I want to write this one in first person.
The product website says CharityIQ is the grounded AI for UK charities. That’s accurate. But websites don’t tell you why someone builds something. This post is the why.
The afternoon that started it
It was a Tuesday in 2024, late afternoon. I was at my kitchen table writing a grant application for the UK charity I run. The application was for a £15,000 community grant. The funder’s eligibility criteria were specific. I’d already spent 90 minutes finding their published priorities, reading the application form, and gathering my charity’s beneficiary data into something coherent.
I opened ChatGPT to help draft the Need section. Pasted in my charity’s context. Asked for a draft. Got one. It was well-written. It included the sentence:
“In our local area, 38% of children live in households below the poverty line.”
I didn’t know if that figure was correct. I asked ChatGPT for the source. It gave me a vague reference to “ONS data”. I asked for a link. It made one up.
I spent the next 40 minutes verifying the actual figure (it was 41%, not 38%). I rewrote the sentence with the correct figure and a real source link. I submitted the application.
The application was successful. But I sat there afterwards thinking: if I hadn’t checked, I would have submitted a grant application with a fabricated statistic. Multiplied across 30,000 small UK charities that probably use ChatGPT for grant writing, how many fabricated statistics are sitting in submitted UK grant applications right now?
That afternoon was the start.
What I see small UK charities doing every day
The UK charity I run is small. About £100,000 income. Two part-time staff, three volunteers, four trustees. We do work I care about. We are also, as an organisation, perpetually overstretched.
The pattern across every charity I know at this size is the same:
- The CEO writes the grant applications, files the annual return, manages safeguarding policies, sends donor thank-yous, and runs the programmes.
- There is no fundraiser. There is no compliance officer. There is no impact analyst. The CEO is all of these.
- AI tools are tempting because they promise time back. ChatGPT is free. Charity Excellence AI Bunny is free.
- But the AI doesn’t know our funders. It doesn’t know our beneficiaries. It doesn’t remember last year’s grant application. Every time we use it, we re-paste everything.
The result: AI helps with first drafts, but the marginal value is small for charities applying to grants regularly. The promise of “AI does the boring bits” partly delivers, partly doesn’t.
I started thinking about what would actually help.
Why generic ChatGPT didn’t fix it (and why free tools won’t either)
ChatGPT is a brilliant general tool. It doesn’t know things specific to my charity, my funders, my regulatory context. That’s not a fault — it wasn’t built for that.
Charity Excellence Framework’s AI Bunny is a structured improvement on raw ChatGPT for charity bid writing. I respect the work Ian McLintock has done — Charity Excellence has been one of the most useful free UK sector resources for years. AI Bunny is a meaningful service for charities that need help with one-off bids.
But the gap I kept hitting wasn’t “I need a better one-off bid generator.” It was: I keep doing the same work over and over for different funders, and none of these tools remember what I told them last time.
Three specific gaps free tools couldn’t close:
1. Profile memory. Every conversation with ChatGPT starts fresh. AI Bunny’s 17 questions reset each session. After my fifth grant application, I was tired of re-pasting my charity’s basic information. There had to be a way to store it once.
2. UK funder integration. Generic AI doesn’t know that NLCF Awards for All caps at £20,000, or that Lloyds Bank Foundation Specialist Programme covers eight specific themes. It doesn’t know which grants are open right now versus closed. Sector-specific data lives in 360Giving and on funders’ websites — but not in any general AI.
3. Audit trail. When a trustee asked me at our last AGM how I’d prepared a particular grant application, I had no answer except “I drafted it.” If a funder had asked the same question, I’d have had nothing. As a Charity Commission-registered organisation, this should be defensible. It wasn’t.
The “grounded AI” idea — what it means in practice
I started reading about retrieval-augmented generation (RAG). The technical principle is straightforward: instead of letting an AI generate text from training data alone, you give it specific, verified data sources at the moment of generation. The output is grounded in those sources rather than in general patterns.
Applied to UK charities: the AI knows your Charity Commission registration, your past applications, your beneficiary records, your funder requirements, your impact data. Every output cites a specific source. Hallucination risk drops dramatically. Audit trail becomes possible.
I started calling this “grounded AI” because it was the cleanest way to explain the difference. Generic AI is ungrounded. Grounded AI cites its sources.
I built a prototype. It pulled my charity’s profile from the Charity Commission Register, plugged it into Anthropic’s Claude API with my data as context, and generated a draft grant application that cited every claim. The first time I ran it, I sat there for several minutes just reading the output. Every fact had a source. The Need section quoted my charity’s actual beneficiary numbers. The Outcomes section referenced my actual impact data from previous years.
It wasn’t magic. It was just AI given the right data.
What we chose to build first — and what we chose not to
I made deliberate choices about what to build first.
Built first: Grant writing. This is where small charities lose the most time. Multiple grants per year, similar information, repetitive drafting. The compound time saving of profile memory plus grounded AI is largest here.
Built next: Charity Commission compliance. Specifically annual return pre-population and SORP 2026 tier-aware Trustees’ Annual Report drafting. This was the second-biggest pain point — annual return time consumes weekends across the sector.
Built after that: Impact measurement. Beneficiary surveys, outcome data collection, master impact reports that adapt per funder. This connects to grant writing — the impact data feeds the next application.
Free Grant Finder. Lower-friction onboarding. No payment, no credit card. Set up your charity profile, get matched UK grants weekly.
Deliberately not built (yet):
- Case management. Plinth does this well. We don’t compete on it.
- Bookings and payments. Not our focus.
- Custom AI agents. Undercroft does this for £500k+ charities. We’re focused on £25k–£500k self-serve.
- Multi-language support. UK English only for now.
- Donor CRM. Many other tools do this; integration was higher priority than reinvention.
I kept the scope narrow on purpose. A small charity needs grants done well, compliance done well, and impact done well. If those three work and connect to each other, the rest can follow.
What’s next, and how to be involved
CharityIQ is at the start of its life. The product is real and works for the workflows I described. We’re learning from every charity that uses it.
If you’re a UK charity at the small end of the income range — £25,000 to £500,000, applying to multiple grants per year, filing under SORP 2026 — I’d be delighted if you tried it. The free Grant Finder is a no-cost starting point. The 14-day free trial of paid tiers is enough time to see if it fits your specific charity.
If you’re a UK funder reading this, please know: every output the platform produces cites the underlying source. We are explicitly trying to reduce hallucination risk in submitted applications, not increase it.
If you’re a fellow founder building in the UK charity space — Plinth, Undercroft, FundRobin, anyone — the sector benefits from more good tools. I respect the work each of you is doing. We don’t need to be enemies; small UK charities have plenty of pain to spread between us.
If you’re a sector journalist or commentator — Civil Society, Third Sector, Charity Digital, NCVO — I’d welcome the conversation. The “grounded AI” framing is genuinely different from generic AI and the difference matters for sector trust.
A practical note
I run a UK charity. I built CharityIQ. The two roles inform each other every day.
Every product decision is filtered through whether it would actually help the charity I run. Every product feature is something I either needed myself or watched another charity director need.
That’s not a unique founder story — many of the best products come from people building for themselves. But it does mean I take the brand line of “grounded AI for UK charities” seriously. If I built a tool that hallucinated grant statistics or gave UK charities advice that wasn’t fact-checked, I’d be making it harder for the people I’m meant to be helping.
The bar is real. The work is ongoing.
If this resonates, come see what we built.
Start a free 14-day CharityIQ trial. No credit card. Your charity, your data.
Join Waitlist →
Ivan Siyanko is the founder and CEO of CharityIQ. He runs a UK registered charity and built CharityIQ to be the AI tool small UK charities deserved. Ukraine-born, London-based.
Connect:
– LinkedIn — Ivan Siyanko
Related posts:
– The Best AI Tools for UK Charities in 2026 (Tested & Compared)
– ChatGPT for Charities: Where It Helps, Where It Hurts
– Grants for Small UK Charities: Complete 2026 Guide