This transcript has been generated using automated tools. It has been checked but minor errors or omissions may remain.
00:00: Chapter 1: Introduction
[Intro music]
[Voiceover, Ka Man]: Welcome to Fresh Humanitarian Perspectives, the podcast brought to you by the Humanitarian Leadership Academy.
[Music changes]
[Voiceover, Ka Man]: It’s absolutely critical that as we rebuild this humanitarian sector in a way that we want it to look like, AI is going to be a part of it, whether you’re using it or not.
[Voiceover, Madigan]: I think we have a really exciting future ahead now that we’re starting to shift into this new world of AI literacy and readiness and actually turning it into practice — but there comes risk along with it, and I think we also need to be well aware of those risks and know how to mitigate them as well.
[Voiceover, Ka Man]: Since May 2025, the Humanitarian Leadership Academy and Data Friendly Space have been leading the first worldwide study to track how humanitarians are using AI in their work. It’s been something of a journey.
I’m Ka Man Parkinson, Communications Lead at the HLA. I’ve been co-leading this research, and in today’s episode I sit down with fellow co-lead Madigan Johnson from Data Friendly Space to take stock and to dive further into some of the nuance of the research insights, as well as to discuss the release of our new Kaya course on Ethical AI use and decision-making in humanitarian work.
In this conversation, we aim to be candid about what we know, what we don’t know, while also being practically oriented with our recommendations in this space, setting out why we believe AI literacy and governance that meets practitioners where they are are critical priorities for the second half of this year.
This episode is the first part of a two-part discussion. In part two, I reconnect with Musaab Abdalhadi, Nour Arab and Ali Al Mokdad — humanitarians who have been an integral part of this research and convening journey.
[Music ends]
Ka Man: Hi Madigan, welcome back to the podcast!
Madigan: Thank you so much for having me back, Ka Man. I’m excited to be here!
Ka Man: On that’s fantastic. So for those listeners who are new to our work, or aren’t familiar with our AI in the humanitarian sector collaboration at large, could you briefly introduce yourself and the work of Data Friendly Space?
Madigan: Of course. My name is Madigan Johnson. I am the program manager at Data Friendly Space where I lead research, communications and AI and data projects. So Data Friendly Space is a nonprofit that helps humanitarian organisations make smarter decisions with data, and so we work with humanitarian practitioners to figure out what’s needed in terms of technology, AI and data.
Over the past year or so we’ve been partnering with the HLA, and we’ve been so happy about this, because I think we’ve genuinely done the first study tracking how humanitarians are actually using AI. So that work has been, you know, quite the eye opener, and I’m so pleased to be back here with you today, Ka Man, to talk about it.
Ka Man: Thanks so much, Madigan. We’ve actually passed a bit of a milestone, haven’t we, because we’ve just passed the one year, the 12-month mark, since we started this collaboration. And like you say, it’s been such an eye opener, and it continues to be such an enriching and engaging part of our work, collaborating together as organisations with the humanitarian community at large, as well as technologists and other people operating in this space, to try and connect those dots so that we can work together for the movement for humanitarian AI.
I thought it’d be nice for us to just briefly reflect a little bit on the second phase of the research that started in January, which kicked off with the pulse survey to check in to see how humanitarians are adopting AI, or not as it may be. And also to reflect briefly on our collaboration on the new Kaya microlearning on Ethical use of AI and decision-making in humanitarian work, which we’re both really excited to collaborate on.
First of all, before we dive into that, for the benefit of listeners who may be new to our work — what we found last year when we started this research is that — it’s what we coined the humanitarian AI paradox, where individual uptake of AI tools like ChatGPT, Copilot and so on, is racing ahead. And that’s really driven by interest, but also pressure — increased pressures of work, in light of the sectoral changes and organisational restructures, and so on. And in contrast, overall organisational readiness is low.
And we checked back in in Jan to see if anything has shifted. And what we found is that the paradox has, shall we say, deepened. Individual uptake has increased — so 75%, three-quarters, are now using AI tools daily or weekly — but organisational adoption as a whole has not really shifted. And, that’s quite sobering, but also not totally unexpected, given the severity of funding cuts and all of the ramifications of that.
What really shifted was the individual conviction and belief in AI, compared to when we first did the survey last year – that was really significant.
So, for those who want to dive into the stats further, our briefing report is on our website that you can download and read at your leisure, and Data Friendly Space also created this fantastic dashboard so you can deep dive into the disaggregated data as well.
06:12: Chapter 2: AI literacy as connective tissue and governance as structural support
Ka Man: I’ll stop there and I’ll bring you back in, Madigan. What surprised you? What were your key takeaways, or anything excite you or concern you from what we found in this phase of the research?
Madigan: Yeah, so much actually surprised me, from the very beginnings of the research to even now with the pulse survey. Maybe I can start with what I expected in this last round that we did with the pulse survey. The individual adoption going up — I did kind of see that coming. I think people are curious, they try the AI tools, that’s sort of human nature.
But what I didn’t expect was, like you touched on Ka Man, how fast that conviction has shifted. It was not just you know “I tried this AI tool once.” People are genuinely starting to trust – well maybe not trust – but starting to believe that AI is making their work better and more efficient and leading to better decision making. And so there is that significant jump in trust, and I think we really need to sit with that, and that happened much quicker than I expected.
I think the other side is where the confidence is sort of concentrated. Something, again, I would love to do follow-up research on — it’s probably not evenly distributed. The practitioners who probably feel like AI’s value is really shifting are the ones who have had the space to experiment.
So I think when you’re in a team where people aren’t openly talking about the tools they’re using, when there’s no shared vocabulary for thinking through a tricky AI decision, there’s no course that kind of fixes that, as much as I’d like there to be. I think you need the culture first, or at least alongside.
And I think AI literacy sits at the heart of that – it’s the connective tissue that holds everything together. You can have a policy document that maybe some of your team members have created, but if the practitioners or the people that are actually working there on the day to day operations don’t understand enough about how these tools work or how that policy is meant to be implemented, then it actually doesn’t end up helping.
What I’m trying to say is that AI literacy isn’t about making everyone an expert – it’s about giving people enough of a mental model to kind of ask the right questions. And I think that’s what we really tried to do within our course as well. Should I trust this output? What would happen if this was wrong? Who should know that I’m using this? Or what data privacy policies do I need to know? And I think those questions really matter.
And what really matters as well is the support that goes alongside of it. Sometimes we treat AI literacy as an individual problem to solve — go take a course, or experiment with AI — but I think the practitioners who might be further behind, it’s not because they haven’t tried, but also because their organisations haven’t created the conditions in which to learn. I think we also really need to focus on that cultural shift as well.
But I don’t know, Ka Man, what about you? What are the key takeaways for you, or where did things really surprise or concern you for that matter?
Ka Man: You’ve just said it so well, really, about that importance of literacy being seen in that really holistic lens. Because what really stood out like you say, is that the humanitarian sector is obviously not a homogeneous group. Sometimes we probably oversimplify things for our way of seeing the world and benefit. But what the research really sort of showed and really reinforced is the huge amount of difference across the sector, and the different behaviours that don’t cut neatly across the ways that we maybe see the sector — like international organisations are like this, and local organisations are like this. Although we did pick up some of these trends, like you say.
I see so much need for structural support, and I think actually, the sector, like I say, the humanitarian AI paradox plays out here as well, where individuals can do – you know, their efforts are really important, crucial, but systemically — that’s what we really, that’s the challenge, and that’s what we really need to drive.
So, for example, on the governance front, we saw such a difference between the proportion of local organisations who have policies through to international organisations and UN agencies. So I think it was 39% for UN agencies and 13% at the local level.
So really, my takeaway from that is we have a responsibility: those with the most resource should be supporting – structurally – those in the sector, and providing that scaffolding for those with less resource. So I’m not necessarily talking about that, in terms of, like I say, across those silos that I talked about, but sector-wide efforts – really thinking about AI literacy and governance as in public interest, rather than individual, organisational – you know – benefits. Although of course that is an incentive and motivation as well.
11:19: Chapter 3: Ethical AI microlearning: a co-creation process with humanitarians
Ka Man: I’m really excited that we’ve developed this course together, this Kaya microlearning module. Now obviously we’re not claiming it’s a silver bullet. It is a small step, because AI literacy – contextualised for humanitarians is still in its relative early days. There are courses and training for nonprofits as a whole — for example, NetHope are spearheading a lot of great initiatives in this area and they have the NetHope Lighthouse, so there’s lots of great stuff going on, and Elrha has just introduced a new humanitarian AI course too. This is something that we really need to focus our efforts on now. So Madigan, what does this course mean to you? Why do you think it matters?
Madigan: I think for me something that really resonated with me while we were doing this is the way that we approached it — the co-design process with humanitarian practitioners. And I think that was something that -— and not just with the practitioners but also the ones that engaged in our survey. What came out of the survey was that so many people were asking for contextualised learning. And so I think what we did was actually listen to their needs and say, okay, how can we build something with them? So it wasn’t for them — and I think it’s really important to make that distinction — it was with them. And so that co-design process, involving so many different perspectives, was really crucial. And I think it’s really how we have to approach AI literacy going forward in the humanitarian sector.
There were so many different points of AI literacy that we could have focused on. When we were going through the survey results and people were writing “we would like a course about human in the loop,” or ethics, or “how do I do the best prompting” — it was really hard to narrow down, and I think we had a lot of conversations as well about that.
And I think what it came down to is that a lot of the ethics guidance — and I want to say this with the utmost respect — is written for more policy audiences, right. It’s really comprehensive, it’s rigorous. And for a practitioner with a deadline, they’ll usually read the first paragraph and then kind of close the tab.
And so I think what we really tried to build in our course is something that lives closer to the actual decision. So you know, here’s the situation, here are the tensions that come with it, here’s a way of thinking it through — but also maybe there are other ways to approach it. So you might see, some answers aren’t clear-cut; it’s a judgement call, right.
So I think there’s a process I want the sector to take from — it’s not just the course itself, but also the model. If you’re building AI literacy and you’re not in constant conversation with humanitarians and with the practitioners, you might be building the wrong thing.
And so what I would love to see the sector do is see AI literacy become as standard as safeguarding training. It shouldn’t just be one course, but rather a whole ecosystem of contextualised, practice-based learning that grows alongside the tools, rather than chasing them — especially with AI and having the rapid development it has.
But I don’t Ka Man, what do you think? Why did this course matter to you, and what did you take away from building this?
Ka Man: Such great points about the critical importance of co-creation — so not just co-design, but co-creation — and working with practitioners to test our assumptions. Because we have a learning framework, but how do we know if that actually – especially because this is unprecedented – this is a new emerging area. So having that feedback from our collaborators was absolutely critical, and we can’t thank our colleagues enough for supporting us on a pro bono and in-kind support basis.
Shout out to my colleague Phil Street, who is our learning design maestro who created and built this beautiful interactive course on Kaya. And we were actually – what served as a bit of a basis of inspiration was a module that he worked on together with a colleague maybe a couple of years ago now called Introduction to Conflict Sensitivity.
We tried to bring through that learning principle into this course too, because people think “AI, AI literacy, it’s about learning how to prompt”. But obviously we had to sort of deconstruct that — it’s about how you’re using the tool, and crucially, for frontline humanitarian practitioners working directly with communities, what implication does using this AI tool today, what might the implications be beyond you, for example, copying and pasting data into a tool? Like what are the implications? So that’s really what we were trying to bring to the fore, especially while frameworks are emerging, organisational approaches are firming up as well. So really we’re drawing on the humanitarian principles that are already established and understood — do no harm, impartiality, neutrality. Really, that was the framework which we worked with, together with humanitarians.
As I say, there’s an absolute need for structural support and sectoral support, and we hope that organisations in their own AI adoption journeys will start to formalise approaches through policies. That’s what I really took away — the need for organisations to develop policies and to socialise that through different mechanisms, such as through Kaya courses. For example, we have Kaya Enterprise Partners who have the ability to design and publish their own courses on Kaya. And we have at least one enterprise partner using that feature to socialise their own organisational AI policy — so I was really excited to see that development.
And also my colleagues, my digital learning colleagues, they also work across the sector to provide specialist digital consultancy. So my wish is that organisations and people across the sector will be able to use Kaya – this is a mechanism to facilitate organisational learning and to support AI literacy. And the really nice thing about the Enterprise partners in general is that you have the ability to make courses public. So many of our partners have chosen to make their internal courses for their staff public. So that’s really a contribution to the sector at large. I think that’s a really beautiful thing in terms of that common understanding and approaches.
And so that’s what I’d really like to see happen alongside — like you said, Madigan — AI literacy happening alongside safeguarding, protection, data management, etc., so that it’s just a core learning that people have these frameworks to draw on.
Madigan: I think Ka Man, you brought up such a good point about that shared learning. With Kaya Enterprise and being able to make that knowledge public — I think that’s such a huge boon for the humanitarian community. So I think that’s really exciting.
16:55: Chapter 4: Priorities for the second half of 2026: AI literacy, governance – and critical scaffolding
What do you think the priority should be for the second half of 2026? With the changes that AI are making, what are the things listeners can do or take away to discuss with their teams and organisations?
Ka Man: There’s so much people can do, but it’s easy to feel quite overwhelmed, because I know that as so many are racing ahead with AI, it can feel very daunting. And if you do feel overwhelmed, just know that that’s completely normal. Everybody working in this space can feel overwhelmed and does with the pace of change. It’s okay to be where you are, but I would really recommend that whether you’re using AI or not, you don’t intend to, AI literacy skills are really critical, because that underpins our overall and collective readiness and preparedness in this space. Humanitarians are familiar with anticipatory action, obviously, and I see AI literacy as part of that toolkit.
So take a course — including ours on Kaya. Talk to others. AI, as we found in the research, people can feel very guarded about talking about it — whether they use it, how much they use it — especially within organisations. Join whatever conversations you can, whether that’s just with a trusted peer or a small group of people, or join webinars. So the HLA regularly hosts humanitarian tech webinars with partners, including with NetHope and Data Friendly Space of course.
Finally, for organisations — governance is your number one priority, and literacy obviously goes hand-in-hand with that, as Madigan’s already talked about. One of the things I found very sobering was the lack of movement on organisational policies, and I completely understand the reasons for that, as organisational leaders are more-or-less in crisis management mode with all of the structural changes across the sector and the impact and implications for organisations. And in larger organisations, lots of teams have to be involved — legal, head office, all those levels and layers. But it really is absolutely critical that there is movement on this. Just 1% movement from last year to our survey was disheartening.
And the last one for organisations: take a look at the SAFE AI framework. It’s the first one that’s been released and developed for humanitarians specifically. And even if you’re right at the start of your journey — for example, deciding Copilot or whatever tool — even if you think “we’re not deploying large-scale AI systems” — please still take the time to read it. There’s lots of commentary and notes that go alongside the actual framework that are really interesting, and it introduces you to lots of key concepts, including “responsible refusal”, responsible refusal of AI. Please do take a look at that. It’s available on the CDAC Network website.
So those are the four things mentioned there, Madigan. How about you? What do you think people need to prioritise now?
Madigan: I think right now the most useful thing you can do is build enough literacy to critically think — like you said about that AI refusal. Do we even use AI? What does it look like if we integrate AI into our work? You don’t have to become a power user, just start asking questions, take one of the courses Ka Man has mentioned, and do that informal learning, even though it would be great if organisations also formalised that.
Related to that — both for organisations and for teams and individuals working across different organisations — is to create conditions where people can be honest about how they’re using AI. Something we saw throughout our research is this fear factor — worrying you’re going to get in trouble for using a tool, or judged for not using one. It kind of kills that learning. And so we need to develop a culture where AI use is visible and discussed and not hidden away.
We’re now starting to see — with the SAFE AI project — frameworks, toolkits and checklists being produced. But I also think we need to focus on how we take those policies and actually transform them into practice.
So one other thing that I think is a priority going forward. And maybe less about what gets written, more about what gets resourced. So this is kind of a call for donors as well: how do we actually resource this so that funders can see the need to treat convening and peer learning as a legitimate line item, and not just built into overhead?Organisations need to actively protect that space for those conversations, and be able to include it in budget lines — not just assume it will happen organically.
So those are my hopes and what I’d like to see the priorities shift to in the latter half of 2026. And I think we have a really exciting future ahead, now that we’re starting to shift into this new world of AI literacy and readiness and actually turning it into practice — but there comes risk along with it, and I think we also need to be well aware of those risks and know how to mitigate them as well.
Ka Man: Like you say, these high-level frameworks like SAFE AI — absolutely fantastic. It feels like there needs to be that structural support for top-down and bottom-up approaches to meet in the middle, to support the whole of the sector, the whole of the space, in what’s often called the messy middle, which is where the work is happening on a day-to-day basis. So we can’t do that alone.
Like I say, our work has been unfunded — this has been done alongside our business as usual in terms of this convening. And it was interesting — I recently joined, last week I joined a NetHope Elrha webinar and they’ve pointed to the same in their research. They found that the convening and those communities of practice really need to be operationalised now and mobilised to support the sector at large, to support those different tracks of people regardless of where they are on their AI journey. And that needs to be funded.
And just generally thinking about the sector at large — because of the hyper-prioritisation, I can understand why that might not feel like an immediate pressing need. But hopefully through this research, when you dive into it – and through the conversations happening in the space at large – AI is rewiring the system, whether we like it or not. So it’s absolutely critical that as we rebuild this humanitarian sector in the way that we want it to look, AI is going to be a part of it, whether you’re using it or not. It’s rewiring it. And I see it as — when you’re rebuilding a house, the wiring needs to happen. You can’t see it, but it’s there. And there needs to be those safety mechanisms, that earthing, that grounding to make the house safe for all of us. That’s how I see AI literacy and governance, and that convening piece really needs to be funded.
I follow someone called Gautam John on LinkedIn, and he used the phrase “relational substrate” — the relational substrate of the sector, how, as he’s watched and observed and been part of the system for a while, he’s seen that the relational substrate is not funded because it’s not visible, right, it’s not always highly visible. But that’s what makes the sector work, right — that’s what makes systems work: those relationships, that communication. So I guess that’s my plea to donors to say: while we’re building this house, let’s get that wiring in and safe, collectively, for all of us.
27:54: Chapter 5: Why every voice counts and keeping the human-in-the-loop
So thank you so much. We’ll wrap this conversation up here. It’s been an absolute pleasure to have this time to connect with you again today, Madigan. And I just want to say, just over a year ago, we recorded a podcast to invite people to take part in the first survey, and our message then was every voice counts, every voice matters. And I think the message is the same again in relation to this AI literacy and AI governance piece. So that’s my final sort of reflection to share. How about you, Madigan?
Madigan: I mean again, you took the words right out of my mouth there. I think the other thing that I would say is — the human-in-the-loop. AI is an incredibly useful tool when used ethically, when used responsibly. And I think that always means having a human with oversight over the AI outputs. So while we delve into more of the AI literacy part, I want to remind everyone that the AI tools that we use should be complementary to the work that we do. And so that comes with the human-in-the-loop component.
As we go forward, I’m really hoping to see more of that, and more shared and collective learnings among the sector.
[Music]
29:18: Chapter 6: Thanks to the community — and looking ahead to part two
A huge thanks to Madigan Johnson — and to everyone in the humanitarian community who is showing up and contributing their voice to this conversation. This collective effort is a testament to the power of collaboration — between the HLA and Data Friendly Space and our partners in this space, and with the wider community whose experiences are at the heart of this initiative. The challenge is now to draw on this body of evidence and our collective voices to unlock and mobilise broader structural support and resource for the sector.
Part 2 is due out shortly — a powerful conversation with Musaab Alhadi, Nour Arab and Ali Al Mokdad. Thank you for listening to Fresh Humanitarian Perspectives from the Humanitarian Leadership Academy.
