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 don't need everything neatly prepared: copy and paste from documents you already have, or just type rough notes directly into the chat.

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

• Who you support: their circumstances, needs, and any barriers they face
• What services you currently offer, and what's working well or less well
• What's changed recently: shifts in demand, new challenges, or gaps you've noticed
• What your capacity looks like: staff, volunteers, budget, and physical space

 Additional context to add depth and detail

• Feedback from the people you support: surveys, complaints, compliments, informal comments
• Any previous service reviews, strategies, or plans
• What similar organisations are doing, if you know of approaches you've admired
• Partnerships or resources you already have access to

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 structured list of ideas, each with a brief explanation. The AI may also ask you clarifying questions, for example about the size of your team, the geography you cover, or what a particular service involves. This is normal and helps improve the output.

As you read through the ideas, ask yourself:

• Do these feel grounded in the reality of our work, or are some too generic or unrealistic for our context?
• Are the people we support described and considered in a way they'd be comfortable with?
• Do the ideas reflect the full diversity of our community, or do some groups get left out?
• Are there ideas here that we've already tried? If so, what does the AI not know about why they didn't work?
• What would our team say about these: would staff and volunteers see them as feasible?
• Which ideas could we explore quickly and cheaply, and which would need more investment?

Not every idea will be right for you, and that's fine. The goal is to have something concrete to react to, discuss, and build from. If something doesn't feel quite right, you can refine it in the next step.

Getting useful ideas often takes more than one exchange. Think of this as a conversation: you can push back, zoom in, or take the ideas in a new direction.

Ask for revisions - see some examples below:

Tone:

•  "These ideas read a bit like a strategy document. Can you make them sound more like something our team would actually say to each other?"
• "Write these in a more encouraging tone, focused on what's possible rather than what's difficult."

Length:

• "Give me just the top three ideas, with a bit more detail on each."
• "Turn the top three ideas into a one-page briefing for our board."

Structure:

• "Can you reframe these ideas as questions we could bring to a team away day?"
• "Group these ideas from quick wins to longer-term projects, so we can see what's realistic first."

Representation:

•  "Check these ideas for any assumptions about what the people we support can or can't do for themselves."
•  "Rewrite idea [number] so it centres the choices and strengths of the people we support, not just their needs."

Remember: you can always start a fresh conversation with different instructions if the first output isn't heading in the right direction.

‍

Add what only you know

AI works with what you give it: it doesn't know your history, your community relationships, or the informal things that make your services work. Once you have a set of ideas worth developing, bring in:

• Insight from frontline staff and volunteers about what's really going on
• Stories or examples from the people you support
• Local context: what's available, what's changed, who you already work with
• Your organisation's values and culture: what kind of change feels right for you

Centre the people you support

Ideas that look good on paper don't always work for the people they're designed to reach. Before taking anything forward, ask: have we checked this with the community? Co-production and user research, even light-touch, will tell you far more than any AI output. Use these ideas to start a conversation with your beneficiaries, not to replace it.

Watch for ideas that assume digital access

AI tools may tend to suggest digital solutions: apps, online portals, self-service platforms. These can be valuable, but they can also exclude people who face digital barriers, including many of the people charities support most. Actively flag any digital idea and ask: how would this work for someone without a smartphone, reliable broadband, or confidence online?

Always check your facts

If the AI references sector examples, best practice, or statistics to support an idea, check them before repeating them. AI can occasionally generate plausible-sounding but inaccurate information. Treat any external references as a prompt to investigate, not as a verified source.

Get a colleague to read it through

Before sharing anything externally, ask a colleague to look it over. Do they recognise your organisation in it? Would the people you support recognise themselves? A fresh pair of eyes catches things you might miss, so never share AI-assisted content externally without a human review.

Be transparent about AI use

If you share these ideas externally, for example in board papers, strategy documents, or funding applications, consider noting that AI tools were used in the development process. More organisations are being open about this, and it's in line with Worthwhile AI values.