AI decisions for purpose-led organisations
Mission Fit:
A defensible stance on AI.
Mission Fit weighs each AI idea against your mission, not a one-size-fits-all checklist. You get a clear answer you can defend to your board, your funders and the people you serve.
The bind
You are being told two things at once. Both are true.
Move fast on AI, or fall behind. Funders ask what your AI strategy is. Trustees read the headlines. Your own staff are already pasting things into chatbots, policy or no policy.
Be careful, or you will do real harm. A company that gets AI wrong issues an apology. A values-led organisation that gets it wrong can lose the trust it runs on. That is the one asset you cannot buy back.
So the conversation stalls: the enthusiast against the sceptic, two strong opinions and no way to settle them. What is missing is not more opinion. It is a method.
Show, don't tell
Watch one idea get three different answers
Here is a proposal we hear in some form from almost every organisation. Run it through a generic AI checklist, and then through Mission Fit, as three different charities.
The proposal
“We want an AI assistant on our website, so people get an answer straight away instead of waiting for us to reply. It is there at three in the morning, and it costs a fraction of a person.”
A generic AI checklist
Approved.
The same result, for every organisation that runs it.
Mission Fit. Choose who you are:
People in crisis type into whatever box is on the page, at three in the morning, when nobody else is awake. That is not an edge case for you. That is your core user on their worst night.
A confident, plausible, wrong answer to someone in that state is not a customer service problem. And you would not find out that it had happened.
The framework prices this direction too, rather than treating caution as automatically safe. People are waiting now, and some of them wait badly.
Decline
This one never reaches the weighing up. Anything that could put AI between your organisation and a person at acute risk is pulled out before scoring begins, and held to a standard closer to a medical device than a website feature. You do not have that standard in place yet, and the framework will not pretend otherwise.
What this is not: a ruling that AI has no place in your organisation. It is a no to this use, on this proposal, today. Something that never answers but gets a person to a human faster is a different proposition, and it would be assessed on its own.
“The checklist gave all three organisations the same answer. Their missions gave three different ones. Which would your board rather defend?”
Simplified for the web from assessments run under the Mission Fit framework, version 1.6. Each panel has been checked against the framework's own decision logic, so the site never shows a result the method would not produce. The panels summarise the questions that decided each verdict, not the full written record.
Why now
Your organisation already has an AI practice.
The only question is whether it has a decision process.
Ungoverned doesn't mean unused. The risky calls are already being made, by whoever happens to be at the keyboard, with no evidence, no record and no board oversight. Waiting doesn't make those calls go away. It just means the first your trustees hear of one may be a complaint, a breach or a headline.
Source: 2026 Nonprofit AI Adoption Report (Virtuous / Fundraising.AI, 346 organisations).
The foundations
The method is built on what the world already trusts
Every question Mission Fit asks traces back to a recognised international standard. That is why its verdicts hold up in front of boards, funders and regulators.
UNESCO · Ethical Impact Assessment
The backbone of the assessment: examine each AI use before adopting it, question by question, with evidence rather than opinion. Agreed by 194 countries.
UN Guiding Principles & human rights impact assessment
The discipline behind “does it harm people”: name who is affected, show how this use affects them, and treat severity seriously instead of averaging it away.
NIST AI Risk Management Framework
The engineering rigour: can you trust what a system produces, how do you check, and how confident are you in your own answer. Every Mission Fit verdict carries a confidence level for exactly this reason.
EU AI rules & data protection law
The legal floor, in force since 2 August 2026. Mission Fit is built to line up with what the rules now ask of organisations using AI, and it feeds your compliance work rather than replacing it. It also asks one question the law implies but generic checklists skip: who supports you can itself be private information about them.
And then Mission Fit adds the one thing none of these provide: weighting to your specific mission, and a clear decision at the end.
The process
How it works
Screen
Two things are settled before any weighing starts: whether this could touch someone at acute risk, and whether it crosses a line no organisation should cross, whatever the benefit.
Score
Value and risk, rated against written descriptions and weighted to your mission. Every score needs evidence and carries a confidence level.
Decide
One of four answers: Adopt, Adopt with controls, Redesign, Decline. A serious, evidenced risk on a mission-critical question stops the idea, whatever the saving.
Note
What can't be scored yet must still be watched. Watch-notes keep tomorrow's risk on today's record.
Underneath the scoring sit four plain questions, asked of every idea:
- 1Does it harm people?Privacy · fairness and dignity · harm to the people you serve
- 2Can we trust what it produces?Accuracy · honesty about what is real and what is generated
- 3Does it harm the world beyond us?Environment · jobs and livelihoods · whether your supplier's conduct fits your values
- 4Does it harm what we stand for?Contradiction of your own mission and public positions
Safeguarding is never a trade-off.
Any use that could touch a person at risk is pulled out before scoring begins and held to a higher, separate standard. A vulnerable person's safety is never placed on the same scale as staff time saved.
Your mission can raise the bar. It can never lower it.
A small number of protections hold for every organisation, whatever the saving and whatever the cause. Weighting decides what you should do on top of those, not what you can talk yourself out of.
What you walk away with
- →A clear answer, not a grade: Adopt, Adopt with controls, Redesign, or Decline.
- →The one thing driving the risk, named in plain language.
- →The lever: precisely what would turn a “no” into a “yes”.
- →Practical safeguards with named owners, so nothing is everyone's job and therefore no one's.
- →A confidence level, so you know how solid the call is and what to check next.
- →A decision that expires, with an owner, a review date and one thing that suspends it. A yes given in March should not still be running unexamined in December.
- →A board-ready record of how the decision was reached, produced as the session runs rather than written up afterwards. The document a trustee, a funder or a journalist could ask to see. You'd be glad they did.
Start with one
See it work on your own example
You don't have to resolve AI in the abstract. You have to make good, defensible decisions about the handful of uses actually in front of you. In one session we'll run a real AI idea of yours through Mission Fit, and you'll see, on your own example, exactly what you'd walk away with. The session is run on the working tool the method is built into, so the record you leave with is the one produced in the room, not a write-up that follows a week later.
Tell us what AI idea or challenge you are weighing. For example: staff want AI help with donor letters, or you are deciding whether to roll out Copilot or an enterprise AI account.
Request a session