← Back to Knowledge Hub

Most guides about using AI at work focus on what the tools can do. This one focuses on what you should sort out before you start — not to put you off, but because a little preparation up front avoids most of the problems that trip people up later.

The checklist is organised into three areas: data and security, compliance and governance, and practical readiness. Work through them in order. Most items take under ten minutes to address.

Section 1: Data and security

The most common mistake in early AI adoption is feeding sensitive data into tools without understanding where it goes. These checks are non-negotiable.

🔒 Must check before using any AI tool
Check whether the tool stores your inputs

Most consumer AI tools have a setting to opt out of using your conversations for model training. Find it and turn it off before you enter any financial data. In Claude, this is under Settings → Privacy. In ChatGPT, it’s under Settings → Data Controls.

Define what data you will and won’t enter

Write a simple rule for your own use: for example, “I will enter financial summaries and draft text, but not donor names, employee details, or beneficiary information.” Stick to it.

Check whether your organisation has an AI usage policy

Larger charities and organisations with public sector contracts are increasingly putting AI policies in place. Check before you start, so you’re not inadvertently breaching an existing policy.

Consider a locally-running alternative for sensitive analysis

If your role involves regular analysis of sensitive financial data, consider tools that run locally in your browser without sending data to any server. Lumino’s tools work this way — it removes the data security question entirely.

Section 2: Compliance and governance

Most small charity finance teams don’t need a formal AI governance framework. But a few quick checks will save headaches later.

📋 Governance — sort these in your first month
Review your privacy notice

If you’re using AI tools in any process that touches personal data, your privacy notice should mention this. A brief, plain-English addition is usually sufficient for small organisations.

Agree a review process for AI-generated outputs

Decide now that AI outputs will always be reviewed before use. AI makes errors. A trustee report or grant application submitted without review can contain factual mistakes that are embarrassing or damaging.

Note which outputs involved AI assistance

Keep a simple record of documents where AI tools were used in preparation. Some funders and regulators may start asking about this in due diligence processes.

Check funder terms for AI-generated content

A small number of funders are starting to include clauses about AI-generated content in grant applications. Check the terms for any active or forthcoming applications before using AI to assist with the writing.

Section 3: Practical readiness

These are about getting the most from the tools you use — avoiding the common pattern of trying AI once, finding it disappointing, and giving up before properly learning how to use it.

✓ Make these habits from day one
Start with one specific task, not a general exploration

Pick one concrete, recurring task — drafting a finance committee update, writing a grant report, building a reconciliation schedule — and focus on using AI well for that one thing before expanding.

Learn to write better prompts

Give context (who you are, what this is for), specify format (bullet points, formal prose, a table), and include constraints (UK charity, FRS 102, 250 words). Vague prompts produce vague outputs.

Build in time to review

Don’t use AI to save time and then skip the review step that makes the time saving safe. Budget ten minutes for review when you use an AI tool.

Keep a note of what works

When you find a prompt that works well, save it. These become your personal toolkit and save setup time for future tasks.

Don’t expect perfection on the first pass

AI tools work best iteratively. Give an instruction, review the output, refine with a follow-up instruction. Treat the first output as a draft, not a finished product.

Where to start

If this checklist has felt a bit overwhelming, here’s the short version. Before you start: turn off training data usage in whichever tool you’re using, and decide what data you won’t input. That covers 80% of the risk for most small charity finance roles.

Then pick one task and try it. The most universally useful starting point is asking an AI tool to turn your bullet-point notes into a paragraph of polished prose — for a finance committee update, a grant progress report, or a trustee paper. It will probably save you twenty minutes the first time you try it.

Already past the basics? See our how-to guide on using AI to prepare your year-end accounts for more advanced applications with specific example prompts.

Finance tools with no data concerns

Lumino’s tools run entirely in your browser — no data ever sent anywhere. Ideal for finance teams where data privacy matters.

Browse tools →
← Back to Knowledge Hub HMRC digital reporting →