Everything your board will ask you
The five answers on the home page, in full. If your board asks something that is not here, ask it and it gets added.
What it is
It reads your whole field: the grants open now, the donors worth going back to, and the companies and charities already showing interest. It reaches a view on which are worth your name and why. It writes the case in your own words, with the source attached to every claim in it. Then it brings you the argument and waits.
- It never sends anything.
- It never agrees to anything.
- It never commits you to anything.
- It never writes to a funder.
Everything leaves in your name, after you have said yes.
It has judgement and it uses it. It reaches a view, argues for it, and shows you what that view rests on. What it never has is the last word. Every decision that binds your charity, to a funder, to a donor, to a partner, is taken by your people. The machine argues. Your people decide. That order does not invert.
What it asks of you
Approve what goes out, and tell it what came back. The second part is real work and we will not pretend otherwise: the loop sharpens because you feed the outcomes back into it. Everything else is carried for you.
No. There is nothing for you to operate.
Your data
Us, and the companies named on our sub-processor list, which says what each one sees, where it is, what binds it, and the date each row was last confirmed. Not another charity, and not a competitor of yours.
While we work for you we hold what the work needs: your documents, and the research built around your field. Where it sits is on the list above. When the contract ends, it is deleted. The only exception is the business records Spanish law requires us to keep for six years, and those go the day that period runs out.
Yes, and structurally, not by anyone remembering to be careful. Nothing from your work is visible inside anyone else's, and we learn across all of our work, so what works for one charity improves the work for every charity. The personal information of a charity or of its donors never crosses to another. Work for you goes out through your own accounts, which you control and can withdraw.
Yes. Where something is too sensitive to go near a general AI provider, it does not go near one: that work runs on a local model instead, on a machine we control. You tell us what falls into that category and it stays out.
Your funders
That is the funder's call, and every one of them is below in their own words. Where a funder asks you to declare it, you will be able to declare it accurately, because every claim in a draft carries the source it came from.
Fifteen of the funders UK charities apply to, and what each one says.
Permits it, explicitly. The National Lottery Community Fund and The National Lottery Heritage Fund both publish the same sentence: "We will not reject an application just because AI was used." Paul Hamlyn Foundation: "Using AI tools alone will not disadvantage your application." Lloyds Bank Foundation for England and Wales: "We understand if you want to use AI to support you." Esmée Fairbairn Foundation has no preference either way and asks only that the application stays "an honest reflection of your work and plans". The Henry Smith Charity welcomes it for structure, and asks you to "write the words yourself".
Asks you to declare it. City Bridge Foundation: "Be transparent", and "tell us what tools you used and why".
Limits how far it can go. UKRI: "Applicants must not use generative AI tools to generate an entire application." The Wellcome Trust, in a joint statement with the Research Funders Policy Group, restricts using it "to generate an entire application, or sections of an application".
Publishes nothing either way. Arts Council England, Sport England, Comic Relief, Garfield Weston Foundation, Joseph Rowntree Foundation and Trust for London. Four of the six say nothing on any page we could reach. Two we could not read at all, and we would rather say that than record a silence we did not verify.
Not one of the fifteen prohibits it.
Where it comes from
Yes, all of it. Every claim in anything prepared for you carries the source it came from, so you read the source rather than take our word. The funders' own pages above are public and we link to them. The sub-processor list is dated, row by row. Nothing here rests on trusting us where you could see for yourself.
Most of the sector, quietly. 91% of charities already use AI in some form. Very few will say so, some because they would rather the charity down the road did not learn how they got better.
So the question in front of your board is not whether to start. It is how it is governed, which is what the rest of this page is for.
Our own clients stay confidential wherever they ask us to. If they are not ready to say it, nobody hears it from us. That holds for you too.
Source: Blackbaud, The Status of UK Fundraising 2026, 9th annual edition.
One person, and the same one at both ends. Marc de Andrés built this and answers for it himself, with a direct address he reads. He came from McKinsey and from building language models, and before that from being the person who decided, every week, which part of a charity's fundraising would not get done. That is the problem this was built against.
What it costs
You pay for the work your charity needs and nothing else, agreed with you before any of it starts. There is no subscription, and nothing renews on you.
Three months. Long enough to see it running at full scale on your real field, rather than a demonstration on a sample.
Stopping takes thirty days' notice in writing, at the end of those three months or at any point after. Those thirty days are the handover: everything disconnected safely, and every document you need put in your hands. You leave with the work.