Gemini Enterprise goes vertical: skills are the reusable unit
Google Cloud shipped Gemini Enterprise for Financial Services and Legal in preview. The reusable skill — not the vertical branding — is the pattern worth copying.
Google Cloud shipped two vertical agent products on August 25: Gemini Enterprise for Financial Services and Gemini Enterprise for Legal, both in preview. The industry packaging is the marketing. The engineering worth stealing is one layer down — Google's unit of agent work is a skill, a versioned package of instructions and context that teaches an agent to do exactly one specialized task the way your organization does it.
What actually happened
Per Google's financial services announcement, the Financial Research agent ships with 50-plus foundational skills and returns outputs with confidence scores, an explicit methodology, a data snapshot for audit, and source citations. It reaches data through MCP connectors configured inside the customer's own environment — FactSet, S&P Global, Moody's, MSCI, PitchBook, SEC EDGAR, Google Workspace, Microsoft 365 — and Google states access is bound by existing role-based entitlements. Deutsche Bank and CME Group are named design partners; BNY, Citi Wealth, Lloyds Banking Group, and Macquarie Bank are named as Gemini Enterprise customers.
The legal edition follows the same shape with different skills: contract redlining against a firm playbook, regulatory horizon scanning, DSAR fulfillment, redaction for motions to seal. Connectors run to iManage, NetDocuments, Docusign, Everlaw, RelativityOne, and CourtListener. Launch customers are Cleary, Freshfields, Weil, and Williams & Connolly. Google says healthcare and life sciences editions follow.
Why reusable agent skills matter for your business
A skill is tribal knowledge with a version number. Every business already has these: the way you quote a job, the fields you check before a PO goes out, the tone your refund emails use. Today they live in someone's head and a Google Doc nobody opens. Written as a skill — instructions plus context plus a defined procedure — they become something an agent executes identically at 2am and something you can diff when it changes.
The entitlement line is the real product claim. "Access bound by existing role-based entitlements" means the agent inherits the permissions of the person who asked, not a service account with the keys to everything. That is the single question to put to any agent vendor you evaluate. If they answer with a shared API token, you are not buying an assistant. You are buying privilege escalation with a chat interface.
Copy the output shape even if you never buy this. Confidence score, stated methodology, data snapshot, citation. Four fields turn an agent answer from a claim into something a human can review in thirty seconds. Most in-house agents return a paragraph and nothing else.
Preview is not production, and design partner is not deployed. Both products are in preview. A named law firm shaping a feature is not the same as that firm running it on live matters. Plan on that timeline.
Key takeaways
- Gemini Enterprise for Financial Services and for Legal launched in preview on August 25, 2026
- The Financial Research agent ships with 50+ skills and returns confidence scores, methodology, data snapshots, and citations
- Data access runs over MCP connectors in the customer's own environment, bound by existing role-based entitlements
- Write your recurring procedures as versioned skills — they outlive whichever agent runs them
- Ask every agent vendor whose permissions the agent runs under. A shared token is a wrong answer
Your procedures are already written — just not where an agent can read them. We turn the checklists your team runs by hand into versioned skills and MCP connectors scoped to real permissions, on a stack you own. See how we build it.
Sources: Google Cloud, Google Cloud.
- #gemini-enterprise
- #vertical-ai
- #mcp
- #agent-permissions
- #google-cloud
Tommy Rush — Founder, Rush Commerce
Operator turned builder. 15+ years running operations — now shipping the systems businesses run on. More
Get The Rush Report weekly — one email, zero fluff.
Keep reading
Thomson Reuters built its own model for $40M — data is the asset
Thomson Reuters trained a proprietary LLM on Westlaw and Reuters content for $40M and owns it outright. The lesson for operators isn't build-your-own — it's what you own.
Read itStability AI raises $76M from the labels: licensing is the product
Universal, Sony, Warner, and EA invested in Stability AI. When rightsholders become shareholders, provenance stops being a footnote in AI image generation.
Read it