Can you see the future yet?
Well, as you might have guessed, AI is already being rolled out across the human and social sector. So, you might be curious about the role of AI in this new age. NGOs are already leveraging AI tools to help sift through data, analyze data, increase their fundraising, translate content, improve communication, optimize logistics, and identify trends that would take teams hours and even days to identify.
On first reading this could just appear to be a story about tech speeding up NGO work. But there’s a lot more going on here.
Use of AI in an NGO is not just an upgrade to software. It’s a piece of technology capable of making decisions that affect real human lives. An automated recommendation can alter the allocation of funds, the design of a project, the inclusion or exclusion of at-risk communities, and even managing private data.
So what’s the real question? It’s not: Can the AI do a thing? It’s: Should the AI do a thing?
How much power should it wield?
And who’s responsible when it swings that way?
Meet ethical AI governance. For NGO advocates, it’s more than “mother tongue” for non-specialists; it’s about human oversight, transparency, privacy, accountability, fairness, community engagement, and local knowledge and expertise. UNESCO’s global non-binding recommendation on AI ethics emphasizes this and discusses human rights and dignity, citing the importance of principles such as transparency, fairness, privacy, accountability, sustainability, and inclusion. It also makes clear that AI should never substitute for ultimate human responsibility.
This is more important for an industry that’s centered on people and communities. What are your thoughts? Can AI truly stand alongside humans in this kind of important work?
AI is changing how NGOs work, but it should not change why they work.
NGOs are under constant pressure to do more with limited resources. Teams may be managing several projects at the same time, working with complicated donor requirements, collecting monitoring data, preparing reports, raising funds, and responding to community needs.
There are many tasks within this work that are repetitive.
Staff may spend hours cleaning spreadsheets, sorting documents, translating material, preparing meeting notes, reviewing large amounts of information, or searching through donor requirements. These are the kinds of tasks where AI can be genuinely useful.
It can reduce some of the workload and help teams work more efficiently.
But the core work of an NGO is different.
NGOs exist to serve people, strengthen communities, respond to crises, protect rights, and address problems that are often shaped by poverty, inequality, conflict, culture, history, and local circumstances.
Those things cannot always be understood by analyzing data alone.
A system can tell an organization that a particular community has a high level of vulnerability. It cannot necessarily tell the organization why that vulnerability exists or what kind of response the community itself would consider appropriate.
That distinction is at the heart of ethical AI.
AI can support the work. It should not quietly take ownership of the work.
Human oversight is not a box to tick.
One of the most important principles of ethical AI is human oversight.
The phrase sounds simple, but meaningful human oversight requires more than having a staff member review a computer-generated result.
The person reviewing the result needs to understand what the system is doing. They need to know that AI can be wrong. They need enough time to investigate unusual results. They need access to additional information when the data looks incomplete. Most importantly, they need to have the authority to reject what the system suggests.
Otherwise, “human oversight” can become a formal step without much real influence.
The World Food Programme provides a useful example of how this can work. WFP has used AI to identify possible duplicate records within food-assistance programs. The technology can help flag potential matches, but staff review the results and make the final determination rather than allowing the AI system to decide independently.
That model is worth paying attention to.
AI is good at processing large quantities of information quickly.
People are better placed to understand unusual circumstances, question assumptions, and consider consequences.
The strongest approach is therefore not to ask which one should replace the other.
It is to ask how they can work together.
Why local context matters more than an algorithm may suggest
There is a simple problem with relying too heavily on technology: data can describe a situation without fully explaining it.
Imagine that an AI system identifies a rural community as being at high risk of food insecurity. The model may be looking at rainfall, crop production, food prices, household data, and other indicators.
All of that information could be useful.
But a local NGO might know that the area’s main road has recently become inaccessible because of insecurity. It might know that a local market has closed. It might know that some families have lost access to a water source or that women in the community are experiencing barriers to receiving assistance.
Those details may not appear in the dataset.
They may also completely change how the organization should respond.
This is why local context matters so much in development and humanitarian work.
A model can provide a broad view.
A local team can provide the context needed to understand that view.
This relationship becomes particularly important when working in places where data is limited, outdated, or collected unevenly. A community that is poorly represented in digital systems can easily become a community that is poorly represented in algorithmic decisions.
That can create a dangerous cycle in which the people who are hardest to measure become the people easiest to overlook.
The risk of making invisible communities even more invisible
AI depends on data.
That creates opportunities, but it also creates a serious risk.
Some communities are easier to capture in datasets than others. People with formal identification, stable addresses, reliable internet access, and regular contact with public services may be easier to represent.
People who move frequently, live in remote areas, lack documentation, experience displacement, or have limited digital access can be much harder to capture accurately.
When this happens, AI systems may work better for the people who are already highly visible in existing systems.
That creates an important question for NGOs: Who is missing from the data?
It is a question that should be asked before an AI system is deployed, not after it causes a problem.
Missing data should not automatically be interpreted as low need. In some humanitarian contexts, the absence of reliable information can itself point to vulnerability.
For example, displaced populations may have incomplete records, different spellings or language variations in registration information, or records spread across systems that do not communicate easily with each other.
AI can help organizations identify patterns in this information, but it cannot solve the underlying problem of incomplete representation on its own.
Sometimes the solution is not more automation.
Sometimes it is better data collection, stronger local partnerships, and simply spending more time listening to communities.
Transparency should be understandable to ordinary people.
When organizations talk about responsible AI, transparency often appears near the top of the list.
But transparency should not mean only publishing a document full of technical language.
For an NGO, meaningful transparency starts with being able to clearly explain why a system is being used.
People should know, where appropriate, that AI is involved and understand the role it plays.
Staff should understand what kind of information is being used and how the system’s output should be interpreted.
And when a decision affects a person directly, there should be some way for that person or the staff member responsible to question the result.
Imagine that someone is told that an automated system has helped determine whether they qualify for assistance.
A reasonable response would be: What information was used? How was the decision made? Did a person review it? What happens if the information is incorrect?
Those are accountability questions.
People are much more likely to trust systems when they know that decisions can be questioned than when they are simply told that “the system decided.”
UNESCO’s AI ethics framework specifically links transparency and explainability with responsibility and accountability while recognizing that transparency must also respect privacy and security.
For NGOs, that balance is particularly important because explaining a system should never mean exposing the personal information of vulnerable people.
Protecting humanitarian data is not optional.
Data protection is even more important when NGOs are working with vulnerable groups.
It is possible that a humanitarian organization is managing data pertaining to information about refugees, children, those affected by violence, individuals requesting protection, health status, familial ties, and place of residence, among other sensitive areas.
This information can have consequences beyond privacy.
There are certainly situations where revealing someone’s name or location could lead to a safety issue.
Put simply, it means that NGOs will need to be mindful of how access to and use of personal information will happen with the AI.
Before engaging an AI tool, be aware of what data your organization is sharing. Find out where data is stored, who has access, if it can be reused, and how long it is held for.
This is especially true for generative AI tools available to the public.
An employee might reason, “I can just upload this report and have the AI generate a summary.”
The summary may take seconds.
However, if it contains information such as names, case details, health information, or other sensitive data, the effort to protect it could pose risks that far outweigh the value of the process saved.
That is exactly why responsible AI governance cannot just involve organizational policy at the top; it must involve day-to-day staff conduct as well.
UNESCO includes privacy and data protection as a fundamental concept in its approach to the AI lifecycle.
The problem with AI that sounds confident when it is wrong
Another challenge for NGOs is the rise of generative AI.
Modern AI systems can produce very polished text. They can write reports, summaries, funding proposals, social media posts and research-style explanations in seconds.
The problem is that confident language does not guarantee accurate information.
AI can make mistakes. It can misunderstand a question, rely on incomplete information, create an inaccurate summary or generate a citation that does not exist.
For NGOs, these errors can have real consequences.
A wrong number in a donor report can damage credibility. A misleading statement in a humanitarian communication can create confusion. A poorly researched funding proposal can misrepresent what the organization actually does.
Human review therefore remains essential.
The fact that a piece of content was created with AI does not remove responsibility from the organization.
An NGO is still responsible for the information it publishes, sends to donors, shares with communities or uses to make decisions.
AI can help create the first draft.
It should not automatically receive the final word.
Authenticity matters too
There is another side of AI ethics that deserves more attention: authenticity.
NGOs often tell stories about real people. Those stories can help donors understand the impact of a programme and help the public understand an issue.
AI can make storytelling faster and more polished.
But there is a line between improving communication and changing someone’s story into something more dramatic because it is likely to perform better.
A real person’s experience should not become fictional simply because an algorithm thinks fictional language will attract more attention.
The same applies to images, quotations and beneficiary stories.
The pressure to communicate impact is real. NGOs need strong fundraising materials and engaging campaigns.
But engagement should not come at the cost of truth.
Ethical communication means protecting consent, dignity and context.
The people represented in NGO communications should remain people, not simply become data points or marketing material.
AI should build trust, not quietly damage it
Trust can take years to build and very little time to lose.
Communities trust NGOs when they believe that the organization will listen to them, protect their information and act honestly.
Donors trust organizations when reporting is accurate.
Partners trust organizations when they believe that decisions are based on evidence and sound judgement.
AI should support these relationships.
That means NGOs need to be clear about why a particular system is being used.
There is a major difference between using AI because “everyone else is doing it” and using AI because it solves a specific problem and the organization has considered the risks.
The second approach is much healthier.
It puts purpose before technology.
It also helps prevent one of the most common mistakes in digital transformation: adopting a tool simply because it is available.
Not every NGO problem needs AI
One of the most important signs of mature AI governance is knowing when not to use AI.
This can be surprisingly difficult at a time when AI is being promoted as a solution to almost everything.
An NGO might be better served by improving its database, training staff, fixing an outdated workflow or improving communication between departments.
Adding AI to a broken process does not automatically fix it.
Sometimes it only makes the process harder to understand.
UNESCO’s ethical framework includes proportionality and a “do no harm” approach, emphasizing that AI use should be necessary and appropriate to the objective rather than introduced without considering risks.
For NGOs, a useful starting question is therefore not:
“Where can we use AI?”
It is:
“What problem are we trying to solve, and is AI actually the best way to solve it?”
That small change can make a major difference.
Ethical AI needs more than an IT department
Ever assume that AI governance should be a task for your tech team alone? No way. You shouldn’t leave this to the sole discretion of the tech team, and here’s the reason.
You, as someone who implements program, are essential to these discussions because you are in the field and experience how work is actually done in the real world. For those in communications, you need to understand what genuine content is and how to use it ethically. For those in finance or procurement, you’re stuck with contracts, subscriptions, and out-of-control tech.
Alright, so how does senior leadership get involved? They’ve got to identify the ‘who’s’ of accountability, and monitoring and evaluation need to do the ‘is’ of performance, to see if the technology is having an effect. And hang on community, you should have a voice when technology really affects you!
UNESCO advocates for a multi-stakeholder model for inclusive governance of AI, urging the involvement of civil society. But why? Because AI is not an “IT matter” but a people matter; an issue of programmes, of governance, and sometimes – of human rights.
Staff need AI literacy, not just AI tools
Giving staff access to AI without training them on responsible use can create unnecessary problems.
Employees should know the basic limitations of AI. They should understand that AI-generated content may contain errors, that data can create bias and that sensitive information should not automatically be uploaded into every AI platform.
They should also know when human verification is required.
For example, an organization might decide that all AI-generated donor statistics must be checked before publication. It might require staff to verify research claims against original sources. It might prohibit sensitive beneficiary information from being entered into external AI systems.
These rules do not need to be complicated.
What matters is that employees understand them.
AI literacy is increasingly becoming part of organizational accountability because a tool can only be used responsibly when the people operating it understand its limitations.
UNESCO also identifies awareness and literacy as one of the core principles of ethical AI governance.
What funders need to understand about responsible AI
The responsibility does not belong only to NGOs.
Donors and foundations also have an important role.
Funders increasingly ask organizations to demonstrate innovation and digital capability. That can be positive, but responsible technology costs more than a software licence.
An NGO may need staff training, stronger cybersecurity, privacy assessments, technical support, monitoring, evaluation and community engagement.
These things can be less exciting to fund than the launch of a new AI platform, but they are essential to using technology safely.
A funding model that supports only the technology may encourage organizations to launch pilots without investing in the systems needed to maintain them responsibly.
A better approach is to fund the wider ecosystem around technology.
That means supporting people as well as platforms.
It means funding training as well as implementation.
And it means allowing NGOs to learn from mistakes rather than expecting every digital project to succeed immediately.
This could be particularly important for smaller and local NGOs, which may have strong community relationships but less access to technical expertise
.
The Global South needs a voice in AI governance
The conversation becomes even bigger when we look at AI from a global perspective.
Many advanced AI systems are designed and developed by organizations in wealthier countries. Their datasets, assumptions and priorities may not always reflect communities in Africa, Asia, Latin America, the Caribbean or the Pacific.
That creates a risk that organizations in the Global South become primarily consumers of technology rather than participants in shaping it.
The question of representation is therefore critical.
- Who decides what information matters?
- Who decides what a successful outcome looks like?
- Who determines acceptable levels of risk?
- Who owns the data?
- Who has the right to challenge an automated decision?
- Who is able to change the system when communities say it is not working?
- And who ultimately benefits from the technology?
These are governance questions, not merely technical ones.
UNESCO’s current AI ethics work emphasizes diversity, inclusion, civil-society participation and multi-stakeholder governance, reflecting the need for AI systems to be shaped by more than a small group of technology developers or powerful institutions.
For NGOs, this creates an opportunity.
Local organizations can bring knowledge that large technology systems may not have.
They can help test tools in real settings.
They can identify cultural and social problems that might otherwise be overlooked.
And they can make sure that communities have a voice in decisions that affect them.
Localization and ethical AI belong together
The humanitarian sector has increasingly emphasized localization and the importance of strengthening local and national organizations.
AI should support that direction rather than reverse it.
A centralized AI system controlled by a distant institution could make decisions even more removed from communities.
But technology can also help local NGOs access information, strengthen monitoring, improve coordination and participate more effectively in wider humanitarian systems.
The key question is who controls the technology.
A local organization should not simply be treated as the place where a technology is tested.
It should have a voice in determining whether the technology is appropriate.
It should be able to share concerns.
It should be able to influence the system’s design.
And communities should have a meaningful opportunity to say when something is not working for them.
That would turn localization into something broader than funding and implementation.
It would make local participation part of technology governance itself.
AI ethics is also about the environment
There is another issue NGOs should not ignore: the environmental impact of AI.
AI systems depend on computing infrastructure, electricity and physical resources. As organizations begin using AI more extensively, the environmental cost of digital infrastructure becomes part of the ethical discussion.
This matters especially for NGOs working on climate change, environmental protection and resilience.
A technology solution cannot automatically be considered responsible simply because it produces a social benefit.
Organizations should also think about efficiency, necessity and environmental impact.
UNESCO includes environmental sustainability among the core values of its global AI ethics framework and calls for AI technologies to be assessed for their sustainability impacts.
The lesson is not that NGOs should avoid AI.
It is that responsible innovation means thinking about the full life of the technology.
Success should be measured by impact, not by AI adoption
There is a tendency to measure digital transformation by asking whether an organization has introduced a new tool.
But launching AI is not an impact.
Having a chatbot is not an impact.
Generating thousands of automated predictions is not an impact.
The better question is what changed because of the technology.
- Did staff save time?
- Did that time allow them to work more closely with communities?
- Did the system reduce mistakes?
- Did services reach people faster?
- Did the organization identify a humanitarian risk earlier?
- Did it make information more accessible?
- Did it reduce costs?
- Did it improve decision-making?
Those are the outcomes that matter.
WFP’s broader AI strategy describes the purpose of AI and machine learning in terms of optimizing resources and accelerating frontline response rather than adopting technology simply for its own sake.
That distinction is important for every NGO.
The goal should never be to say, “We use AI.”
The goal should be to say, “We used technology to improve something that matters to the people we serve.”
What responsible AI could look like inside an NGO
For a small or medium-sized NGO, ethical AI governance does not have to begin with an expensive consultancy or a complicated technical framework.
It can start with a simple set of organizational principles.
The NGO can identify what AI is being used for and why.
- It can decide which types of information are safe to use with external AI tools and which must remain protected.
- It can establish which decisions always require human approval.
- It can train staff on common AI risks.
- It can create a process for reporting errors and unexpected behaviour.
- It can ask communities for feedback when technology directly affects them.
- It can regularly review whether the system is actually delivering value.
And it can stop using a tool when the risks become greater than the benefits.
This last point matters.
Governance is not something that happens once when a tool is approved.
AI systems are updated. Vendors change their policies. Models change. New risks emerge. Organizations change their priorities.
Responsible governance therefore needs to continue throughout the life of the technology.
The difficult question: who is responsible when AI gets it wrong?
Are you ready to get down with AI for NGOs?
As new technology gets adopted, you’ll find yourself facing difficult accountability questions. What if an NGO’s AI tells you a family doesn’t need aid because of a low priority level.
When that goes wrong, who’s responsible?
Is it the company that created the software, the NGO that used it, the staffer who evaluated the recommendation, the data provider, or the donor supporting the program?
They’re not simple ones, are they?
And here’s the thing: you can’t offload responsibility onto an algorithm. An AI system isn’t going to sit in a community consultation to take responsibility, or say sorry to those affected.
It’s human beings and organizations who need to take this on board.
That’s where UNESCO’s five features come in: responsibility, accountability, auditability, traceability and human oversight. Do you think those will be even more important as the technology gets more advanced?
The future of AI in humanitarian work needs both innovation and restraint
Feeling the heat of speed to respond? So are we. Humanitarian front line staff like you have the dilemma of a rapidly changing situation, sifting through information from thousands of sources and at the same time the needs growing faster than you can fundraise. No wonder your staff is overwhelmed with information overload.
Yeah but AI might Save you. To pull this off, the World Food Programme (WFP) is creating a global AI strategy built on AI and machine learning to boost efficiency and speed up front-line operations. Right up our alley isn’t it? Not so fast-you should go smarter, not faster.
Sure, an AI can produce an answer within seconds. But what if it was wrong? A model can go through millions of records but if those records are based on inaccurate or biased data, a model just ends up amplifying those inaccuracies.
So, how do you balance the mix? Gutsy – be brave on technology when you need to, wise back-off, review & remove the wrong ideas.
Conclusion: Good AI governance keeps people at the centre
AI is going to remain part of the NGO and humanitarian conversation.
The technology can bring real benefits. It can help teams handle information, reduce repetitive work, improve operational efficiency and potentially respond to crises more effectively.
But responsible adoption depends on much more than technical capability.
NGOs need systems that respect privacy.
- They need people who can question AI-generated recommendations.
- They need transparent processes.
- They need staff who understand the limitations of the technology.
- They need communities to have a voice.
- They need local knowledge to remain part of decision-making.
And they need clear accountability when something goes wrong.
UNESCO’s global approach makes human rights, dignity, transparency, fairness, sustainability, privacy and human oversight central to ethical AI governance.
For NGOs, the most important lesson may be surprisingly simple:
The goal is not to use more AI. The goal is to use AI better.
Technology should help organizations become more effective without making them less connected to the people they serve.
AI can process data. It can identify patterns. It can automate repetitive work. It can help teams make sense of complicated information.
But it cannot replace community trust.
It cannot fully understand lived experience.
It cannot take responsibility for the consequences of a decision.
And it should never become an excuse for organizations to stop asking questions.
The future of ethical AI in the NGO sector will therefore not be about choosing between people and technology.
It will be about finding the right balance between the two.
Let AI do what it does well. Let people do what only people can do. And make sure that, throughout the process, the communities at the centre of NGO work never disappear behind the technology.
