Setting Up Your AI Workflow

Having access to AI tools isn't the same as having an AI workflow. A workflow means AI is genuinely integrated into how you work — not something you remember to try occasionally.

This lesson covers the practical habits, structures, and setups that make AI a consistent part of your finance work day.

The Three-Layer AI Workflow

Think of your AI integration in three layers:

Layer 1 — Quick Tasks (under 5 minutes) Prompts you run immediately as tasks arise: rewriting a paragraph, explaining a term, drafting a quick email, generating a formula.

Layer 2 — Structured Tasks (5-30 minutes) Dedicated AI sessions for specific deliverables: variance commentary, board report drafts, document summaries, research briefs.

Layer 3 — Saved Templates (ongoing) Prompts you've tested and refined that you reuse for recurring work: month-end commentary, KPI narratives, investor update drafts.

Most finance professionals only ever operate at Layer 1. The real time savings come from building Layers 2 and 3.

Building Your Prompt Library

The single highest-ROI thing you can do is maintain a personal prompt library. Every time you craft a prompt that produces a great result, save it.

Where to save prompts:

• A simple Notion page or Word document titled "AI Prompts"
• A pinned note in your phone
• A text file on your desktop

What to include in each saved prompt:

• The use case (e.g., "Month-end variance commentary")
• The prompt text with [placeholders] for the parts that change
• Notes on what works well

Example saved prompt:

Use case: Monthly management report commentary
Prompt:

"Act as a senior finance manager. Write a 3-paragraph management commentary for [month] results. Key facts: Revenue [£X] vs budget [£Y] ([+/-Z%]). Gross margin [X%] vs budget [Y%]. Overhead [£X] vs budget [£Y]. Key driver of variance: [explain]. Tone: professional, direct. Audience: senior management team. Avoid jargon."

Daily Integration Habits

Morning (2-3 minutes):
Use AI to scan and summarise any long emails or reports that arrived overnight. Prompt: "Summarise this email thread. What action is required from me and by when?"

During analysis work:
Keep a browser tab with your AI tool open. When you hit a question — "what's the IFRS treatment for this?", "how should I word this?" — ask it immediately instead of interrupting your flow to search.

Before a meeting:
Paste the board pack, management report, or agenda into Claude and ask: "What are the 3 questions I should be prepared to answer in this meeting based on these numbers?"

After writing a first draft:
Paste any narrative text into AI with: "Edit this for clarity and conciseness. It's for [audience]. Keep the facts exact."

Setting Up Claude Projects or ChatGPT Custom GPTs

If you're on a paid plan, create a dedicated finance assistant:

In Claude (Projects):
Create a project called "Finance Work" and add a system prompt like:

"You are my personal finance analyst assistant. I'm a [role] at a [industry] company with [£X] annual revenue. Always be concise, professional, and flag when you're uncertain about a number. Format outputs for professional business use."

In ChatGPT (Custom GPT or My Instructions):
Go to Settings → Personalization → Custom Instructions and add your context there.

This context means you don't have to re-explain yourself every conversation.

Your Turn

This week, create your prompt library. Start with just three entries:

1. One prompt for a task you do every month
2. One prompt for a type of document you regularly write
3. One prompt for explaining financial concepts to non-finance stakeholders

Test each one, refine it, and save the version that works. That's your foundation.