I've spent my career working with charities whose demand is bigger than their capacity. There's always more they could do. More people to reach, better support to offer - and countless great ideas that never get tried because everyone's already stretched.
And right now, that gap is getting harder to close. For many, income is flat while costs keep climbing. At the same time, there is rising demand for services.
AI is the biggest opportunity I've ever seen to give charities more room to act.
It can help us do the work we already do better, and it can make things possible that weren't possible before. We need to pursue both.
That's why I'm so positive about this technology. I can see an opportunity for charities to become more capable, for people's work to become more rewarding, and for our missions to come within greater reach.
We're already seeing it in practice in the pilots we’re running across the sector. At Centrepoint, frontline workers in our case-note pilot reported getting around five hours to one whole day back per week on admin tasks. They then spent more time with young people and on planning better support. These are staff-reported pilot findings; we're still learning about the longer-term effects - but that is a very worthwhile start, and we’re seeing similar results from pilots across a variety of charity teams.
And there's so much more within reach. The gap between what today's tools can do and what we're doing with them is enormous. And we often overestimate how difficult it will be to do something useful, when there are hundreds of low-hanging-fruit opportunities staring us in the face.
At Good Innovation, we've developed Conscious Adoption through our work supporting more than 20 UK charities to adopt AI. We're seeing it work, and I want other charities to take it, adapt it and put it to use.
AI is too important to the sector for us to keep what we're learning to ourselves. Which is why transparency is a principle of our work: sharing what works, what doesn't and what we're still figuring out. It's why I'm constantly banging on about AI on LinkedIn, and why I want to write more long-form articles like this if they're useful. Whether you work with us or not, I'd love your charity to benefit from what we learn along the way.
Now let's get into it: I'll start with three incomplete AI Adoption playbooks, then set out Conscious Adoption - our view on what good AI Adoption looks like for the sector. It combines leadership, broad capability and focused work - which together delivers better work, new possibilities and mission outcomes.
Better work & new possibilities
Better work means improving what we already do. Less effort on admin, higher quality analysis, clearer comms, and more reliable processes. Ultimately: more room for people to use their judgement and build relationships.
New possibilities means asking what we could do that we can't do today. With AI at their side, can a small team now punch well above their weight, doing work they only previously dreamed of? Could we make useful information accessible to people we struggle to reach? Could we test a new service that we've never had the capacity to explore?
Those are questions to investigate. We won't know which ideas work until we try them - but if we only look for faster versions of today's work, we'll never find out.
We should explore both forms of AI adoption - and neither should be judged on time saved alone. Throughout this work, we look at Value, Trust and People: what improves, what we can stand behind, and what changes for the people involved. I'll come back to those.
Three incomplete playbooks
The 2026 Charity Digital Skills Report found that 79% of responding charities were exploring or using AI, but only 4% were using it strategically. Now that’s a gap.
In our work at Good Innovation, three approaches keep coming up, consciously or otherwise. Each has something useful in it, but I’ll be clear from the top: none of these will lead to good adoption, and the kind of results we’re all looking for.
1. The tech rollout
This is generally the approach when IT is in charge. Licences are rolled out, and everyone gets some one-off training. This approach would work if AI was normal software (hint, dear reader: it is not).
Don’t get me wrong: IT teams are essential, especially when it comes to data security. Everyone should get access to a quality AI tool, and AI literacy is the foundation of everything: people need enough understanding to use these tools well, recognise their limits and question what comes back.
But too much of the training I see is a tour of a chatbot’s features that go out of date within a month.
People need to learn sustainable skills: how AI fits into their actual work. What makes it different to work with than any other piece of software? When should it take on a task? When should it stretch your thinking? When do you need to think for yourself, or bring other people into the room? What needs checking, and how? When should you leave AI out entirely?
People need support to practise on real work, and support after the training ends.
A good chatbot can make someone's working day considerably better. Give people that opportunity, but no matter how good the training is: a collection of individual gains doesn't automatically change how an organisation works.
If everyone writes reports faster and the same three people still have to approve them, you may just have made the queue longer. Congratulations - you've automated a bigger backlog.
A chatbot subscription is a start, but by no means good AI Adoption.
2. The efficiency trap
There's a private-sector playbook I don't want charities to borrow: cut headcount, put AI in the announcement, call it progress.
Call me cynical, but these announcements look more interested in the share price, the next earnings call or an IPO than in showing what AI has actually delivered.
We don't have a share price to prop up or an IPO to dress up. We have a mission to deliver.
Efficiency matters enormously. Sometimes reducing costs is what keeps a service alive. But the trap is treating time and cost as the only measures of success.
Even within existing work, faster isn't the only improvement worth making. (And many charities exist because of the failures of “efficient” markets in the first place.)
Better quality, fewer mistakes, less frustration and a more useful result all matter. Then there's the work we could do for the first time.
Most charities I work with have people they can't reach, thousands of donors they haven’t engaged personally, relationships they can't invest in and ideas they haven't got time to try. Before treating every gain as a reason to cut headcount, I'd want to know what it could make possible.
What does the time become? Who benefits from the new capability? Those are leadership decisions.
Making work better for people is part of the point, too. Less drudgery, and more energy for the work they came here to do.
It also matters enormously for whether adoption fails or succeeds. If every improvement looks like evidence that someone's job is unnecessary, why on earth would anyone help you find the next one? Good adoption needs people willing to experiment, admit mistakes and share what they're learning.
Leaders have to earn that trust - be clear about why you're using AI, involve staff in changing the work, and give them time to learn.
Let's be honest about jobs
Like many topics in the social media age, this is a nuanced topic we’re forced into binary camps about. Granted, it feels less nuanced for the call centre worker being replaced by AI, or the taxi drivers under threat by robo-taxis - but for most jobs, we appear to be looking at roles evolving and changing, rather than being replaced. How is that, then?
Jensen Huang makes a useful distinction between the purpose of a job and the tasks within it. A taxi driver’s and a call centre worker’s purpose and tasks are the same thing - which puts those types of roles at risk of automation in the coming months and years (regardless, it is still a human decision whether we want that to happen).
But most roles have a much more complex set of tasks involved with them. For instance, a fundraiser's purpose is to grow support and income - some tasks might include researching prospects, drafting appeals and updating CRM records.
AI can take on some tasks, assist with some, help expand the possibility of others - giving the fundraiser more time for the tasks that stay fully human. As you can probably now see, automating one task doesn't remove a role - and every use of AI still needs an expert human overseeing it (no matter what out of touch tech CEOs say).
Another point worth flagging here: the kinds of things AI is currently good at, also happen to be a lot of work that more junior colleagues use to learn. Sure, this could make us more efficient - but what happens to the future of our missions? Future expertise needs somewhere to come from, so there will be a job to redefine what a junior means in the age of AI we’re entering.
To bring it back to the here and now, it’s within our control to make this transition beneficial to the mission and our people. For example, we've recently worked with a supporter engagement team who wanted more time to call supporters. Sorting emails and copying routine replies kept getting in the way. Simple automation and AI helped take on that admin, while people checked replies and handled sensitive queries. The team described putting that time back into supporter calls - and ended up enjoying their jobs more as a result of the AI pilot, not less.
That’s the world we should be aiming for - and have full control to create. The technology doesn't decide whether saved time becomes a better conversation, a bigger workload, a redundancy, or more hires. People do.
Sometimes the finances leave leaders with painful choices and very little room to manoeuvre. I get it. But as soon as you tie AI adoption to redundancies, one thing will be absolutely clear: nobody is going to want to use AI, and you’ll be a worse organisation for it.
3. The policy masterpiece
Our final incomplete playbook is a bit tongue-in-cheek, but also: I’ve seen it happen dozens of times in the sector. It typically looks like this: you spend months (or, sometimes, years!) writing a very long AI policy. After another few months, it gets signed off by the various committees. You then send it on a six-month governance pilgrimage - and finally, at last, put it on the intranet.
Then…wait for responsible adoption to occur…and wait…and wait…
Policies matter, and you 100% need one. But staff don’t need a policy day to day - they need something they can actually use: one or two pages explaining which tools they can use, what data they can put in, which uses are off limits, what needs human review and who to ask when they're unsure.
Give them that guidance now, because most of them are using AI whether you know it or not. Put real examples in it, keep it current, and make it straightforward to try useful, low-risk things within some sensible boundaries.
And whilst you’re at it, find out what's already happening. In charities we work with, people are using tools outside the official offer. If the approved tool doesn't meet their needs, people look elsewhere. A rabbit hole I’m not going to go down, but if the only tool people have is the free version of Copilot (or even the paid version), I can guarantee that a large percentage of your staff are using ChatGPT, Claude or Gemini on the side. Your guidance needs to assume that.
I could go on, but I think we can agree at this point that having a long AI Policy and some guardrails does not in and of itself lead to good adoption. So let’s turn over our attention to what does.
Introducing Conscious Adoption
Ambitious about AI, and accountable for its impact.
Charities should be among the most ambitious organisations using AI. We've got more demand than resources, and an extraordinary opportunity to close that gap.
I think we should help set the standard for responsible AI, too. That means creating real mission value, protecting trust and making work better for people, while understanding and managing the risks. Trust is the lifeblood of our missions - and our staff, volunteers, supporters and the people we serve should see our values in the choices we make about AI.
And our resources need to count. Buying licences, telling everyone to use AI for everything and hoping the benefits follow isn't a strategy.
So leaders need a clear point of view on AI: what it's for, what it could make possible, where the boundaries are and what it means for people's work. Communicate it clearly and often, and update it as you learn.
Give people the tools, skills and confidence to use AI well. Alongside that, find worthwhile opportunities and test them safely. Keep the first steps proportionate, and use Value, Trust and People to judge evidence and decide where to invest next.
Then, when something works, properly back it. Build it into the work, and help more teams benefit.
If we can show that AI helps us deliver more of our mission, leaving those benefits stuck in a pilot is a failure of ambition, and a failure of our mission.
Some of this adoption work should make today's work better. And some should make entirely new things possible. We need both.
All of that combined is Conscious Adoption: Good Innovation's approach to delivering lasting mission value through responsible AI. It connects clear leadership and communication, capability across the charity and focused work that delivers.
Let’s dive into the three connected lenses of Conscious Adoption:
Set direction: leadership, strategy and governance
Leaders need to be clear about what AI is for, and communicate relentlessly before people take what they hear in the news and assume that AI is simply coming for their jobs. What do we want to improve or make possible? What must we protect? Who can make decisions, and who owns the results? That direction helps people choose worthwhile work and know where the boundaries are.
Then back it with suitable tools and practical guidance.
Go wide: horizontal adoption
This is the work of building AI Literacy across the charity. Access to tools, practice on real work, local support and shared learning. People need the confidence to use AI, question it and know when to leave it out.
It's sustained people and culture change - so naturally, this work takes time and effort. And leaders, staff, volunteers and the people building reusable tools need different kinds of support. Individual wins can come quickly from this type of adoption; but changing habits, culture and processes across an organisation for AI to have wider benefits takes time.
The skills for this form of adoption are teaching, communication, organisational change and making technology usable. IT, learning and development, people teams and managers all have a part to play. Look for better judgement and useful changes in everyday work, alongside confidence and participation.
I think of this type of adoption as setting the floor, or the foundations. It’s slow work, and you’re not going to see transformative results from it. Because, in essence, the work hasn’t changed - you’ve just introduced a new tool for people to use.
Go deep: vertical adoption
This means working with a team on a particular service, workflow or new capability to run a pilot and improve or transform how work is done. It requires finding a genuine problem, and testing how AI can help - with the work around the tool itself just as important. Measure the results, and if successful, roll it out further.
In our work, this is where we're seeing the clearest gains right now. A focused pilot can give you useful evidence much sooner than waiting for habits to change across the whole charity. That's why we often encourage charities to start here while building wider capability alongside it.
This form of adoption needs an experimental, innovation mindset: understanding needs, framing problems, mapping workflows, prototyping, redesigning work and evaluating results. Bring together people who know the work with the technical and change skills it needs. The pilot is just the first step - embedding the improvement takes ownership, support and continued learning.
Different skills, one approach
Horizontal and vertical adoption are different kinds of work - and require different skillsets. Running excellent AI training and leading an experiment to redesign a service aren't the same thing, even though they’re often lumped together under ‘adoption’. The same people can cover both, especially in a small charity, but both sets of skills need to be there.
Culture runs through all of it. People need time to try things, permission to challenge an output and enough trust to say 'that didn't work'. Leaders need to make that normal.
Ultimately, it is the interconnectedness of the three lenses of adoption that make for good adoption. Wider literacy helps people spot opportunities, whilst focused pilots produce evidence and examples that improve the training - and start to change culture when people see how AI can be tangibly helpful.
Leaders then use that learning to update guidance, set priorities and back what works. Wider culture change then helps useful approaches spread informally. In time it all moves together, with mission outcomes keeping them pointed in the same direction.
How to judge progress
These are our common tests across the whole approach, from everyday use to focused experiments and the wider adoption programme.
Value. What gets better or becomes possible, and for whom? Look at quality, usefulness, time saved and mission contribution, alongside the total effort and cost. If you free up capacity, decide where it goes and check what it enables.
Trust. Can we stand behind the work? Check reliability, safeguards and accountability. People affected need a way to question the result and get human help. Be honest about what the evidence does and doesn't tell us.
People. Is work becoming better, and are people developing the skills and confidence to do it well? Involve staff, volunteers and, where relevant, the people using your services. Protect the judgement, learning and relationships the mission depends on.
Efficiency is one measure within that picture, but a useful improvement doesn't have to save time to count. And match the scrutiny to the consequences: a low-risk everyday task within agreed guidance needs much less than a new service affecting vulnerable people.
Responsible AI
Here's my working definition of responsible AI: using AI to advance a worthwhile purpose, understanding who could benefit or be harmed, putting appropriate safeguards in place, and staying accountable for what happens.
I want the sector to be ambitious enough to explore the opportunity, and informed enough to make good choices. That means taking five areas of risk seriously:
Reliability. AI can invent facts, sources and plausible nonsense. Check important claims against original evidence and test it on real work. A human reviewer needs expertise, time and authority to intervene - and you need to be clear that humans are also 100% accountable, blaming the AI is never an excuse.
Bias and exclusion. The training data of these tools is as biased as the internet: Western, white, and male - and your AI literacy programme needs to address that, alongside mitigations. AI tools can also work well on average, and badly for a particular community, language or disability. Involve people affected by a tool, test who it works for and provide an accessible alternative if required. If used well, AI can have a positive impact on inclusion, but that’s only if you plan for it.
Privacy and security. Understand what the tool stores, who can access it and what permissions it has. Connecting AI to an inbox creates different risks from improving a paragraph. Sensitive information needs appropriate protection; and you definitely already have policies covering all of this, so there’s no need to reinvent the wheel.
Environmental impact. AI is an incredibly efficient technology, and it is becoming more efficient. But it uses energy, water and hardware - and the sheer scale and growth of it are an environmental concern. The impact varies with the model, task, infrastructure and scale. So use it where the benefit justifies the cost, avoid needless generation and review the data alongside the suppliers you work with.
Ethics and human impact. Consider creators' rights, the labour behind the systems and the concentration of power in suppliers (top tip: don’t use Grok). Ask what happens to skills, jobs, dignity and human connection - and ensure that your AI Literacy programme talks about how to keep your critical thinking skills sharp in the age of AI. Be clear when people are interacting with AI, and preserve human support where it's needed.
Some risks can be reduced through training and better work design. Others need technical controls or specialist judgement. We have less control over supplier practices, but we can scrutinise them, choose differently and keep reviewing the evidence.
Charities also have a role in shaping what happens next. AI is changing the world our communities live in, including the problems we exist to address. We can use it thoughtfully while challenging its harms and campaigning for better rules.
Start with what's within reach
You can get all three parts moving without turning this into a huge programme. Agree what you're trying to achieve, make the boundaries clear and choose a useful place to start.
Help people use the tools already available to them, and give a willing team space to test something worthwhile. Sensible everyday use within agreed guidance doesn't need its own pilot. Keep the effort proportionate to what you're trying to learn.
To help you put this into practice, we've made a one-page Conscious Adoption checklist. It brings together leadership, going wide and going deep, with Value, Trust and People running through all three. Use it with your team to spot what's already in place and pick your next move.
Take it, and make it yours
We're seeing Conscious Adoption help charities turn AI into better work and new possibilities. That's why I'm sharing it - and I would love to see you use it.
Make it fit your charity, your people and your mission. Get leadership, wider capability and focused work moving together - and use Value, Trust and People to judge what's improving and what needs to change.
Start with something worthwhile, and share what you learn. We're still learning too, and I'd love to hear what you'd do differently.
And if you'd like some help putting it into practice, that's what we do at Good Innovation. We can help you set a clear direction for AI, build your team's capability and turn a promising idea into something that works. Get in touch.
There's so much this sector wants to do: more people to reach, better support to offer, good ideas that deserve a chance. AI gives us more room to do it. We’d be silly not to make the most of it now, wouldn’t we?