Digital · Practical guide
AI for small charities: choose a useful first task
Try a bounded task, check the result and keep responsibility with the people doing the work.
AI can be useful for a first draft or a different way of organising information. A good first experiment has a clear task, suitable material and a person who can judge whether the result helps.

Choose a task with a checkable result
Start with something small: improving the structure of a public event notice, suggesting headings for a briefing or turning a list of non-sensitive notes into a draft checklist. Decide what a useful result would look like before using the tool.
Avoid beginning with a decision that determines someone’s rights, safety or access to support. Choose work where a colleague can compare the output with the original material and correct it before anyone relies on it.
Use suitable information
NCVO identifies potential uses in communications, fundraising and governance, alongside risks relating to data protection, cybersecurity and misleading information. Its guidance is a useful starting point for discussing both benefits and limits.
For an initial experiment, use public or fictional information. Do not paste beneficiary stories, confidential minutes, staff records or donor details into a tool simply because it can accept them. Ask the person responsible for information governance to approve any proposed use of organisational records and check the provider arrangements.
Give a clear brief
Describe the audience, the job and the source material. For example: “Using this public event information, draft a short notice for local residents. Keep the date, location and booking instructions exactly as supplied. Identify any missing information.”
Ask for a draft rather than an authoritative answer. If the tool supplies a claim, quotation or reference, check the original source. A confident-looking sentence does not establish that the information is correct.
Review before sharing
Read the result against your brief. Check names, dates, facts and tone. Remove unsupported statements and ask whether the wording reflects what your organisation actually does. The person approving the final work remains responsible for it.
- Does it answer the task you set?
- Can each factual claim be checked?
- Has any important meaning been lost?
- Is the language appropriate for the audience?
- Who will approve the final version?
Decide whether the experiment was worthwhile
Compare the whole job, including briefing and checking, with your usual method. If the review takes longer than writing the notice yourself, choose a different task or stop using the tool for that purpose. Record what helped and what needed correction.
Use the result to agree a small working practice: suitable tasks, permitted information, review responsibility and a point for reconsidering it. You do not need to commit to a large digital project to learn whether one carefully chosen use is helpful.
Sources and further reading
Take the next step
Explore the consultancy scope or prepare an enquiry. Availability, scope, timescale and fees would be agreed before any paid work begins.