New: AI Audit Assessment. Discover every AI opportunity in your business. From £999. Learn More →
Prime Ai
AI Tools
10 min read

Claude Projects for Finance Teams: A Complete Setup Guide

Most finance teams using Claude are not using it at full capability. They paste context in every conversation, repeat their cost centre structure each month, and start from scratch on every board pack. Claude Projects, Skills, and Connectors change that entirely. This is how to set them up for finance workflows, from the knowledge base to recurring task automation.

What Claude Projects Actually Changes

The standard Claude.ai conversation is stateless. Every time you open a new conversation, Claude knows nothing about your business, your reporting structure, your management pack format, or your commentary style. If you use Claude for monthly reporting, you are probably re-explaining these things every month, or copying in the same context block from somewhere else. That friction is exactly what Claude Projects removes.

A Claude Project is a persistent workspace. You load it once with your finance context — chart of accounts, cost centre descriptions, board pack templates, prior commentary examples, the writing style your CFO expects — and that context lives in the Project. Every conversation you open inside the Project starts with Claude already knowing all of that. The first month you set it up, you do the configuration work. Every month after, you just run the workflow.

Skills extend this further. A Skill is a reusable tool you build inside a Project — a structured task that runs a specific workflow when invoked. Connectors go one step further, giving your Project live access to documents in SharePoint, Google Drive, or other connected sources rather than relying on uploaded snapshots. Together, the three features turn Claude from a general-purpose AI into a configured finance tool.

How a London Finance Team Set This Up

One of our clients is a financial services business in London with a compliance-heavy finance function. Their monthly reporting cycle involves a management pack covering six entities, a board narrative covering regulatory capital, and a set of variance commentary sections written to a specific format that the board had approved three years ago and expects consistently each month.

Before Claude Projects, their Finance Director was spending three hours each month briefing Claude afresh before using it for commentary drafts. She had a notes document she would copy from, but it never quite captured the nuance of the regulatory context or the board's preferred framing. The output was inconsistent month to month.

We built a Claude Project called Monthly Management Pack. Into the project knowledge we loaded: their chart of accounts (PDF export from their system), the board pack template (Word document), two prior board narratives as style examples, a summary of the regulatory capital framework they report against, and a one-page document explaining the board's preferred framing for different types of variance. We then built Skills for each commentary section. Now the Finance Director opens the Project, runs the relevant Skill with the month's numbers, and has a first draft in under five minutes. Total setup time: one afternoon. Monthly time saving: two and a half hours.

Want to go deeper? Our AI for Finance Leaders course covers this in detail with practical templates and exercises.

Setting Up Claude Projects for Finance

You need a Claude Pro or Team plan. Team is recommended for finance use because it includes a data processing agreement and does not use your data for model training. Create a new Project from the Claude.ai dashboard and give it a specific name — one Project per major workflow works better than one Project for everything.

What to add to project knowledge

  • Chart of accounts: Export from your ERP as a PDF or CSV. Include cost centre descriptions, not just codes. Claude uses these to understand context when you paste in financial data.
  • Reporting template: Your management pack or board report structure. Claude uses this to understand what sections need to be produced and in what format.
  • Style examples: Two or three prior commentary sections that represent the tone and format the board expects. These are the most important piece of knowledge in the Project.
  • Business context: A one-page document covering your business model, key metrics, reporting period, and any recurring themes (e.g. "we are in a growth phase and the board accepts higher marketing spend as a strategic investment").
  • Project instructions: The system prompt for the whole Project. Set the role, tone, and any absolute rules (e.g. "never use passive voice", "always quantify variances in both absolute and percentage terms").

What NOT to add to project knowledge

  • Live financial data — add this per conversation, not as static knowledge (the numbers change every month)
  • Confidential individual employee data — project knowledge has the same access controls as your Claude account
  • Very large documents (100+ pages) — context length has limits; summarise long policy documents rather than uploading them in full

Building Finance Skills

Skills are the feature most finance teams do not discover until they have been using Projects for a month or two. They are worth setting up from the start. A Skill is a saved prompt workflow — a structured task that appears as a tool inside your Project conversations. You invoke it by name and it runs the defined workflow, often prompting you to fill in specific inputs.

For finance workflows, we build Skills for every recurring task in the monthly cycle. The setup is a structured prompt that uses the RACEF framework — Role, Action, Context, Examples, Format — with input fields for the dynamic data that changes each month.

Variance Commentary Skill

Runs the RACEF commentary template for any variance line. You input: line item name, actual, budget, variance, known drivers. Claude produces the commentary section using your project knowledge for context and style.

See our RACEF variance commentary prompts guide for the exact template structure to use when building this Skill.

Reconciliation Summary Skill

Takes reconciliation output (unmatched items, total matched, exceptions) and produces a formatted reconciliation summary section for the management pack. Works for bank reconciliation, intercompany, and sub-ledger tie-outs.

Board Narrative Skill

Takes month-end summary numbers and produces the opening board narrative section: headline performance, key variances, outlook statement. Uses the style examples in project knowledge to maintain consistent tone.

Journal Narrative Skill

For ERP journals requiring a narrative description. You provide: journal type, amount, period, and reason. Claude produces a structured journal narrative meeting your ERP's description format and your audit requirements.

Claude Connectors: Live Data Access

Connectors are available on Claude Max and Team plans. They give your Project live access to external data sources rather than relying on documents you manually upload. For finance teams, the most useful Connectors are SharePoint (for your reporting templates and prior packs) and Google Drive (for teams on Google Workspace).

The practical advantage: when your board pack template changes, your Claude Project automatically reads the updated version without you having to re-upload it. When a new prior period board pack is added to SharePoint, Claude can reference it in the next month's workflow without manual intervention. This is particularly valuable for teams where the reporting template is maintained centrally and updated periodically.

Setting up a Connector: in your Project settings, select 'Add Connector' and authenticate with the relevant service. You can restrict which folders or files the Connector can access — for finance, we recommend restricting to specific SharePoint document libraries (e.g. Finance Reports, Board Packs) rather than giving access to the entire SharePoint.

  • SharePoint Connector: Access reporting templates, prior pack archive, policy documents. Best for M365 organisations.
  • Google Drive Connector: Access shared finance folders. Best for Google Workspace organisations.
  • Notion Connector: Useful if your finance team uses Notion for process documentation or assumptions tracking.
  • GitHub Connector: Relevant if your team has structured financial data or automation scripts in GitHub repositories.

A Complete Month-End Claude Project

Putting it together: here is what a fully configured month-end Claude Project looks like for a mid-market finance team.

  1. Project name: Month-End Reporting — [Year]
  2. Project instructions: Role (Senior Finance Business Partner), tone (direct, professional, no jargon), rules (quantify all variances, flag any data inconsistency, never use passive voice), and format preferences (paragraph length, whether to include outlook sentences)
  3. Knowledge — permanent: Chart of accounts, cost centre descriptions, business context document, two prior board pack sections as style references
  4. Knowledge — updated monthly: Prior period actuals summary (one page), any strategic context relevant to the current month (e.g. budget revision, one-off items)
  5. Connector: SharePoint Finance Reports library (read-only access to current board pack template and prior pack archive)
  6. Skills: Variance Commentary, Reconciliation Summary, Board Narrative, Journal Narrative
  7. Month-end workflow: Export actuals from ERP, paste into a new conversation in the Project, run each Skill in sequence, review and edit output, paste into pack template

The one-time setup takes a half-day. The monthly workflow after setup takes 30-90 minutes depending on pack complexity, down from a typical 3-5 hours for manual drafting. For the consulting and implementation support to build this, see our AI consulting service or our finance function audit which identifies exactly which workflows to prioritise first.

If you are comparing Claude against ChatGPT or Copilot for this setup, see our tool comparison guide for the decision framework. If you are migrating from a ChatGPT workflow, see our ChatGPT to Claude migration guide. For the specific RACEF prompts to use in your Skills, see our variance commentary prompts guide. For the AI for Finance Leaders training that covers Claude Projects in a finance context, see our AI for Finance Leaders course.

Work with Prime AI Solutions

Self-Paced

AI for Finance Leaders

8 modules covering FP&A, reporting, automation, and governance. No coding required. From £99.

View course
Team Training

Live Workshop for Your Team

Half-day or full-day live sessions for finance teams. Tailored to your tools, workflows, and industry.

See workshops
Audit

AI Finance Audit

We map your finance function, identify what to automate, and give you a prioritised action plan.

Learn more
Done For You

AI Consulting

We design and build the workflow, configure the tools, and train your team. 90-day typical engagement.

Learn more

Frequently Asked Questions

Get AI insights for business leaders
Subscribe Free

Related Resources

Blog
Variance Commentary Prompts: RACEF Templates

The RACEF prompt templates to use when building your Claude Skills for variance commentary.

Learn More
Blog
ChatGPT vs Copilot vs Claude for Finance

Tool comparison to help you decide whether Claude is the right fit for your finance team.

Learn More
Training
AI for Finance Leaders Course

Hands-on training covering Claude Projects setup and RACEF prompting in a finance context.

Learn More
Set Up Claude Projects for Your Finance Team

We configure Claude Projects, Skills, and Connectors for your specific finance workflows. Most setups are complete in a half-day session.