Finster AI

Agents do the core work. Explorer reasons across documents and data to answer analyst questions, Tasks run research workflows on their own when a schedule or an event fires, and PowerPoint Automation turns the analysis into decks in a firm's format. All of it sits on Finster's own retrieval pipeline, which blends semantic and keyword matching across its document collection. Without the models there is a document store and nothing that reads it.

Tasks run multi step workflows such as earnings analysis, peer comparison or risk screening on a schedule or when an event lands. A new earnings transcript is detected within minutes of a call ending, the saved workflow runs without anyone starting it, and an email says the output is ready for review. Coverage agents alert bankers to new filings and developments and can update pitch decks and CIMs automatically. Task agents show their plan and the steps they took, but Finster describes no approval step before an automatically updated deck is used.

A proprietary citation method opens the exact table cell or sentence behind a figure in one click, and every insight carries an auditable source trail. Task agents show the plan and steps they took, Explorer keeps its reasoning visible, and Finster's own agent framework keeps an audit trail of data and AI flows. Retrieval blends semantic and keyword matching, with pre filtering and reranking tuned to financial questions. Finster's terms warn that the model can produce incorrect output and leave users to judge its accuracy, and no accuracy figure or evaluation result is given.

FactSet invested in Finster and built FactSet AI for Banking on its technology, launched in alpha for select clients in March 2026, and Kate Stepp, FactSet's chief AI officer, is quoted on the partnership. UBS Investment Bank made a strategic investment as part of Finster's Series B in August 2026, and Greg Peirce, co head of global banking for Asia Pacific and the bank's AI business sponsor, is quoted on it.

Finster says it works with tier one global investment banks and asset managers without naming them. Weaviate, its vector database supplier, adds that a single tenant test with one such bank began within a day. Finster gives no usage or outcome figure.

Finster's terms of service commit that it will not train its AI model on customer data, though they allow customer data and feedback to be used to improve the model's retrieval and accuracy. Each client can run in its own single tenant environment or in a containerized deployment inside its own cloud, and data controls work at the user or organization level within the firm's governance perimeter. Customer data remains the customer's property under the terms. Clients can also bring their own language model.

The privacy policy of Finster AI, Inc., effective July 20, 2026, covers the website and the software platforms. Where staff use Finster under an employer's agreement, that agreement governs the data and the employer controls it, and for users in Europe, the UK and Switzerland Finster acts as controller of the personal information it collects. Data can move to the United States and other countries under safeguards the policy does not name, and retention runs as long as reasonably necessary. No service providers are named, and a data protection officer is appointed through Workstreet. Under the terms of service, customer data remains the customer's property.

Finster states SOC 2 Type II compliance and describes a zero trust model built on least privilege access and multi factor authentication with continuous verification. Fine grained role based access control and audit logging support incident response and threat detection, and vulnerability scanning, patching and remediation run continuously. Third party firms carry out regular penetration tests and security audits. The report itself, its auditor and its period are not available, and there is no trust center.

Finster holds no license. It acknowledges that its banking users routinely handle confidential information and material nonpublic information, yet ties no rule on information barriers, recordkeeping or client communications to the product. Those users include broker dealers whose decks and client messages fall under FINRA communications rules, and Finster does not say how generated output reaches a firm's supervised records.

Companies and markets are the subjects of this work, so the governance questions are accuracy and which developments the agents choose to surface. Proactive agents decide which events across a coverage universe deserve attention, and Finster does not explain how that choice is made or checked. Output also adapts to each user's role, templates, style and tone, with no account of how those learned preferences are reviewed.

A deck carrying a model error reaches clients as the bank's own work. Finster's terms of service, updated September 2, 2026, provide the subscription as is, warn that the model can produce incorrect output and put the duty to check accuracy on the user. Liability is capped at the fees paid in the three months before a claim, or $1,000 for a free evaluation, and the customer indemnifies Finster rather than the reverse. Third party data from FactSet, PitchBook and Crunchbase is used at the customer's own risk.

Data suppliers are named, from FactSet, PitchBook, Preqin and MT Newswires to Crunchbase, whose terms sit as exhibits to Finster's own, and Weaviate names Morningstar and SEC filings among the feeds behind Finster's pipeline. Weaviate Enterprise runs the vector database, holding 42 million vectors in production by Weaviate's account. Clients can bring their own language model, and Finster describes the platform as model agnostic, but the default models and the subprocessors are not named.

Finster connects to a firm's own data, CRM systems and outside providers. PowerPoint Automation builds decks natively, and FactSet AI for Banking carries Finster into Excel and PowerPoint for FactSet's clients. The PitchBook Premium Connector brings in companies, deals, investors, funds, key people, cap tables, deal terms and the limited partners around a company or fund, for clients with their own PitchBook license and single sign on.

Preqin adds fundraising, transaction and fund performance data across private asset classes, and MT Newswires streams real time news on equities, credit, commodities, FX and macro. Identity runs through SAML single sign on, SCIM provisioning and directory sync with Microsoft Entra ID and Google Workspace. No public API documentation is offered.

Single tenant environments are tailored to each customer, and containerized deployments run inside the customer's own cloud, with data regions in the United States and Europe. Customers keep control of their security configurations and policies. The vector database runs on Weaviate Enterprise in single tenant or multi tenant form depending on the customer, and FactSet AI for Banking offers virtual private cloud and single tenant options to FactSet's clients.

Finster prices by quote, and cost per user, per team or per deployment is not given, nor what a single tenant deployment adds. Its terms of service, updated September 2, 2026, cover the paid subscription and allow a free evaluation. Either side can end the agreement for convenience on 30 days' notice, but an active order form runs until it expires. PitchBook data comes through the buyer's own PitchBook license, whether seat based, unlimited or trial, and FactSet AI for Banking reaches banks as a FactSet product.

Investment banking, asset management and private credit teams are the users. Bankers use Finster across the deal cycle for comps, company primers, strip profiles, CIMs and industry reports, and coverage agents keep them current on filings and news. Private markets teams screen funds and deals on Preqin and PitchBook data across private equity, venture, private debt, real estate, infrastructure, natural resources and secondaries.

The document pipeline covers more than 8,000 companies with deep EMEA and APAC coverage, and FactSet AI for Banking carries the engine to FactSet's banking clients. Lending operations, insurance and retail banking are outside it.