Bridge, MT Newswires and Crunchbase, with FactSet data and broker research inside the FactSet edition.

Cautions

• Customer evidence. We did not locate a named production customer, case study or user review.
• Attestation detail. SOC 2 Type II compliance is claimed without a report date, auditor or trust centre; ISO 27001 and a subprocessor list were not located.
• Information barriers. No explicit MNPI or information-barrier control was located, which matters for sell-side deployment.
• Data-use terms. The terms prohibit training on user data but allow its use to improve search and accuracy, and the sharing of anonymised aggregate usage data with data providers.
• Standalone versus FactSet edition. The strongest Excel, PowerPoint and broker-research features appear tied to FactSet AI for Banking.
• Scale. Headcount is not disclosed, no roles were open at our review, and the Series B amount was not announced.
• Evaluation evidence. No benchmark or accuracy methodology has been published.
• Pricing. Pricing is not published; fees are set by order form and paid annually in advance.