Most organisations answer the same grant questions from scratch, every time

What is your organisation? Who do you support? What difference does your work make? Why now? These questions appear in almost every grant application. And most small charities, CICs and community organisations answer them again from memory each time, in slightly different ways, under time pressure.

AI tools like ChatGPT and Claude can help. But they only produce strong, specific output when they have real knowledge to work from. Without it, the text is fluent but generic - and you end up editing heavily anyway.

This guide shows you how to build up from scratch: starting with a free tool that asks you the right questions, then moving to something you own and can reuse, and finally pointing to what a fully structured system looks like.

Start with the AI Bunny: a quick free proposal draft

1
Starting point

The Charity Excellence AI Bid Writer

The Charity Excellence Framework offers a free AI bid-writing tool - the AI Bunny - that works well as a first experiment. You register, answer a structured set of questions about your organisation and the project you need funding for, and it produces one proposal-style draft of up to 1,000 words.

The questions it asks are a useful prompt in themselves. Thinking through your answers carefully - even before touching AI - is already valuable work, and the output can give you a fast first version of the story.

When this is the right starting point: you want to try AI grant writing without any setup or cost, or you want a quick proposal-style draft to see how your project sounds before going further.

Its limitation: the AI Bunny is not a reusable grant-writing workspace. It does not hold your organisation's knowledge over time, answer individual funder questions one by one, or pause to ask targeted follow-up questions before drafting. Its input fields and word count are also limited, so you still need somewhere better to store the fuller evidence behind your bids.

Try it, see what it produces, then come back: Level 2 below shows you how to turn the same thinking into something you keep and reuse.

Access the Charity Excellence AI Bid Writer →

Build one grant knowledge document you can reuse

Level 1 gives you a single generated proposal. Level 2 gives you a reusable drafting workflow. Instead of squeezing your organisation into a short form each time, you write down the core knowledge once, keep one maintained version, and draw on it for each new application.

Importantly, Level 2 builds on the same Charity Excellence question foundation used by the AI Bunny. The difference is not the basic questions themselves. The difference is what you can do with the answers once they are stored in an AI-ready format: reuse them, expand them, update them, and apply them to individual funder questions rather than generating one fixed proposal-style draft.

The practical gain is control. You can give the AI the full list of funder questions, ask it to plan the application as a whole, then work through each answer one at a time. When the knowledge document is not enough, the AI can ask targeted questions before drafting instead of filling the gap with plausible but generic text.

2
Reusable approach

Five steps to your own reusable grant knowledge workflow

A
Choose a consistent place to use it

The most important thing is not the tool. It is having one maintained knowledge document and a repeatable way of using it. You can attach the latest version manually to each grant conversation if that works for you. A persistent workspace is recommended, called a Project in Claude and ChatGPT and a Notebook in Gemini, because it keeps the document in one place, reduces setup each time, and makes the routine easier to repeat.

For a single document like the one in Level 2, the free tiers of several major tools now work reasonably well.

Claude Free or Pro
Free · or Pro ~£18/month · Anthropic
Projects are available on the free tier, with unlimited files per project (up to 30MB each). For a single knowledge document, the free tier works. Pro adds higher message limits and access to stronger models.
Gemini Free or paid plan
Free · or AI Plus ~£4.49/month · Google
Notebooks are Gemini's closest equivalent to a Project. A Notebook holds your document, answers from it, and shows which part of your text each point came from. This is the feature that used to be called NotebookLM; it was renamed Gemini Notebook in July 2026 and now sits inside the Gemini app. The free tier holds up to 50 sources per notebook, so one knowledge document is comfortably within it. If you also want your drafting instructions saved rather than pasted each time, a Gem can sit on top with the Notebook as its source. A natural fit if your organisation already uses Google Workspace. Google AI Plus is the cheapest first paid tier of the three and doubles the free usage limits.
ChatGPT Plus recommended
Free limited · Plus ~£20/month · OpenAI
The free tier allows only 3 file uploads per day, making consistent project use more cumbersome. ChatGPT Plus is the more reliable option here, with up to 20 files per project and the most widely tested workflow for this approach.
Microsoft Copilot
Included in some M365 plans
May already be available if your organisation uses Microsoft 365. Check with your IT contact - it could mean no extra cost.

Where free tiers hit a wall is not storage. All of these will hold a single knowledge document comfortably, and a free Gemini Notebook will hold up to 50 sources. What you run into first is usage: free plans cap how many messages you can send in a day, and drafting a full application is dozens of exchanges, so a long session can stop part-way through. That is the practical reason to consider a first paid tier, rather than running out of room for files. For now, any of the above will get you started.

B
Set it up safely first

Use AI with organisational information, not sensitive personal information. Before you upload anything, do these three things.

Turn training settings off. Check the privacy controls before adding organisational material.
A first paid tier is a sensible step up. For regular use, ChatGPT Plus, Claude Pro or Google AI Plus usually give better limits and file handling than free accounts.
Keep sensitive data out. Use descriptions, outcomes and anonymised evidence, not beneficiary records or personal details.

On consumer accounts, turn off any setting that allows your chats, files or activity to be used to improve or train models. In ChatGPT, this is "Improve the model for everyone" under Data Controls. In Claude, check the "Help improve Claude" or model-training preference. In Gemini, turn off "Keep Activity" for future chats.

For many small and medium-sized organisations, a first paid tier such as ChatGPT Plus, Claude Pro or Google AI Plus is a realistic step up from a free account: better file handling, higher limits and clearer control over model-training settings. For ordinary organisational material - descriptions of your work, outcomes, policies, impact, governance and past grant wording - this is usually a proportionate starting point when those settings are turned off.

It is still not the same as a private in-house system. Even with training turned off, AI providers may retain, process or exceptionally review content for service operation, abuse prevention, safety, legal compliance, feedback handling or technical support.

Never enter:

Names of individual beneficiaries Contact details or addresses Safeguarding records Medical or health information Bank details or financial data Staff personal information

Safe to enter:

Organisation description Activities and services Outcomes and impact Aggregated statistics Anonymised quotes (role only) Published income figures Past successful grant language
C
Build your knowledge document - one section at a time

Create one master knowledge document in plain text. You can upload it to a Project, a Notebook or a Gem, or attach the latest version manually when you start a grant conversation. The template uses the Charity Excellence question set as its starting framework, then turns it into an AI-ready knowledge document: answer-led, easy for the AI to read, and designed for repeated use beyond a single form submission.

Aim for at least 50 words per answer. Use plain paragraphs and avoid heavy formatting like bullet points. Write what is true and evidenced, not what sounds impressive.

A pre-formatted template is ready to download - all four sections already structured with every question, plus the prompts and final checks that help you use it safely and consistently with your AI tool.

Download the knowledge document template

These four content areas are the sections of the knowledge document. Open each one to see what to include.

The funder probably knows nothing about you. Tell them about your charity, what you do and how. Avoid jargon or acronyms they may not understand.

What is your organisation's name? Write: The name of our charity / community group / CIC is…
What kind of organisation are you? Small charity, registered charity, volunteer-led community group? If you are not registered, make sure to say that you have a constitution and trustees or management committee.
How big are you? Something like: our annual income is… Or, if newly set up: we only recently formed and estimate that our income this year will be…
Who does your organisation support and where? Make sure the groups and location are eligible for the funder you are applying to. If you specifically support any groups that are a priority for the funder, make sure to include this.
Write a short statement describing your organisation and the services or support it provides. Tell the funder about your charity, what you do and how. If the bid is for an area or activity in which you have particular expertise, access or capabilities, include this.

You can reuse most of these answers across future bids - write them well once and adapt, not rewrite.

This is the core of any application. It is not about what you want - it is about what the funder wants and what your beneficiaries need.

Why is your charitable project or activity needed? What is the unmet need you will address? It is not about what you want, but what the funder wants and what your beneficiaries need. Use data and evidence to support this.
What makes funding your project so urgent and/or important? There is huge competition for grant funding - what makes your project particularly urgent and/or important? If there has been any recent media coverage or research reports that make this issue topical, you might reference these.
Tell me about any key dates and, if applicable, how long it will last. Include any key dates, bearing in mind that many funders meet quarterly or even annually.
Describe what your project will involve, how it will be delivered, where and/or by whom, including any partners. Funders want to see that you have a practical, deliverable plan that will address the need they fund. Give them the information needed to show that you do.
Outcomes are the difference your project will make - list these, including who will benefit, the numbers, and how big an impact this will have on them. Funders want to make a difference, so show them that your project will. How many will benefit and how big an impact will this have on their lives?
Will there be other benefits, such as impact on the wider community, helping you make the activity sustainable, or creating best practice that will be shared? Your project may have wider or long-term impact - for example, reducing anti-social behaviour in the local community, or making a change in someone's life that will impact them for the rest of their lives.
How will you know or measure that your project has been a success, and how will you report that to the funder? That might be the numbers who benefit and/or feedback surveys, or some other measure of the change you will have made. Funders like to hear about the impact they have - will you produce a report, or write to them? Will you recognise their support on your website, materials or at an event?

For project-specific questions (dates, partners, particular outcomes), write a version for your current project. For the need and urgency, write something that can be reused and updated over time.

Funders want to know what they will be paying for and that your budget is reasonable. Be clear and proportionate.

What amount of funding are you requesting? If it is only some of the funding you need, write "a contribution of…" - for example: We are seeking grant funding of £3,000, or a contribution of £500 towards the total cost of £3,000. If you have already secured some income, mention that to show others think your project is worth funding.
Provide a simple breakdown of what you will spend this funding on and the amount for each item. Include the main headings and amounts for each budget line. Funders want to know what they will be paying for and that your budget is reasonable.

Never enter bank details, account numbers or payment information in any AI tool. Budget totals and narrative descriptions are safe.

These are the details that move a competent application to a compelling one. Build this section up over time as you gather evidence and feedback.

Is there anything else that will make your funding bid really compelling? Have you won an award, achieved accreditation, work to a quality framework or hold a quality mark that demonstrates the quality of your work? What about facts or data on previous activities or years that shows how effective you are? Or perhaps that demand for your services has increased significantly, or that you have had really positive feedback.
Do you have any quotes that show the impact you have on people's lives? Including a relevant quote can help to engage the funder and demonstrate the impact you have - attribute by name or "anonymous." People give to people. For example: a parent who attended last year's sessions said "It was absolutely brilliant and I made lots of new friends."

For quotes, use role titles or "anonymous" - never include full names or any identifying details about the people you support.

D
Three steps to use every session

The AI does not always automatically read your knowledge document when you open a new conversation, even inside a Project or Notebook. Follow these three steps in order at the start of every drafting session.

First - load your document into context

Paste this when you open a new conversation "Please read [name of your file] before we begin. This is my organisation's grant writing knowledge document - it contains the context I want you to use. If the file is in this Project or Notebook, read it from there. If I have attached it to this conversation, read the attached version. Let me know when you have read it and I will tell you what we are working on today."

Replace [name of your file] with the filename you used. If you downloaded the template here, it will be: Grant Knowledge Document Template.txt

Once the AI confirms it has read the document, keep the knowledge document as it is. Do not paste the application questions into the file. The questions are provided separately in the next message.

Then - share all the application questions at once

Paste this as a new message, followed by the funder's full question list "Here is the full list of questions from the application form. Please read them as a set so you can see the complete shape of the application, but do not start answering them yet. Just confirm when you have understood the full list and wait for my next instruction.

[Paste all application questions here]"

Finally - set the rules for how it works through them

Paste this after sharing the questions "Now that you can see all the questions, please read through them as a set before we begin. Plan how to distribute information across answers without repeating yourself. Then work through each question one at a time: before drafting each answer, ask me any targeted questions if the knowledge document does not give you enough to write a specific, evidenced response. Do not move to the next question until we have agreed on the current answer. I would rather go back and forth than receive a full set of vague drafts."

This produces a back-and-forth rather than a single dump of all answers. Each answer gets a targeted exchange - the AI surfaces what it needs, you supply it, and the draft is built on real information rather than plausible filler.

E
Expect iteration - and make it compound

First drafts will still need work. The AI will identify gaps; you supply them; the draft improves. When you add new information to answer the AI's questions, add it back to your knowledge document too. Each application, your document gets richer. Each subsequent draft, less iteration is needed.

A genuine step up from Level 1

Level 2 is a meaningful improvement over Level 1 because it changes the shape of the work. You are no longer asking a tool to turn limited form answers into one long proposal. You have one maintained source of grant knowledge, a repeatable way to use it, and a process that lets the AI draft individual answers, ask for missing detail, and improve over time. For organisations with the confidence to maintain that routine, this is already a significant shift in how grant work feels.

  • The Charity Excellence questions cover roughly 10% of the knowledge a fully structured Funding Knowledge Base holds. That 10% is genuinely useful, but you will feel the gap the moment a funder asks about your Theory of Change, your safeguarding approach or your governance arrangements. Those questions were never in the set, so the knowledge is not in your document, and the AI has nothing to draw on.
  • Repeating yourself remains part of the game. For each new application you still need to guide the AI toward the right angle for that funder, adapt the language, and review the draft. The document reduces repetition but does not eliminate it.
  • Maintaining one master document and using it consistently takes discipline. A Project, Notebook or Gem can make that easier, but it does not replace the need to keep the document current.
  • There is no funder intelligence built in. The AI does not know what one funder prioritises versus another. You guide it each time.

If any of this feels like more of a lift than your team can manage right now, that is not a failure of the approach. It is a reasonable point at which to get some guidance. A short conversation can help you decide the right next step →

From a flat document to a complete system

The difference between Level 2 and Level 3 is not about which tool you use to store things. Whether that is a Project folder, a Gemini Notebook, Notion, or something else is a practical choice, not the point. Storage has stopped being the constraint: free tools will now hold dozens of files quite happily. The difference is the data model: the architecture of what you store, how it is structured, and what an orchestration layer can do with it.

Building a Funding Knowledge Base is a guided process that takes time, effort and expertise. It is not a solo afternoon task. The good news is that most of the raw material already exists inside your organisation: past applications, monitoring reports, evaluation data, impact case studies, funder correspondence. The build is not about creating content from scratch - it is about structuring what you already have into a form the AI can use precisely and repeatedly.

1 From memory
📝
KnowledgeWhat you remember + documents you can find in time
🔄
Per applicationGather from scratch
AI asks for moreNearly always - very little to draw on
Generic output
2 Flat document
📄
KnowledgeOne document with four CE sections
🔄
Per applicationLoad, prompt carefully, iterate
AI asks for moreOften - gaps remain, funder intel missing
Better, needs editing
3 Knowledge Base
🏗️
KnowledgeStructured modules + orchestration layer
Per applicationLoad modules, orchestrate, draft
AI asks for moreRarely - it is already there by design
Application-ready
3
Grant Resource Studio™

What a Funding Knowledge Base is - and why it changes the output

This is what Grant Resource Studio™ builds and maintains with organisations as a guided engagement. A Funding Knowledge Base is not a bigger document - it is a structured architecture of distinct, purpose-built modules, each with a specific job, maintained separately, and usable independently or in combination. Open any section below to go deeper.

Organisational IntelligenceYour mission, Theory of Change, delivery model, outcomes logic, governance, partnerships, policies, systems, working methods and organisational voice, structured as reusable source material.
Case for SupportYour funder-ready narrative, covering evidence of need, activities, beneficiaries, outcomes and impact, measurement and evaluation, track record and added value.
Grant Application Reference InformationPast application responses, successful wording, reusable examples, and requirements from previous funders and application forms, so future bids can build on what has already been submitted.

These modules are held together by an orchestration layer - a set of rules that tells the AI which modules to load for which funder and application type, how to weight them, what the funder prioritises, and how to adapt your voice and evidence to their specific programme. It also gives the AI two operational modes: a fast first-draft mode for lower-stakes applications, and a deeper mode for strategic bids where it reviews modules carefully and asks clarifying questions before drafting.

Why the "ask me questions first" prompt becomes largely unnecessary at Level 3. In Level 2, that instruction is essential because gaps are common in a flat document. In Level 3, those gaps are filled by design. The orchestration layer already knows what a specific application needs and which modules provide it. When the AI does ask for more, it signals the knowledge base needs updating - not that you need to provide more context in the moment.

The Theory of Change and Case for Support are the two components that make the biggest difference to output quality, and the two that take real time, expertise and iteration to develop well. A flat document typically lacks both, or contains thin versions. Getting them right - with the right evidence, the right logic and the right voice - is most of the work of a guided build. Historical applications and funder intelligence then layer on top, turning the system from a strong knowledge base into one that learns and compounds with every application submitted.

The guided build draws on everything the organisation already has: past bids that worked, monitoring data sitting in spreadsheets, evaluations written for funders, strong passages from old applications. None of that is wasted. It gets structured, approved and made reusable rather than left scattered. That is what makes the process less painful than it might sound - and what distinguishes it from starting from scratch.

A guided Funding Knowledge Base build is a one-off, fixed-price engagement. There is no subscription and no day rate. The price is set to be accessible and realistic for small and medium-sized charities, CICs and community organisations: many fund it from unrestricted or core grant money, as a single piece of work rather than an ongoing cost. And the knowledge base it produces is yours to keep, works with any AI tool, and becomes more valuable with every application cycle.

There are two ways the build can happen. Most organisations commission it as a defined, one-off engagement. But if your organisation is bringing in broader fundraising support, the Funding Knowledge Base can be built as part of that work instead: same methodology, same result, delivered within an engagement shaped around your wider fundraising needs. Either way, what you end up with is the same: a structured system your organisation owns.

If you have been working at Level 2 and starting to feel its limits - the repetition that remains, the gaps the AI keeps surfacing, the sense that your knowledge is there but not quite organised enough - a complete Funding Knowledge Base is the natural next step. The raw material is already in your organisation, and nothing you have already written at Level 2 is wasted.

See the output difference concretely - the same funder question answered from memory, pieced together by hand, and from a structured Funding Knowledge Base:

See the difference for yourself →

In the 18 months after their Funding Knowledge Base went live, Early Years Cocoon - a CIC in Barking and Dagenham - secured 52% of their all-time grant income, with application preparation time down from 10 to 20 hours to 3 to 6:

Read the Early Years Cocoon case study →


Want help building something more robust?

If you would like to find out what a full Grant Readiness engagement involves - or just to talk through where your organisation is with its grant knowledge - book a free discovery call.

A free resource from Grant Resource Studio™ for UK charities, CICs and community organisations.