Gemini Enterprise for Financial Services and Legal: Industry Agents, Skills, Connectors, and Governance
Gemini Enterprise for Financial Services and Legal: the Financial Research agent, skills, data connectors, legal integrations, and governance.
Checked against Google's official documentation.

General-purpose AI assistants struggle in regulated professions for predictable reasons: they don’t have access to licensed data like market feeds or legal databases, they don’t know how your firm does the work, and they can’t prove where an answer came from. On August 25, 2026, Google launched two industry solutions on Gemini Enterprise to address exactly that (Gemini Enterprise for Financial Services and Gemini Enterprise for Legal), both in preview.
This guide explains what’s in each solution, how they’re built (skills, MCP connectors, agents), the governance and confidentiality controls, and what IT and risk teams should evaluate before a pilot.
Both integrate with Google Workspace and Microsoft 365, so results land in the documents and spreadsheets professionals already use.
Key Takeaways
- Both solutions launched in preview on August 25, 2026.
- Financial Services adds a Financial Research agent, 50+ skills, and 13 licensed data connectors.
- Legal targets contract review, regulatory scanning, playbooks, and DSARs, with inherited matter permissions.
- Outputs include citations; the finance agent adds confidence scores and audit snapshots.
- Pilot narrowly and test ethical walls and entitlements first.
Table of contents
Open Table of contents
How Industry Solutions Are Built

Both solutions follow the same pattern on top of Gemini Enterprise:
| Building block | What it adds |
|---|---|
| Skills | Reusable instructions that teach agents to do specialized tasks the way your institution does them (skills guide) |
| MCP connectors | Secure integrations to licensed data and industry platforms, configured in your own environment (MCP guide) |
| Google agents | Google-built, Google-managed agents for core tasks |
| Partner agents | Out-of-the-box agents from industry software providers |
| Platform controls | Gemini Enterprise identity, audit, model routing, VPC-SC, CMEK, residency |
Because they’re built from standard Gemini Enterprise components, firms can also use the same skills and connectors in their own agents.
Gemini Enterprise for Financial Services
What’s included
Financial Research agent: Google-built and Google-managed, automating end-to-end research. Available in the Gemini Enterprise app or through A2A APIs to plug into other agent workflows (A2A guide).
50+ foundational skills: reusable packages of instructions and context that teach the agent to run a task the way your institution does, such as applying your report formatting, pulling a specific data cut, or following a defined research methodology. Available in the Financial Research agent and to agents your teams build.
13 data connectors (licensed sources, configured in your environment; access follows existing entitlements):
| Category | Sources |
|---|---|
| Market data and fundamentals | Daloopa, FactSet, Finnhub, Fiscal.ai, Guidepoint, LSEG, S&P Global |
| Risk, ratings, private markets | Moody’s, MSCI, PitchBook |
| Regulatory and corporate records | SEC EDGAR, Dun & Bradstreet |
| Digital assets | CoinDesk Data & Indices |
Plus productivity integration with Google Workspace (Docs, Sheets, Slides) and Microsoft 365 (Word, Excel, PowerPoint) for exporting results.
Partner agents:
| Agent | Purpose |
|---|---|
| D&B Business Verification | Commercial onboarding and KYC |
| FlowX | Loan pack completeness checks, document reconciliation, and other core banking processes |
| Obin Financial | Faster complex financial analysis for asset management, commercial lending, and insurance teams |
| S&P Global Data Retrieval | Multi-step analysis, report generation, and research workflows |
| S&P Global Horizons | Energy and sustainability data for finance workflows |
Use cases Google highlighted
| Use case | What it does |
|---|---|
| Advisor insights | Insights, recommendations, and materials for relationship managers |
| KYC and onboarding | Ingest PDFs, spreadsheets, and filings to map corporate hierarchies and resolve ultimate beneficial owners |
| Portfolio resilience | Bond portfolio risk exposure analysis in under five minutes (Google’s claim), with duration-hedging suggestions |
| Credit opportunities | Identify potential mispricings for trade ideas |
| Bond issuance pitches | Cut pitch preparation from days to minutes (Google’s claim) |
Explainability and governance
The Financial Research agent’s outputs include confidence scores, explicit methodologies, data snapshots for auditing, and source citations. Google describes a single dashboard for IT and risk teams that enforces security policies (VPC, CMEK), keeps data isolated, and holds outputs to traceable citations. It says customer data, business rules, custom agents, and model outputs stay private to your organization and are never used to train or fine-tune Google’s foundation models.
Who’s using it
Google developed the solution with financial institutions including Deutsche Bank (a design partner for the Financial Research agent) and CME Group. Google also cited BNY, Citi Wealth, Lloyds Banking Group, Macquarie Bank, and Signal Iduna as Gemini Enterprise customers.
Gemini Enterprise for Legal
What it does
| Capability | Description |
|---|---|
| Contract review | Reviews vendor agreements, NDAs, and M&A documents against your playbooks; flags high-risk clauses, suggests redlines, routes for approval |
| Regulatory scanning | Monitors regulators, legislative updates, and court dockets; cross-references changes with internal policies; drafts policy updates for review |
| Playbook creation | Extracts key terms, fallback positions, and institutional knowledge from past agreements |
| DSAR fulfillment | Finds and compiles personal data across connected systems to meet data subject access request deadlines |
Integrations
Open MCP connectors link Gemini to Microsoft 365, Google Workspace, and legal platforms, including Courtroom5, Docusign, Everlaw, CourtListener, Harvey, iManage, Legora, NetDocuments, RelativityOne, Solve Intelligence, and Thomson Reuters HighQ. Google says these work across existing systems rather than replacing them.
Confidentiality and ethical walls
This is the make-or-break question for law firms. Google says:
- Connectors inherit existing permissions, including matter-level permissions, so ethical walls hold automatically: the AI only retrieves what the attorney is already authorized to see
- Responses are grounded in verified internal documents and authorized legal databases, with traceable citations
- Client data, firm playbooks, custom agents, and model outputs stay private and are never used to train or fine-tune Google’s foundation models
- Client data, prompts, documents, and outputs stay within the customer’s private cloud perimeter, and a single dashboard for legal IT and risk teams enforces security policies (VPC, CMEK)
- Google’s product page lists VPC Service Controls, CMEK, Access Transparency, and data residency on Standard and Plus editions
Underneath, Google says the solution runs on a governed control plane, and it works within the permissions and access controls your firm already maintains rather than asking you to rebuild them.
Who’s using it
Google named Cleary Gottlieb, Freshfields, Weil, and Williams & Connolly as firms working closely with the platform. Google also names partners Everlaw and Deloitte as early adopters. Cooley announced at Gemini at Work 2026 a collaboration with Google to develop a Gemini Enterprise agent for complex litigation.
What IT and Risk Teams Should Evaluate
For financial institutions
| Question | Why it matters |
|---|---|
| Which data connectors are we licensed for, and how do entitlements flow through? | Licensed data must only reach entitled users |
| Where is data processed for our regions? | Residency rules for client data (compliance guide) |
| How are confidence scores and data snapshots retained? | Model risk management and audit evidence |
| How do we validate the Financial Research agent? | Model risk frameworks require testing and monitoring (evaluation guide) |
| Which partner agents touch customer data, and under what terms? | Third-party risk management |
For law firms and legal departments
| Question | Why it matters |
|---|---|
| Do matter-level permissions sync reliably from our DMS? Test with walled matters | Ethical walls are non-negotiable |
| How quickly do permission changes take effect? | A lateral hire or conflict must be reflected immediately |
| How are citations verified? | Professional responsibility for accuracy |
| Where are prompts and outputs stored, and for how long? | Client confidentiality and retention obligations |
| How do we handle client consent and engagement terms for AI use? | Client expectations and outside counsel guidelines |
Audit evidence from the platform
Both solutions run on Gemini Enterprise, so its usage audit logs in Cloud Logging are the base audit trail. They record activity such as Search, Assist and StreamAssist requests, agent create, update and delete actions, and data connector updates. An admin with the Gemini Enterprise Admin role (roles/discoveryengine.agentspaceAdmin) turns them on after the app’s observability settings, and reviewers need the Logs Viewer role. Prompt and response text is logged only when Enable logging of prompt inputs and response outputs is on; otherwise those fields show as <elided>. Decide that setting with your records and privacy teams, because turning it on puts client and customer content into your logs. Our compliance guide covers the related residency and key controls.
A Pilot Plan for Regulated Teams
- Pick a narrow use case: KYC research for one desk, or NDA review for one practice group
- Connect only what’s needed: one or two data sources plus the relevant document repository
- Test permissions first: use test accounts on both sides of an ethical wall or entitlement boundary
- Build an evaluation set from past work with known correct outcomes
- Run in parallel with the existing process for four to six weeks
- Measure accuracy, citation validity, time saved, and reviewer confidence
- Document controls and results for model risk, compliance, or general counsel sign-off
- Expand to more desks or practice groups
Turn on Model Armor from day one: documents from counterparties and public sources can carry prompt injection (Model Armor guide).
Building Your Own Industry Agent From the Same Parts
You don’t need to wait for Google to package your industry. The components behind these solutions are available to any Gemini Enterprise customer:
- Write skills for your firm’s methods (your credit memo format, your NDA playbook, your regulatory change template) as
SKILL.mdpackages with reference files. - Connect licensed data through MCP servers. Many data providers already offer them, or your platform team can build one for an internal system.
- Build the agent in Workflow Builder for simple flows, or with ADK on Agent Runtime for complex logic (ADK tutorial).
- Register it in Agent Registry and expose it through A2A so other agents (including Google’s Financial Research agent) can call it.
- Govern it with Agent Identity, Agent Gateway, Model Armor, and an evaluation set built from past work.
For example, an insurance team could combine a claims-assessment skill, an MCP connector to its policy administration system, and an ADK agent that drafts claim summaries for adjusters, following the same pattern Google used for finance and legal.
Common Concerns and Straight Answers
“Will the AI see matters or accounts it shouldn’t?” Only if the source system allows it. Connectors inherit permissions and entitlements, so the real work is making sure those are correct in the DMS, CRM, and data platforms, and testing walled matters during the pilot.
“Can we trust the numbers?” Treat outputs as analyst drafts. The Financial Research agent provides citations, methodologies, confidence scores, and data snapshots precisely so a person can verify before acting.
“Who’s accountable when an agent gets it wrong?” The professional who relies on the output, as with any research tool. Build review steps into workflows and keep audit trails.
“Does our data train Google’s models?” For both solutions, Google states that customer data and model outputs are never used to train or fine-tune its foundation models. Still confirm the data terms in your contract.
“What about cost?” Google hasn’t published pricing for the preview solutions. Ask about licensing for skills, partner agents, and data connectors (data provider subscriptions are usually separate).
What’s Next
Google called these the first in a series of packaged industry solutions. Both launch blog posts said solutions for Healthcare, Life Sciences, and other professional services are on the horizon, and at Gemini at Work 2026 Google said Government, Healthcare, and Retail specializations are coming soon, without dates.
Conclusion
Gemini Enterprise for Financial Services and Gemini Enterprise for Legal show where enterprise AI is heading: packaged skills, licensed data connectors, and specialized agents on a governed platform, rather than a smarter chatbot on its own. The Financial Research agent’s confidence scores and audit snapshots, and Legal’s inherited matter permissions, address the objections that have kept regulated professions cautious.
Both are in preview, so treat them as pilot candidates: narrow scope, rigorous permission testing, evaluation against known outcomes, and documented controls. Firms already running Google Workspace and Gemini Enterprise can add these solutions on top of the identity, data, and security foundations they already have.
Official Google sources:
- Introducing Gemini Enterprise for Financial Services
- Google Cloud launches Gemini Enterprise for Financial Services (press release)
- Gemini Enterprise for Legal (product page)
- Introducing Gemini Enterprise for Legal
- Welcome to Gemini at Work 2026
- Set up usage audit logs for Gemini Enterprise
Related reading:
- Gemini Enterprise Skills: the skills model behind industry solutions.
- MCP Servers in Gemini Enterprise: how licensed data connectors plug in.
- Gemini Enterprise Data Residency, VPC-SC, and CMEK: the controls regulated teams rely on.
This guide contains referral or affiliate links. If you sign up through one, Workspace Blog may earn a commission at no extra cost to you. Read our referral and advertising disclosure.
Frequently Asked Questions
- What is Gemini Enterprise for Financial Services?
- It's a packaged agentic AI solution built on Gemini Enterprise for capital markets and corporate banking, announced on August 25, 2026 in preview. It includes a Google-managed Financial Research agent with more than 50 purpose-built skills, 13 licensed data connectors such as FactSet, LSEG, Moody's, MSCI, PitchBook, S&P Global, and SEC EDGAR, and third-party agents from providers such as Dun & Bradstreet, FlowX, Obin, and S&P Global.
- What is Gemini Enterprise for Legal?
- Gemini Enterprise for Legal, launched in preview alongside the financial services solution, is Google's legal solution on Gemini Enterprise, delivered inside the Gemini Enterprise app. It adds legal skills, agents, and connectors for law firms and in-house legal teams. It targets contract review, regulatory scanning, playbook creation, and data subject access requests, and integrates with systems such as iManage, NetDocuments, RelativityOne, Thomson Reuters, Harvey, Everlaw, Docusign, and CourtListener.
- How do these solutions protect confidential client data?
- Google says connectors inherit existing permissions from connected systems (for legal, including matter permissions and ethical walls), and that customer or client data, custom agents, and model outputs stay private to your organization and are never used to train or fine-tune Google's foundation models. Both solutions run on a governed control plane with a single dashboard for IT and risk teams that enforces security policies such as VPC and CMEK and keeps data isolated. Google's legal product page adds that Standard and Plus editions support VPC Service Controls, CMEK, Access Transparency, and data residency.
- Is the Financial Research agent auditable?
- Google says the Financial Research agent's outputs include confidence scores, explicit methodologies, data snapshots for auditing, and source citations, and that access follows existing role-based controls and data entitlements.
- Are other industry solutions coming?
- Yes. Google called Financial Services and Legal the first in a series of industry solutions. Both launch blog posts said solutions for Healthcare, Life Sciences, and other professional services are on the horizon, and at Gemini at Work 2026 Google said Government, Healthcare, and Retail specializations are coming soon, without dates.