Somewhere right now, an arts professional is on their fourth evening writing a funding application, wondering whether letting AI help would be cheating — while suspecting that other applicants stopped wondering months ago. Let's take the question seriously, because it's really three questions wearing one coat: is it allowed, does it work, and is it right?
Is it allowed?
Increasingly, funders have positions on this, and they're less prohibitive than people assume — several explicitly accept AI assistance, not least because banning it would penalise applicants who use it for access reasons: dyslexic applicants, applicants writing in a second language, small organisations with no development staff. But positions vary and change, so the boring answer is the right one: check your funder's current guidance before you start, and if they ask whether AI was used, answer honestly. What no funder anywhere welcomes is the thing we should really be discussing.
Does it work?
Here the news is genuinely useful, as long as you keep one distinction straight: an application is thinking made visible. The thinking — what the project is, who it's for, why it matters, why you — cannot be generated, because it doesn't exist anywhere until you do it. What AI can transform is the visible-making:
- Structure. Paste in the questions and your rough notes; ask for an outline of where your material answers what. The blank page problem, solved cheaply.
- First drafts from your own material. Give it your project notes, your evaluation data, your previous applications — then ask for a draft answer to question 3. It's reshaping you, not inventing for you. (This is the transformation principle again.)
- The word count problem. "Cut this from 480 words to 300 without losing the participant voice" is a task AI does in seconds and humans do through gritted teeth.
- Translation between registers. One core case for support, retold for a foundation that cares about young people and a trust that cares about heritage. The project doesn't change; the emphasis does.
- The cold read. Before submitting, paste in the finished draft and the assessment criteria and ask: "Score this as an assessor would. Where is it weakest?" Uncomfortable. Worth it.
Where it goes wrong
Two failure modes, both fatal. The first is invention: ask AI to write your application from a thin prompt and it will cheerfully supply plausible outcomes, invented statistics and a partnership with a library you've never contacted. Everything it drafts gets checked against reality, because assessors do check, and fluent wrongness is what these tools produce when starved of your facts. The second is the word soup: generic prose about vibrant communities and transformative outcomes. Assessors read hundreds of applications; they could smell boilerplate before AI existed, and machine boilerplate smells identical, only more so. An application that could have been written by any organisation will be funded like it was written by none. (The same convergence problem afflicts marketing copy — we wrote about it here — and the same fix applies: you originate, it tidies.)
Is it right?
Our view: using AI to reshape your own thinking is not just defensible but healthy — those four evenings were never a meaningful test of artistic merit, and the sector's grant-writing arms race taxes exactly the organisations with the least capacity. Using AI to simulate thinking you haven't done is wrong, and also — some comfort — doesn't work. The application was never the point. The project is the point. Let the machine carry the prose, and spend the recovered evenings on the thing you're asking to be funded for.
The image above was AI-generated — as ever, we'll always tell you.