This is the second article in an AI4NGO series for civil society. Article 1 walked through the five-step path for adopting AI without an IT team. This article is about what happens next — the human skills that quietly become more valuable, not less, when AI arrives in your work.

Summer is one of the rare stretches of the year when the noise lowers enough that we can actually think about something meaningful after reloading a little energy on holiday.

One question is worth thinking about slowly. “AI” and “automation” are the buzzwords of the year — carrying a lot of fear and uncertainty with them — even as AI quietly becomes a normal part of how our teams work. So what is our place in this transformation? Which human skills become even more valuable, not less? That is what this article is about. It is written to be read in one sitting, and to leave you with a sketch of what your organisation could invest in when September turns the lights back on.

Two things we already know:

1. History repeats itself. Every new technology triggers the same fear: this one will finally replace us. It never has — it has moved us instead. The printing press ended the scribe’s role, but created the editor’s job. The calculator took arithmetic away from accountants and engineers, and gave them time for pondering harder questions. Each wave of new technology has shifted people from doing the task to deciding what the task should be. AI is the next such wave — faster than the ones before, but moving in the same direction.

2. The people building AI say the same. In Anthropic’s data, people collaborate with Claude more than they delegate to it (52% vs 45% of conversations); Microsoft finds 86% of users treat AI output as a starting point, not a final answer; and 83% of 2,000 CEOs told IBM that AI success depends more on people than on technology.

So the starting principle of this article is simple:
AI moves the work. Humans stay at the centre.

Now to the part nobody puts in the brochure.

When AI invents a fact — a report that was never written, a number nobody measured — we call it a hallucination, and it is the technology’s most embarrassing flaw. It happens because of how these models are built: they do not look facts up — they predict what a plausible answer should sound like, word by word.

Often, plausible and true overlap. But sometimes they don’t — and no setting fully switches this off. Nobody will fix this next year — the tool simply works this way. Which is one more reason every AI draft needs a human reader.

But here is the strange part: humans also see things that don’t exist yet. The difference is — we do it on purpose:

“A world where no child dies of a preventable disease.” “A city where nobody sleeps outside.”
— Every mission statement in civil society describes something that does not exist — yet. That is not naive language – that is the job. AI can summarise, remix and extend what already exists — drawing only on what it was trained on. It cannot want the world to be different.

The difference fits in one line: the machine hallucinates by accident and produces errors. Your people do it on purpose and produce missions. Keep that distinction in mind — the rest of the article is about protecting it.

The first trap:
your skills fade quietly

For an NGO, this matters twice as much. The moments where your judgment counts most — what a beneficiary really needs, whether a report tells the honest story, what a partner’s silence means — are exactly the moments where a shortcut costs the most.

None of this is a reason to reject AI. Use it with your eyes open — and treat every hour it saves as an hour to reinvest in the work only your people can do.

The second trap:
everyone starts to sound the same

Now, picture a programme officer at a foundation reading two hundred grant proposals, most drafted with the same three AI models — all fluent, all well-structured, all remarkably similar. Funders say they already notice the similarity. In that pile, polish is worthless. What stands out is specificity: the proposal that sounds like a real place and real people — the thing the community has said – actual people, real stories. Your organisation’s unique voice is becoming more important than ever. 

The third trap has a name:
workslop

Stanford and BetterUp researchers recently named a phenomenon most of us have been on the receiving end of: workslop — AI-generated work that looks good on the surface, but has no actual substance beneath. Their findings reported:

• 40% of employees received workslop in the past month.
• It damages the sender most — colleagues rated people who sent workslop as clearly less creative, less capable, and less trustworthy.

Read that as an NGO. A company that produces workslop loses hours. An organisation that lives on donor confidence and community trust loses the only currency it has. The report that “reads fine” but says nothing of significance, the AI-polished update a funder can “smell” from the first paragraph — those are withdrawals from the trust account. 

This is not just a moral position. The EU AI Act (Article 14) makes human oversight of high-risk AI systems a legal requirement from August 2026 — and both the NIST AI Risk Management Framework and ISO/IEC 42001 are built on the same principle. Every AI architecture should assume every system has a named human accountable for it. The person, not the model, carries the weight.

For an NGO, this is more than a question of compliance. It is a mission-alignment guarantee:

• The colleague who reads an AI-drafted grant narrative before it goes to the funder.
• The programme worker who checks whether an AI summary of beneficiary feedback is honest enough to act on.
• The communications officer who decides whether a community would recognise itself in the line AI wrote about it.

These are the moments where your organisation stays accountable to the people you serve — and they happen at every level of the team.

The person is not there just to supervise the machine. The person is there to keep the mission in the room.

The five human skills that quietly become more valuable

If AI takes on the mechanical layer of knowledge work, five human skills grow in weight. They are not new. They are the ones your NGO already values — close cousins of the analytical thinking, curiosity and collaboration the World Economic Forum’s Future of Jobs Report 2025 ranks among the fastest-growing skills in importance for the next five years.

1. Knowing what actually happened

AI knows what has been written down. Most of your organisation’s knowledge was never written down anywhere: what the community actually said in the meeting, why the 2019 project failed, which partner delivers late but always delivers. That ground truth is data no model has — and collecting it (the field visit, the listening session, the honest field note) is becoming your biggest information advantage. The model guesses. Your people were there.

2. Asking the right question

AI makes answers cheap — which makes questions expensive. Ten fluent drafts of the wrong proposal are worth less than one honest paragraph of the right one. Framing the problem before the prompt — is this actually a fundraising problem, or a programme design problem? — is the step AI cannot do for you.

3. Checking the source before it leaves the building

Quality control of AI output is the top skill gaining importance in Microsoft’s data. For an NGO, it is existential: one hallucinated number in a donor report can cost a funding relationship built over ten years. So make source-checking a reflex: where did this number come from? Is it from our own approved data — or did the model invent something plausible? If a claim cannot be traced to a source you trust, it does not leave the building.

4. Carrying responsibility

Every AI use case needs a named human who answers for it — legally (the EU AI Act) and morally. Stopping a use case that is technically working but ethically drifting. Telling a funder about a mistake. Deciding whose story does not get told because the person did not consent. A model can generate options. It cannot carry the weight of choosing one.

5. Building trust between worlds

NGOs sit between donors, regulators, partners and communities that do not speak each other’s language — literally and figuratively. Sensing that a partner went quiet because they are hurt, and not just busy. Translating a community’s concern into a board decision. Keeping a relationship alive through a challenging year. Trust moves at the speed of relationships — and no model attends the meeting.

Notice a pattern: these are the skills the sector has always prized. NGOs never valued their people for being fast at arithmetic. They valued them for their judgment, sense of purpose, and ability to sustain meaningful relationships. AI does not threaten that — it makes it more valuable. In the end, that is the human job description: to give meaning to information that has none on its own, consciously. 

And to say it plainly: intuition, emotion, creativity, empathy — AI does not have them, and is not close. That is your terrain. Investing in it is not an argument against AI. Grow both, deliberately — human depth and AI skills. You don’t have to choose — growing both is exactly what makes you more powerful in the work.

The commercial sector is spending billions to rediscover something the mission-driven sector never forgot: the reason for doing the work matters more than the pace of doing it. Every NGO has already answered questions many corporations are only beginning to ask — what human dignity is at stake, what red lines we will not cross for efficiency, who is not represented in these numbers. An NGO that adopts AI thoughtfully is not catching up. It is showing the sector how to do it with the person, the mission, and the community at the centre.

And the sector’s oldest habit — describing a world that does not yet exist, and then going out to build it — is exactly the capacity the machine lacks.

Practical action —
one exercise per skill

You don´t have to start with building the whole training program. Pick these five small exercises — one per skill — and put them in the calendar. The same discipline that made Article 1’s five-step method work applies here: small, named, practical.

• 1 → Write down what the room said. After every field visit, partner call or community meeting, take ten minutes to capture what was actually said — before your AI summarises anything. That growing file is your organisation’s approved data: the ground truth you check every AI output against, and the source of the specific detail no model could invent.
• 2 → Start with the question, not the prompt. Before anyone opens a chatbot for a serious task, write one sentence: what are we actually trying to find out, and for whom? If the team cannot agree on the sentence, that conversation was the real work — and no tool could have done it.
• 3 → Give every number a passport. Any statistic that leaves the organisation gets a source next to it — your own data or a named external one. If nobody can say where a number came from, it does not go out. Name the person who checks; make it a role, and not optional. 
• 4 → Keep a one-page AI register. Once a quarter, update a single page: what we use AI for, who owns each use case, and what we have decided not to automate. A use case with no name next to it is your first fix. The red-line list is your moral courage, written down before you need it.
• 5 → Reinvest saved hours in people. For every hour AI saves, put a slice back into the human side: the partner call that wasn’t urgent, the 1:1 where you ask each person what they want to learn, the field visit that kept slipping. And once a quarter, do one piece of work with no AI at all — you will learn more about your team’s real strengths from that session than from any external report.

And one for the organisation’s voice: once a season, take one text you have published and ask — could any organisation have written this? If yes, rewrite it until only you could have.

These exercises are cheap, culturally light, and compatible with the frameworks your organisation may already be using — the Prosci ADKAR change-management model and the Prosci PCT framework both treat individual capability-building as a first-class investment, not an afterthought.

In closing

The people who will do best in this decade are the ones who neither fear AI nor worship it — the ones who invest deliberately in the human skills that have always mattered most, and at the same time learn to use the new tool well.

And through all the noise of the coming year, keep the one distinction from the start of this article: the machine hallucinates by accident; your people imagine on purpose. One produces errors. The other produces missions. That difference is the sector’s entire advantage — and no upgrade will close it.

For NGOs, this is home ground. You already know it. September will bring the strategy conversations, the funder calls, the team meetings and the new colleagues. If any part of this article stayed with you into the shade of a tree or the corner of a café, treat it as the first pencil sketch of what you — and the people around you — could invest in when the season turns.

If your organisation is beginning to think about which human skills to invest in as AI enters your work, three modular offers from our portfolio map to that path — each grounded in Prosci’s certified change-management methodology and adapted for civil-society pace and capacity:

• P2 Leadership Workshop — a facilitated strategy and governance session for your leadership team, designed to align on where AI belongs in your organisation and where the human decision must stay in the room.
• P3 AI Literacy for All Staff — a self-paced, plain-language literacy programme that builds a shared vocabulary and confidence across the whole team, not just the digital champions.
• G4 Steering & Ownership — a light governance model that names owners, defines review rhythms, and makes human accountability visible for every AI use case in the organisation.

Each one stands alone. Combine over time. Grow as confidence and capacity grow.

If your NGO has one question about AI and your people you have not been able to answer — should we invest more in staff development or in tools, how do we prepare our team for what is coming, what does responsible AI adoption actually feel like from the inside — summer is a good time to open that conversation, so autumn can be the time to act on it.

Reach out to us at info@ai4ngo.org or visit advisory.