When the AI asks if you have any drafts or notes to share, you can upload documents or paste in information to help it understand your work more deeply. You can upload documents directly, or paste text, bullet points or rough notes. Even incomplete information is helpful.

Important context (add as much of this as you can)

• Service descriptions: what each service does, who it's for, and how it's delivered
• Key numbers: how many people use each service, frequency, delivery costs if known
• Feedback and outcomes data: satisfaction scores, case notes themes, outcome monitoring results
• Any previous reviews or evaluations, even informal ones

Additional context to add depth and detail

• Referral data: where people come from, and any patterns in who you're not reaching
• Staff or volunteer feedback: what frontline people say about what's working - or not working
• Funder or commissioner requirements, especially if services are tied to specific grants
• Your current strategy or theory of change, to help assess strategic fit
• Waiting lists or unmet demand data: evidence of need that isn't being met
• Financial information: cost per person supported, funding source stability
• Your latest annual report or impact report

Before you upload files: check for sensitive data

Review your documents for personal or sensitive information about beneficiaries. Anonymise names and identifying details before uploading. Even though Worthwhile Chat is privacy-first, good data practice starts with you. If you're using other AI tools, check their data settings too — many train on your data by default. How much you share depends on your existing privacy and data policies, and how much control you have over the tools you use.

If you’re unsure, ask yourself: “If this document were seen by someone outside our organisation, could anyone be identified or harmed?” If yes, anonymise first. 

You'll receive a first analysis and the AI may ask clarifying questions - for example, what a service name means, or how to interpret a particular figure. This is useful: your answers will improve the accuracy of the review.

Read through the output and ask yourself:

• Does this reflect what we know from experience on the ground?
• Are there gaps in the evidence that need flagging to leadership?
• Has it picked up on tensions between services, or missed them?
• Are any services being assessed too positively because the data we shared is incomplete?
• Does the analysis surface anything surprising, or confirm something we suspected but hadn't named?
• Is the framing fair to all the communities we work with, including those who are harder to reach or measure?
• Would a trustee or senior leader find this useful for a real decision?

If the output feels too surface-level, that's usually a sign that more context is needed, not that the tool isn't working.

A service review is rarely finished in one pass. Use the conversation to dig deeper, challenge the analysis, or reframe for different audiences.

Ask for revisions - see some example prompts below:

Depth:

• "The analysis of [service name] feels shallow. Can you look more critically at the evidence?"
• "What questions should we be asking about [service name] that we haven't answered yet?"
• "What would we need to know to make a confident decision about this service?"

Tone and format:

• "Rewrite the summary table so it's easier to present to trustees."
•  "Produce a one-page version of this for a board paper."
• "Write a short narrative version instead of the table."

Challenge and stress-test:

•  "Play devil's advocate: what's the strongest case for stopping [service name]?"
• "Are there any services here that look good on paper but might not be reaching people who need them most?"

Representation:

• "Does this review reflect the experience of [specific group]? If not, flag where evidence is missing."
• "Are any services being assessed primarily on what's easy to measure, rather than what matters most?"

If this isn't quite hitting the mark, remember you can always start a fresh conversation with different instructions; sometimes that gets you closer than repeated refinements.

Add what only you know

AI works from the information you give it. Once you have a solid draft, you'll need to layer in:

•  Informal knowledge from staff and volunteers that isn't in any document you’ve shared with the AI tool
• Recent developments: a service that's just changed, a key person who's left, a new funder
• Your read of the politics: which services are harder to question internally, and why
• Community relationships and trust that don't show up in any data

The best service reviews combine rigorous analysis with honest conversations. This tool can help you structure and prepare for those conversations. It can't replace them.

Agree your criteria before you start

The output is only as useful as the framework you give it. Before running the prompt, take 15 minutes with a colleague to confirm: what are we actually trying to decide? Are we looking to cut costs, strengthen impact, refocus on mission, or something else? Tailoring the criteria to your real question will make the review much more actionable.

Treat this as a first draft, not a verdict

AI can help you organise and surface what you already know, but a service review has real consequences for staff, beneficiaries and partners. The output should inform your conversations, not replace them. Always discuss findings with the people closest to the work before drawing conclusions.

Watch out for data gaps masquerading as clean assessments

If your monitoring data is thin, the AI may produce an analysis that looks confident but isn't. It will flag uncertainty if you ask it to, but it's worth actively prompting: "Where is the evidence genuinely weak in this review?" 

Some of the most important things your services do - building trust, reducing isolation, being there consistently - don't show up in outcome data. Make sure your review process includes a way to capture these, whether through staff conversations, case studies, or community feedback.

Check facts and figures against your original sources

Always verify any statistics or specific claims in the output against your original documents. The AI can occasionally misread a figure or misinterpret a data point. Before sharing the review with trustees or senior leaders, cross-check the numbers.

Be transparent about AI use

If you share this review externally or use it to inform a board paper, consider noting that AI tools supported the analysis. More organisations are being open about this, and it's in line with our Worthwhile AI values. The analysis will be stronger for the human judgement you've applied, and that's worth being clear about too.