AI Use Cases
What the route looks like for a charity where a funder has asked about AI, the services team holds case records and the trustees want an answer they can defend. Then twenty-one places the work could change once the rules exist.
Certified in the Claude Partner Network Microsoft Solutions Partner ISO 27001 ISO 9001 Independent since 1999
Worked example
Six steps, from the first question to AI in daily use and kept running.
A grant funder sends its annual assurance questionnaire and, for the first time, it has a section on AI. Does the charity have a policy, which tools are in use and who decides what goes into them? The chief executive knows the answer to none of the three, and the trustees will ask the same questions at the next meeting once the funder has.
That is the moment the route is built for, with a deadline on the questionnaire and a trustee meeting behind it. What follows is what happens next for a charity with a services team that supports people and holds their records, a fundraising team writing to supporters and to funders, a communications lead, a volunteer coordinator, Microsoft 365 or Google Workspace underneath it all, and a board that meets quarterly. These examples are invented, and the findings are the ones we would expect in a charity like yours.
Govern
The readiness snapshot finds personal accounts on free tiers in fundraising and communications, one used to draft an appeal letter, and a volunteer coordinator who has been pasting scheduling questions into a chatbot. None of it is misconduct. The half-day workshop produces the policy, the register of approved tools and the one-sentence rule about case records.
You hold: Three answers for the funder's questionnaire, a one-page summary for the next trustee meeting and a policy with the case-records rule as its first line.
Stays human: How wide the case-records rule reaches, and whether volunteers are in scope. Your leadership team settles both in the room and, if you want it, a trustee sits in.
Adopt
The chief executive, the head of fundraising and the communications lead connect to your Microsoft 365 or Google Workspace under a signed waiver that names them and caps the team. A company voice skill is built from how the charity writes, so a first draft sounds like you and never describes a person you support in words they did not agree.
You hold: A connection for those three, the voice skill and the result of a test on one of your supporter letters and one of your funder reports.
Stays human: Every story about a beneficiary. The skill carries your voice, and consent and the person's own words govern the story.
The review
The services team is where the most repeated writing is, and it is also where the case records live. The review checks what the team's accounts can already open, and finds two former volunteers whose accounts were never closed and a shared folder of assessments open to everyone in the charity.
You hold: Closed accounts and a locked assessments folder first. Then a connection for the services team scoped to service descriptions and published information, with the case record system outside it.
Stays human: Whether the services team is connected at all. Some charities decide not to, and the route is complete without it.
Discover
Workshops with fundraising and the grants team produce a ranked list. Replies to supporter questions rank top, because the before is measurable and a wrong draft gets caught before it leaves. First drafts of grant applications rank second, because four previous applications already describe the same project. Gift Aid claims are sent back, because the fundraising database does that with a report.
You hold: The ranked list, with Gift Aid marked as sent back and the report that replaces it named.
Stays human: Where the list starts. It is scored against the charity's capacity this quarter as much as its priorities, and the head of fundraising agrees it before anything is built.
Make and Test
The supporter reply skill is built from your own replies, the grant drafting skill from your previous applications, both in the tools your people already open. Test is us running a month of supporter questions and the last application through the two skills and setting the output against what the charity sent, with the fundraising team and the grants lead seeing the result. If either skill does not do the job, it goes back to Make.
You hold: The reply skill and the grant skill, each with the result of a test on your own work, and no figure from us about hours saved.
Stays human: Every number in an application and every claim about outcomes. A person signs the representation to the funder.
Trackside
We keep the connection and the two skills working, watch for new unapproved tools on the domain, keep the policy current as tools change, and feed the supporter replies that got rewritten back into the skill. Fundraising, the grants team and any volunteer who wants to come join the monthly shared session, which is where the next idea usually arrives from.
You hold: A service with a named person watching for drift, and a policy that is current the next time a funder or the regulator asks.
Stays human: The whole of the services team's case work. It never touches any of this, and the policy says so in its first rule.
Potential AI candidates
Twenty-one pieces of a charity's week across fundraising, grants, services, volunteers and the board, each with what changes and what stays with a person. The services team's case work appears once on the list and is marked never.
Part of the business
Verdict
21 candidates, across 7 parts of the business.
AI candidate
A draft in your voice for each supporter that refers to their own giving history and the work it funded.
AI candidate
A first draft from your published answers, with the supporter's details filled in, so the job becomes checking and sending.
AI candidate
First drafts of the letter, the email and the landing page from your case studies and your voice guide, for that person to shape.
Plain workflow wins
No AI involved. Your fundraising database or accounts package builds the claim from a report and a rule.
AI candidate
A first draft from your previous applications and project records, cut to the funder's own questions and word counts.
AI candidate
A first draft of the narrative in the funder's format, built from the project's own records.
AI candidate
Published funder criteria matched against your projects, with a shortlist for a person to judge.
Plain workflow wins
A calendar entry and a register line on the day the grant is signed. There is no AI in that and there does not need to be.
AI candidate
A draft of how the service runs for the person who runs it to correct, written up from an hour with them and from the shift schedule, the referral forms and the reports it already produces.
AI candidate
Drafts from your published service information, with anything the information does not cover flagged for a person.
Stays with a person
Nothing at all, including drafts, summaries and searches. These records are about individuals, often in the most difficult moments of their lives, and no AI tool goes near them on any account.
AI candidate
A reply drafted from the roles you have advertised and the steps that follow, with the application form attached, for the coordinator to send.
Plain workflow wins
One form, filled in once, that creates the schedule entry, the mailing list membership, the training record and the DBS tracker line. That is process automation with no AI in it, and for a charity that runs on volunteers it usually gives back more time than any candidate above.
AI candidate
The induction session recorded and turned into a short written guide with a handful of check questions at the end, for the coordinator to edit and hand to the next intake.
AI candidate
A one-paragraph summary at the top of each paper, and a first draft of the minutes from a recording of the meeting, for the company secretary to correct before the chair sees them.
AI candidate
Draft answers pulled from the policies and positions the charity has already written, with every question it has no written position on left blank and flagged.
AI candidate
A first draft of the narrative sections in the required structure, built from your own records.
AI candidate
A first draft built from the month's published updates, event pages and project notes, in the charity's voice, so the communications lead edits instead of chasing.
AI candidate
Six or eight posts from one story, each from a different angle, for the communications lead to choose from and schedule.
Stays with a person
Nothing. A charity's reputation is its ability to raise money, and a tool-drafted line about a safeguarding matter is a risk with no upside at all.
Plain workflow wins
An export from each platform, an import into the accounts package and a matching rule. No AI is needed.
Nothing on this page matches both of those. Try another part of the business or another verdict, or tell us what you had in mind.
The hard rule
If your charity supports people, the records of that support are the first rule in your policy and it is one sentence long. No AI tool touches them, no account connected to them is widened and no candidate on this page goes near them. That includes writing summaries, searching and tidying, because the risk sits in what the tool has read.
It is also where the permissions review earns its keep. A services team that has grown through volunteers and short contracts usually has accounts that were never closed and folders that are open to everyone, and that is true whether or not anyone connects an AI tool. Connecting one would turn a dormant permission into a live one, which is why the review comes first.
The policy for your sector
Proof
Independent since 1999, audited to ISO 27001 and ISO 9001, and already supporting teams like yours.
Charities and not-for-profits are part of the managed IT base, and we do pro bono work in the sector. The process behind it is the same ISO 27001 and ISO 9001 audited one that has run every client engagement since 1999.
No charity is named here, because none in the sector has yet given permission to be. Proof carries the full position.
Thirty minutes with the person who has to answer for it. We will tell you where a charity like yours would start, which of the candidates above are worth doing first and which we would send back.
Ask Cardonet AI
Ask in your own words, and we will point you to the part of this site that answers it.
Search runs in your browser, and we do not record what you type.