Anthropic expands Claude AI for financial services
Anthropic is pushing Claude deeper into day-to-day finance work. The latest release brings an Excel add-in, real-time market data connectors, and pre-built Agent Skills for modeling and reporting. Backed by stronger financial reasoning, Claude Sonnet 4.5 posted 55.3% accuracy on the Finance Agent benchmark by Vals AI, outperforming other systems on portfolio analysis and equity valuation.
The signal is clear: Claude is moving from chat to actual workflow automation across banking, asset management, insurance, and PE.
AI embedded directly in spreadsheets
Claude for Excel is now in beta. It sits in a sidebar inside Microsoft Excel so you can chat with it while it reads, analyzes, and edits spreadsheets in real time. It explains its steps, tracks every change, and links you to the exact cells it touched.
Analysts can debug formulas, preserve dependency chains, build new sheets, and document logic as they go. This addresses a common blocker for AI in finance: trust and traceability in sensitive calculations. The beta is a research preview for Claude Max, Enterprise, and Teams, with 1,000 early users giving feedback before a wider rollout.
- Use it to audit model logic, reconcile line items, and surface broken links.
- Generate scenario tabs with documented assumptions and switchable cases.
- Convert ad hoc analyst notes into clean, labeled inputs and outputs.
Connecting Claude to live financial ecosystems
Claude now plugs into live data through new connectors, turning it into a market-aware companion. Integrations span Aiera, Chronograph, Egnyte, LSEG, Moody's, and MT Newswires. Aiera brings real-time earnings call transcripts and event summaries, plus expert interviews via Third Bridge. Chronograph adds private equity data on valuations and fund performance. Moody's contributes ratings and coverage on 600M+ companies.
This pulls Bloomberg-style immediacy into a conversational interface. The upside: faster synthesis of headlines, filings, and call commentary into concise takeaways, while cutting manual research and copy-paste error.
- Aiera: Earnings calls, investor days, expert transcripts.
- Chronograph: Private equity fund metrics and valuation insights.
- Moody's: Credit ratings and company data at global scale.
- LSEG, MT Newswires, Egnyte: Market data, news flow, and secure document retrieval.
New Agent Skills for finance
Anthropic rolled out six finance-focused Agent Skills-modular workflows with scripts, data structures, and best-practice logic. They automate discounted cash flow modeling, comparable company analysis, and due diligence checklists. Claude can also produce company profiles, earnings summaries, and full initiating coverage drafts.
Outputs are dynamic: refresh data, tweak assumptions, and drop results directly into your existing docs, decks, and models. This shifts junior-level repetitive work into automated flows, so teams can focus on judgment, client context, and risk.
- DCF builder with explicit assumption tracking and sensitivity outputs.
- Comps builder that keeps methodologies consistent across coverage.
- Due diligence packs with source links, variances, and open questions.
- Earnings recaps with KPI tables, guidance deltas, and management tone analysis.
Implications for banks, asset managers, and PE
Expect workflows that took hours to compress into minutes-model fixes, comps updates, coverage notes, board-ready slides. With AI embedded in Excel and wired to live data, the friction moves from gathering to deciding. That also means governance matters: access controls, model audits, and human-in-the-loop signoff.
Real talk: this shift will likely reduce some analyst throughput needs and lead to job cuts in specific functions. Teams that upskill the fastest will gain leverage-one analyst can now do the work of several, with better documentation and fewer manual errors.
How to pilot it responsibly
- Start with contained use cases: earnings summaries, comps refreshes, disclosure extraction.
- Keep a human in the loop for any valuation, client deliverable, or regulatory output.
- Lock data governance: PII handling, vendor entitlements, audit logs, and versioning.
- Define measurable KPIs: time saved per model, error rates, revision cycles.
- Create reusable prompts/templates so outputs are consistent across teams.
Access and rollout
Claude already integrates across Microsoft 365 for editing decks, drafting reports, and fetching files, emails, and Teams messages. The Excel add-in is in beta; broader access is expected after feedback from the initial 1,000 participants. Data connectors will matter most for research, PE, credit, and macro teams that live on live feeds.
Anthropic also announced a major expansion of its collaboration with Google Cloud, marking one of the largest compute scale-ups for AI infrastructure. More compute means faster iteration on reasoning, latency, and scale-useful if you're planning firm-wide deployment.
Next steps for your team
- Nominate a pilot squad: 2-3 analysts, 1 associate, 1 risk/compliance reviewer.
- Pick three workflows to automate end-to-end in Excel with clear acceptance criteria.
- Wire in one or two data connectors your team actually uses daily.
- Document final outputs and signoff steps so compliance is built in, not bolted on.
If you want structured training and playbooks for finance teams adopting Claude and similar tools, see our resources:
Bottom line: Claude's Excel integration, live data connectors, and task-specific skills push AI from "helpful chat" to practical automation. The firms that standardize workflows, enforce oversight, and upskill their teams will capture the most value-and do it safely.
Your membership also unlocks:
