Your board doesn’t understand AI (and that’s a problem)

See the problem?

Everyone’s adopting technology. Hardly anyone’s planning for it. And the people who should be overseeing this; your trustees, your board, your leadership, often have the least understanding of what’s happening.

This isn’t a criticism. Boards are dealing with financial crises, safeguarding issues, regulatory changes, and rising demand for services. AI feels like one more thing to get their heads around when there’s already too much.

But it matters because boards make decisions about risk, spending, and strategic priorities. If they don’t understand AI, those decisions will be wrong.

Not catastrophically wrong necessarily, just wrong enough to cost you opportunities, or create problems you weren’t expecting.

What this actually looks like

It’s not that your trustees think AI is magic or robots. It’s more subtle than that.

It’s the CEO who says “we should probably do something with AI” without being able to explain what problem it would solve.

It’s the finance trustee who approves spending on AI tools without understanding what they actually do or whether you’ll use them.

It’s the governance discussion about AI policy where nobody really knows what they’re governing.

It’s approving a chatbot for your website because “everyone’s doing it” without asking whether your users actually want to talk to a chatbot.

Boards are making decisions about things they don’t understand, based on anxiety about being left behind rather than clear strategic thinking.

That’s the problem.

Why this matters right now

A few things are happening simultaneously:

AI tools are genuinely useful for charities: Automating responses to common queries. Drafting funding applications. Summarising reports. Helping with content. These aren’t theoretical – charities are doing this now and saving real time.

AI tools are also genuinely risky: They make things up. They can encode bias. They raise data protection questions. They create dependency on tech you don’t control. They can go wrong in ways that damage trust.

The sector is moving fast without thinking strategically: Half of charities don’t have an AI policy. A third haven’t done anything to move forward strategically with AI. People are just… trying things.

This is normal early adoption behaviour. But it means decisions are being made without proper oversight. And boards are supposed to provide that oversight.

What boards actually need to understand

Not the technical details. Not how large language models work. Not the difference between GPT-4 and Claude.

They need to understand five things:

1. What AI can and can’t do

AI is good at:

Pattern recognition
Processing large amounts of text
Drafting content based on examples
Answering questions based on information it’s been given
Automating repetitive tasks

AI is bad at:

Nuanced judgement
Understanding context it hasn’t been told
Dealing with edge cases
Knowing when it doesn’t know something
Anything requiring real-time verification

Boards need to know this so they can evaluate proposals sensibly. “Let’s use AI for fundraising” means something very different depending on whether you’re talking about drafting thank-you emails (good use) or deciding which donors to prioritise (questionable use).

2. The actual risks

Not science fiction risks. Real, today risks:

It will make mistakes confidently: AI doesn’t know when it’s wrong. It just sounds very certain. If your team is using it to draft funding applications or respond to service users, someone needs to check the output.

It raises data protection questions: What happens to the data you put into these tools? Where’s it stored? Who can access it? If you’re using ChatGPT to summarise case notes, you might be breaking GDPR.

It can perpetuate bias: AI trained on existing data reflects existing biases. If you’re using it to screen applications or write job descriptions, you need to think about this.

It creates dependency: Once you’ve built processes around AI, what happens if the tool changes, gets more expensive, or stops working?

Boards need to ask: “What could go wrong here?” and make sure someone’s thought about it.

3. What it actually costs

AI tools range from free to very expensive, but cost isn’t just the subscription.

It’s the staff time to implement, the training, the policy development, the ongoing monitoring and the mistakes that need fixing.

A “free” ChatGPT account isn’t free if you’re spending 10 hours a week figuring out how to use it.

Boards need to understand total cost of ownership.

4. The opportunity cost of not doing it

Yes, there are risks, but there are also risks in not using AI.

Your competitors are using it to respond faster, produce more content, and work more efficiently. If you’re not, you’re at a disadvantage.

The question isn’t “should we avoid AI because it’s risky?” It’s “how do we use it responsibly so we don’t fall behind?”

Boards that only see the risks will make overly cautious decisions. Boards that only see the opportunities will make reckless ones. You need both.

5. Their role in governing this

AI isn’t an IT question, it’s a governance question.

The Charity Commission’s guidance on decision making makes clear that trustees are responsible for major organisational decisions – and adopting AI qualifies.

Boards need to be asking:

Do we have a policy on AI use?
Who’s responsible for implementing it?
How do we monitor whether it’s being used appropriately?
What are our red lines – things we won’t use AI for?
How do we make sure we’re using it ethically?

This doesn’t require technical knowledge, it requires strategic thinking and oversight. Which is exactly what boards are for.

How to actually explain AI to your board

Forget the jargon. Start with problems.

Not this: “We should implement an LLM-based chatbot with RAG architecture to improve supporter engagement.”

This: “We get 500 emails a month asking the same ten questions. An AI tool could answer those automatically, freeing up our team for complex queries. Here’s what that would cost, here’s the risks, here’s how we’d monitor it.”

Frame everything in terms of:

The problem you’re solving
How AI helps solve it
What could go wrong
How you’ll know if it’s working
What it costs (time and money)

And be honest about what you don’t know. “We’re still figuring this out” is much better than pretending certainty.

Real examples to use

When you’re talking to trustees who don’t have time for abstract discussions, use concrete examples from your organisation:

Fundraising: “Instead of our fundraiser spending two hours writing individual thank-you emails, AI drafts them based on our template and donor data. She reviews and personalises them. Takes 30 minutes instead of two hours.”

Service delivery: “Our advice line gets calls asking if people are eligible for our support. An AI chatbot on the website answers basic eligibility questions 24/7. Complex cases still go to trained staff.”

Admin: “Board papers take hours to produce. AI can summarise long reports into 200-word summaries for board packs. We check them, but it’s much faster.”

Content: “We need social posts about our impact report. AI generates 10 options based on the report. Our comms team picks the best ones and edits them.”

These aren’t theoretical, these are things charities are actually doing. And they’re concrete enough for boards to evaluate properly.

What good oversight looks like

A charity whose CEO wanted to implement AI for donor communications. The board’s response was really good:

They asked:

What specific communications? (Thank you emails, not appeals)
Who checks the AI output? (Fundraising manager, every single one)
What data goes into the AI? (Public donor names and gift amounts, no addresses or payment info)
What if it makes a mistake? (Nothing goes out unchecked, mistakes caught before sending)
What’s the fallback if it stops working? (Templates exist, can revert to manual process)
How will we know if it’s working? (Track time saved, track donor feedback)

That’s not technical interrogation. That’s good governance. They understood enough to ask the right questions.

The minimal AI literacy boards need

Your trustees don’t need to know how AI works. But they should know:

AI is software that generates text, images, or decisions based on patterns in data
It’s very good at some things and terrible at others
It requires human oversight because it makes mistakes
It has data protection and ethical implications
Most charities are already using it in some form
The question is how to use it well, not whether to use it at all

That’s it. That’s enough to provide meaningful oversight. NCVO’s trustee resources can help frame this as part of broader governance responsibilities.

What to do if your board is resistant

Some boards hear “AI” and shut down. Too risky, too complicated, too expensive, not for us.

If that’s your board:

Start small and show results: Don’t ask for a big AI strategy, instead, ask for permission to trial one specific thing. Show it working, then expand.

Frame it as risk management: “Other organisations are using this. We need to understand it so we can govern it properly, even if we decide not to use it.”

Use their expertise: Someone on your board understands risk management, or data protection, or change management. Ask them to apply that lens to AI. They’ll engage better.

Get external input: Sometimes boards listen to outsiders more than staff. Bring in someone who works with charities on AI to talk them through it.

What to do if your board is too enthusiastic

The opposite problem: boards who want to “do AI” without thinking it through.

If that’s your board:

Slow them down with questions: “That sounds interesting. What problem are we solving? What are the risks? How will we know it’s working?”

Require business cases: Any AI tool should have a clear case: time/money saved, problem solved, opportunity created. If you can’t articulate it, you shouldn’t do it.

Insist on trials: “Let’s trial this with one team for three months before rolling it out” gives you space to learn without committing fully.

The conversation you should be having

At your next board meeting, put this on the agenda:

“AI Use and Governance: Discussion”

Questions to address:

Are we currently using AI tools? (Staff probably are, even informally)
Do we have a policy on what’s appropriate use?
What are our red lines, things we won’t use AI for?
Who’s responsible for oversight?
Do we need training or external advice?

You don’t need answers immediately, but you do need to start the conversation.

Because right now, decisions about AI are happening anyway. They’re just happening without board oversight, without policy, without strategic thinking.

Human

Kind