For UK charities · Funding bids & grant applications
You can, but a generic chatbot writes generic bids.
And the way most people fix that, pasting in their real monitoring data and beneficiary stories, quietly hands vulnerable people's data to a stranger. You don't have to choose between a bid that wins and one you can stand behind.
Here's why public AI fails a funding bid twice over, and how charities draft winning bids on private AI that's grounded in their own evidence and never lets it leave.
It usually starts as something reasonable. A deadline is close, the team is small, and bid-writing is gruelling. So someone opens a public AI tool like ChatGPT, Gemini or Claude and asks it to draft the application. Nothing about the intent is reckless. The problem is that it fails you in two directions at once.
A public chatbot doesn't know your charity: your outcomes, your beneficiary numbers, your theory of change, your past bids. So it writes plausible boilerplate that an assessor reading their hundredth application spots instantly. It doesn't touch the hard part, which is marshalling your evidence to this funder's criteria.
To make it specific, you paste in the real material: monitoring reports, beneficiary case studies, last year's accounts. But case studies about service users, often involving children, health, abuse or immigration status, are personal and frequently special category data. The moment they're pasted in, they land on a US company's servers with no lawful basis.
An unpublished bid is confidential, and funders increasingly ask applicants to declare how they've used AI, with some restricting it. A bid that reads as machine-written, or one where you can't account for where the data in it went, quietly undermines trust with the very funder you're trying to win over.
If a funder asked how you'd drafted your bid, and what had happened to the beneficiary stories in it, would your charity have a clean answer?
None of this means writing every bid from a blank page by hand. The drafting help is genuinely worth having. The job is to get it grounded in your evidence without handing that evidence away.
Imagine an assistant that has already read your past winning bids, your impact reports and your monitoring data, and drafts answers to a specific funder's questions from them, citing where each claim came from, with none of it ever leaving your own systems. Three things change.
It drafts from your outcomes, your beneficiary numbers, your previous bids, so the application sounds unmistakably like your charity and holds up to scrutiny, rather than reading as filler that could belong to anyone.
Reframe the same project for each funder's questions and outcomes language. The slog of re-shaping one piece of work for ten different applications, done in minutes instead of evenings.
Every claim cites the document it came from, so you can check it. It runs in infrastructure you control, in the UK, and your beneficiaries' stories never leave the building.
That's private AI for funding bids: drafts grounded in your own evidence, tailored to the funder, on data that never leaves your control.
The build is a private AI assistant, a chat tool that looks and works like ChatGPT but runs inside your own systems, that has read your charity's own documents, your past bids, impact reports, monitoring spreadsheets and theory of change, and keeps them in a private knowledge base inside your own tenancy. You give it a funder's question; it retrieves the relevant evidence from your documents and drafts an answer grounded in them, citing which document each claim came from so you can verify every line. It drafts from what you actually do; it doesn't invent, and it doesn't send your evidence anywhere.
Same blank-page dread you'd have opened ChatGPT to escape. None of your evidence leaving the building, and a draft that actually sounds like you.
I've built a sample drafting assistant against a fictional charity's documents here, to demonstrate the kinds of thing a private AI build can do for you. Below are a few everyday situations a charity might bring to it. Each one starts with what you're actually trying to do, then the question you'd type, then a short recording of the assistant drafting and citing the documents it drew from.
A funder asks you to evidence the need for your project, and you want it backed by real local figures rather than generalities.
You'd open the assistant and ask:
"Draft the 'why this project is needed' section using our area's deprivation stats and our waiting-list numbers."
This funder wants outcomes written in their own format, and you don't want to hand-rework yours for the fourth application this month.
You'd open the assistant and ask:
"Rewrite our three project outcomes as SMART outcomes in the National Lottery Community Fund's format."
The application has a track-record section, and you need to evidence your impact from what you actually delivered last year.
You'd open the assistant and ask:
"Pull our strongest results from last year's monitoring report to evidence our track record."
You won funding for this work before, and a new funder is asking similar questions, so you want to adapt what already worked.
You'd open the assistant and ask:
"We won a grant for this work two years ago. Adapt that case for support to answer this new funder's questions."
Want to see it pointed at your own bids and impact reports, or have a funder question of your own in mind? Tell me what you're working on →
You can, but the public version fails you twice. It doesn't know your charity, so the draft is generic and assessors increasingly spot it; and to make it specific you'd paste in monitoring data and beneficiary stories, which sends personal data to a US company with no lawful basis. A private build that drafts from your own evidence solves both at once.
No. Case studies about service users, especially involving children, health, abuse or immigration status, are personal and often special category data. Pasting them into public ChatGPT sends them to OpenAI in the United States, into logs you can't see. You'd need a lawful basis, a contract with the provider and usually a data protection impact assessment; a paste into a public tool has none of those.
Policies vary and are changing quickly. Some funders now ask you to declare how you've used AI, and a few restrict it. The safer position is a build you can fully account for: drafts grounded in your own evidence, with every claim traceable to a source document, and no supporter or beneficiary data sent outside your charity in the first place.
Assessors read hundreds of applications and increasingly recognise generic AI prose: vague outcomes, no specific local evidence, claims that could apply to any charity. The answer isn't to hide that you used AI; it's to ground the draft in your real numbers and stories so it reads as unmistakably yours.
Only if it can see your charity's evidence. A generic chatbot can't; a private build that has read your past bids, impact reports and monitoring data can draft from them and cite where each claim came from. That's the difference between boilerplate and an application grounded in what you genuinely do.
It shouldn't, and that's the whole point of grounding it in your documents. It draws figures from your own monitoring and impact reports and cites the source for each, so you can check every number before it goes anywhere near a bid, rather than trusting a plausible-sounding claim a public chatbot invented.
Got a question that isn't here? Ask me directly →
Tell me who you are and what your organisation does. If any of this sounds like your situation, that's a good place to start. I'll let you know honestly whether I can help. Even a 30 to 45 minute call often leaves people with a clearer picture of the path forward, whether or not we end up working together. From there it's whatever fits: sometimes you don't need me, sometimes a short piece of scoping work makes sense first, and sometimes you already know what you want and we get straight to the build. There's no set process you have to follow.
For context: I work mainly with UK charities and non profits, with chief executives, operations and finance directors, programme leads, and the people who look after data and IT. Respectfully, I don't work with recruitment or development agencies.
Email: peter@peterbrady.co.uk
PNB Technologies Limited
55 Yew Tree Road, Ormskirk
Lancashire, L39 1NT, United Kingdom
Company Reg: 07166600
VAT Reg: GB986726850
© 2026 Peter Brady. All rights reserved.