AI Charity Assessment Template: The Framework Grant-Makers Actually Use
Every grant-maker eventually discovers that a generic checklist is not an AI charity assessment template. A checklist tells you what to look at. A template tells you what matters, how much it matters, what good evidence looks like, and when to stop looking. The difference is the difference between a list of ingredients and a recipe.
As AI-assisted due diligence becomes more common, the gap between funders with a well-structured template and those without one is widening. AI tools are only as good as the instructions they are given. If the criteria are vague, the scoring is arbitrary, and the evidence types are undefined, then AI just generates vague, arbitrary output faster than you could have produced it yourself. A proper template is what turns raw AI capability into a repeatable, defensible process.
This article sets out what a working AI charity assessment template should contain, how each part functions, and where generic approaches tend to break down.
Why generic templates fail in practice
There are dozens of freely available charity assessment frameworks online. Most of them share the same structural problem: they were designed to describe what charity assessment involves, not to drive a consistent process. They list categories — governance, financial health, impact — without specifying how to weight them, what evidence counts, or how to resolve disagreements between assessors.
When you hand a generic template to an AI tool, the same problem compounds. The AI cannot make a meaningful distinction between "strong evidence of impact" and "some evidence of impact" unless you have defined what each phrase means in terms of document types, source quality, and specificity. Without that, it defaults to surface-level pattern-matching: presence of an annual report scores higher than absence of one, regardless of what the report actually contains.
A bespoke rubric solves this. It specifies, for each criterion, what evidence you accept, how you grade it, and what weight it carries in the overall score. That is what lets AI do useful work.
The five components of a working AI charity assessment template
1. Criteria and their definitions. Each criterion in the template should have a plain-language definition that an AI can apply without interpretation. "Financial sustainability" is not a definition. "The charity holds at least three months of unrestricted reserves and has not drawn on them significantly in the past two years" is a definition. The more specific the definition, the more reliable the AI scoring.
Most working templates cover five to eight criteria. Common ones include: financial health and reserves, governance and trustee independence, impact evidence quality, leadership and organisational stability, geographic or thematic fit, and safeguarding policy. The right set depends on what your foundation actually funds and how you prioritise risk.
2. Evidence types accepted for each criterion. For each criterion, the template should specify which source types count as evidence and how they rank. For financial health, the hierarchy typically runs: filed accounts (highest weight), Charity Commission financial summary (medium weight), charity-provided financial statements (lower weight, self-reported). For impact evidence, it might run: independently evaluated reports, published longitudinal studies, internally produced outcome reports, output counts. An AI cannot apply this hierarchy unless it is written down.
3. Scoring scale and what each level means. A five-point scale is standard. What is not standard — and what generic templates almost always omit — is a behavioural anchor for each score level. "3 — meets expectations" tells an assessor nothing. "3 — the charity reports outcomes with a clear beneficiary count but provides no independent verification and does not describe how the measurement was conducted" tells an AI assessor exactly what to look for. Anchoring the scale removes subjectivity from what should be an objective step.
4. Criterion weightings. Not all criteria are equally important to every funder. A foundation focused on early-stage charities may weight governance more heavily than reserves, because small organisations are expected to be financially lean. A trust making multi-year commitments to established organisations may weight financial resilience above all else. Whatever the priorities are, they should be expressed as explicit weightings in the template rather than applied informally at the end of the process.
5. Red flag rules. Some findings should disqualify a charity regardless of its overall score. Known red flags include: regulatory action by the Charity Commission in the past three years, reserves below one month of operating costs, absence of a current safeguarding policy for organisations working with vulnerable beneficiaries, and unexplained related-party transactions in the accounts. A good AI charity assessment template encodes these as automatic flags rather than leaving them to the discretion of individual assessors.
Structuring the template for AI use
When an AI tool works through your template, it needs to be able to find evidence, match it to criteria, and apply the scoring scale without ambiguity. The practical implication is that the template should be structured as a series of explicit questions, each with a defined evidence scope and a scoring guide.
For example, a criterion on governance might produce the following prompt structure for the AI: "Review the charity's Charity Commission filing, trustee list, and most recent trustees' report. Does the board have at least five independent trustees? Have conflicts of interest been declared and managed appropriately? Has the organisation been the subject of any Charity Commission regulatory action in the past three years? Score according to the following anchors..." followed by the defined scale.
This approach — sometimes called a scoring rubric — is what separates an AI charity assessment template that produces consistent, audit-ready results from one that produces inconsistent narrative summaries. The rubric is not just documentation. It is the operating logic the AI runs.
What to build before you deploy
The work of building a good AI charity assessment template happens before any AI is involved. It requires the people who make funding decisions to agree, explicitly, on what they care about and how much. That conversation is often harder than the technical implementation. Trustees sometimes disagree about whether a charity's financial resilience should outweigh its impact evidence. Sometimes there are historic commitments to particular sectors that are not written down anywhere. All of that needs to surface and be resolved before the template can do its job.
Once criteria, weights, evidence types, scoring anchors, and red flag rules are in place, deploying AI against the template is straightforward. The AI gathers evidence from public sources — the Charity Commission register, filed accounts, web presence — maps it to each criterion, applies the scoring anchors, flags any red flags, and returns a structured output ready for human review. The template does not replace trustee judgement. It ensures that judgement is applied to the right question at the right time.
Keeping the template current
A charity assessment template should be reviewed annually, at minimum. The evidence landscape changes: Charity Commission filing requirements have evolved, SORP guidance has been updated, new data sources have become available. More importantly, what your foundation has learned from previous assessments — which criteria predicted problems, which ones did not, where the red flag rules caught something important — should feed back into the template over time.
Templates that are never revised drift out of alignment with practice. The rubric says one thing; the trustees do another. Eventually the template becomes decoration rather than infrastructure. The version history of your template is itself part of your grant-making audit trail.
Getting started
If you are building an AI charity assessment template from scratch, start with your most recent contested grant decision. What were the criteria that trustees could not agree on? Where was the evidence insufficient? What would have resolved the disagreement earlier? Those questions will surface the criteria that belong in your template faster than any generic framework.
ClearGiving implements exactly this model: a customisable scoring rubric that you define once, AI-assisted evidence gathering from the Charity Commission and public sources, and structured output for each assessed charity ready for trustee review. You can read more about how the methodology works, or explore the charity due diligence guides for more on building a consistent assessment process.