The challenge for not-for-profits in an era of outsourced cognition

At 1.17am, someone searches for help. Not strategically, as a customer on a journey, or progress through the proverbial fundraising funnel, because human panic rarely respects a diagram. Just a person in bed, screen turned right down, heart doing that small animal thing it does when something is wrong.

They type something simple:

mental health support near me 

how to help my mum leave 

best charity to donate to 

what actually helps after a disaster

That is not search behaviour. It is exposure. A small gap between fear and action, between not knowing and needing something to hold, between the private ache of being a person and the public systems we still half-hope might know what to do with us.

For anyone who works in a not-for-profit sector, this is the moment that matters most. Not because the person wants information. Because for a few seconds, they are reachable.

And that is the part AI is quietly changing. Not by making search worse, but by making it too good. Too fast, too clean, too resolved. Because the answer arrives before the question has had time to do anything to the person asking it.

The new layer between need and care

Most people watched this year’s Google I/O 2026 and saw products. AI Mode, Gemini Live, search agents, Project Astra, ambient assistants, checkout that happens on your behalf. Useful, fast, frictionless, the kind of thing Silicon Valley unveils with a calm smile while quietly rearranging the furniture of how we behave.

But underneath the product theatre was something larger. Google is not just shipping features. It is rebuilding the layer we reach through whenever we do not know something and have to decide who to believe. Tools we no longer use so much as think through.

There is a name for this now. Steven Shaw and Gideon Nave call it System 3 thinking. For decades, marketers and behavioural economists have borrowed the familiar model of System 1 and System 2, the fast, emotional, instinctive mode and the slow, reflective, deliberate one. System 3 is neither. Shaw and Nave argue that AI introduces something else entirely: artificial cognition that operates outside the brain, runs alongside, and more and more, supplementing or replacing internal human reasoning. Not instinctive or reflective. External. A layer of outsourced cognition sitting between humans and reality itself. This may sound too academic until you see where it’s showing up. Not just in shopping and search, but in the small frightened hours, the moments people used to sit inside long enough to work out what they believed, who they trusted, what they owed somebody else.

That is the shift that matters. Because AI does not only compress information, it compresses hesitation. And hesitation is where a surprising amount of humanity lives. 

When care becomes a recommendation

Not-for-profits do not simply compete for clicks alone. They are after something slower and harder to win, moral attention. Empathy. Trust. Participation. Belief. Human proximity.

A domestic violence service is not a category. A refugee charity is not a verified source. A crisis line is not a knowledge panel. These are organisations that live in difficult human places, the ones made of fear and shame and grief and hope, the things nobody types cleanly into a search bar.

For a long time, someone looking to support a cause had to do some of the work themselves. The internet made people move through some of that friction before they reached resolution. They searched and explored and encountered stories, faces, language, mess. Sometimes they close the tab. Sometimes they cry. Sometimes they donate twenty dollars they couldn’t really spare, because another human being had suddenly stopped feeling abstract. Trust came gradually. And that process mattered because it created proximity, not efficiency.

System 3 changes the shape of that encounter. The machine filters, summarises, prioritises, and recommends the conclusion, faster, cleaner and more finished.  And the user receives resolution instead of exploration.

Imagine someone asking: Which charity should I donate to? Who actually helps? Is this fundraiser legitimate? What causes matter most right now? Which organisation is most effective?

The answer comes back at once, with none of the wandering through stories and faces and slightly-too-earnest copy that used to do something to a human on the way. At that point a not-for-profit is no longer competing to be found. It is competing to be the one the machine trusts enough to name. And that is a different game. A colder one.

There is a strange, undramatic sadness in that, the quiet flattening that happens when something complicated gets turned into something easy.

A woman on the train home from the hospital does not need a page of healthcare advocacy results. She needs to feel less alone for five minutes. The teenager looking for help at 2am is not running a query, they are reaching out from inside a panic. And the man searching for food relief the week after he loses his job is not moving down a funnel, he is trying to survive the embarrassment of needing it.

These are not moments that should become frictionless too quickly. Because friction is where humanity enters the room. And meaning often lives inside the friction.

Although I should be honest about the other side, because not all friction is sacred. That same teenager at 2am is better off with a fast, accurate answer than a wall of dead links, and the same machine that flattens all this might be the thing that finally surfaces some tiny, underfunded service nobody could find before. Friction is not a virtue. Some of it is just an obstacle. The skill is telling the two apart, knowing which delay is doing something to a person and which is only wasting their time, and getting rid of the second without losing the first.

Here is the part the sector has barely begun to say out loud. The machine may turn out to be better at seeming to care than the organisations that actually do. It will learn the language of care fluently, the warm tone, the patience, the perfectly judged reassurance, the reply that lands in a second at three in the morning and never once sounds tired. Set that against a real service running on a skeleton team, an ageing database, and funding held together with spiritual duct tape, and the machine simply feels easier. No waiting, or awkwardness, or a sense that you might be a burden.

The deeper risk is not that AI makes humans care less. It makes caring feel complete too early. A recommendation replaces exploration. A summary replaces reflection. A neatly packaged answer creates the emotional illusion of understanding without requiring the messier labour of staying present long enough to understand anything.

The user feels resolved. Which is not quite the same as being changed. Digital culture is already full of these small simulations of moral completion. Sharing instead of helping. Reacting instead of participating. Knowing about a problem instead of standing inside it. And AI risks industrialising that feeling. Not maliciously,  just very efficiently. Which is almost worse because nobody wakes up trying to become emotionally disconnected.

And we take the simulation more readily than we like to admit, as long as it makes the loneliness recede quickly enough. Half of us already tell an algorithm we are sad by playing the same song seventeen times. So the danger is not that people stop caring. It is that they start to mistake feeling better for being held by someone. The chatbot soothes on demand, the engine points to the “best” cause, everything works and everything resolves, and the thing that quietly goes missing is the oldest one of all: being seen by another person who is also tired and flawed and somehow moved by the fact that you exist.

Humans are tired, overstimulated, and economically strained. We are trying to survive our own lives, while carrying a low-level awareness that the world is, broadly speaking, on fire. And AI arrives offering relief from cognitive overload. So of course, we accept it. Because let’s face it, we do not want infinite complexity. We want reassurance that someone or something competent is handling it. And this is what makes the future so psychologically seductive. The machine begins to absorb not only informational labour, but emotional labour too. And eventually it may even begin softening reality on our behalf. Less ambiguity, contradiction and exposure means a smoother emotional experience. A more manageable world. And who wouldn’t want that?

But many social issues are not manageable.  Poverty is not neat. Trauma is not neat. Displacement is not neat. Loneliness is not neat. Grief is not neat. And the danger of AI-mediated systems is that they may become exceptionally good at converting deeply human realities into emotionally manageable interfaces. Because optimisation tends to drift toward comfort, speed, and resolution. I don’t know about you but I have never once watched someone be moved by an optimisation.

Relational invisibility

Many organisations will measure the obvious things first. Traffic down, clicks down, fewer sessions, a thinner month for donations. All real, none of it the actual problem.

The actual problem is much quieter. The bigger risk is relational invisibility. Your research may still train the system. Your organisation may still be technically present. And yet the human relationship may never form. The machine keeps the interaction. The platform keeps the trust transfer. You become informationally useful but emotionally distant. Part of the infrastructure rather than part of anyone’s memory.

For purpose-led organisations this is dangerous because emotional connection is not a side effect of the work. It is the work. And real care has weight. It costs attention, time, and energy. That is why people remember the volunteer who sat with them after everyone else left. Or the crisis worker whose voice stayed calm at 2.11am. Or the advocate who remembered their name three months later.

Because human care leaves fingerprints but AI leaves completion.

None of this is an argument against technology or AI. That would be boring, and frankly a bit too late. AI can remove barriers, surface help, translate needs, find services, reduce shame, and get someone to support faster than the old internet ever could.

But if AI becomes the layer through which people increasingly reach through for help or support, then organisations built around care will have to protect something more fragile than visibility.  They will have to protect the emotional conditions under which care becomes possible.

Not just information or even attention. Emotional residue. The feeling that follows you into the shower after reading something you cannot quite shake. The silence after hearing someone describe grief properly. The uncomfortable recognition that another person’s life is connected to yours in ways that are difficult to optimise or explain.

This may be where purpose-led organisations become strangely important again. Because real care is repetitive. It’s slow, emotionally inconsistent, difficult to measure cleanly, and sometimes irrational, exhausting, and unresolved for years. Which makes it almost perfectly incompatible with platform logic and completely essential to being human.

What not-for-profits need to protect

Build a point of view, not just a content library. Generic information is endlessly compressible; distinctive meaning survives longer. Years of service information, campaign prompts and impact statistics still matter, but in a System 3 environment you need a clear view of the world. What do you believe? What do you see that others miss? What language belongs unmistakably to you? If your voice is indistinct, your meaning becomes portable, and someone else’s interface can carry it without you. Sameness is not neutral here. It means disappearance.

Make trust visible before the ask. Trust cannot live only on the donation page. It has to exist in the public record, the evidence base, the partnerships, the lived experience, the consistency over time, in how you behave when you are not campaigning. System 3 infers credibility from the whole environment around you, not just from what you say about yourself.

Own the relationship beyond the platform. Every platform eventually optimises for itself; Google is simply being more explicit about it. The organisations that survive will own direct infrastructure: communities, memberships, events, newsletters, volunteer networks, local relationships. The machine may introduce someone to you, but it should not be the only place the relationship exists.

Protect human friction. A donation form should be easy. A story about systemic harm should not be reduced to a convenient interaction. Some experiences need slowness. Some decisions should resist instant resolution. In machine-mediated environments, humanity itself may become the advantage. The future will not belong only to organisations that make everything easier. It may belong to those who know which moments are supposed to leave a mark, and then to leave them alone.

What may this look like in practice? Imagine a small homelessness service that, instead of publishing the same impact statistics as everyone else, runs a monthly letter written by someone who has slept rough, in their own voice, unedited. It is bad for scanning and almost impossible to summarise cleanly, which is exactly why it works. The people who read it come back, forward it, turn up to volunteer, and remember the organisation’s name long after a tidy AI answer would have resolved and been forgotten. The strategy is not louder content. It is content that leaves a residue the machine cannot compress.

The hardest thing to forget

This is bigger than marketing, search, or platforms. We are entering an era where institutions compete inside outsourced cognition. Not only for visibility, but for interpretive authority, trust transfer, recommendation priority, and emotional memory inside machine-filtered environments. 

Once machines become cognitive partners rather than tools, the future might end up less human not because the machines turn cruel, but because they turn helpful in exactly the way that removes the conditions under which we learn to care at all.

The challenge stops being technological and becomes cultural, emotional, and most importantly, human. Because many human problems don’t need instant resolution, they need reflection, context, discomfort, ambiguity, and moral tension. Social change rarely arrives in an instant. And empathy rarely forms through optimisation. So the real question is not what happens when AI gets smarter. It is what happens to us when we stop having to think things through alone. When uncertainty is resolved before it has time to become responsible. When care is simulated before it becomes relational. When another person’s suffering can be handled efficiently enough that we are free to move on.

The organisations that survive will not be the easiest to summarise. They will be the hardest to forget. And for not-for-profit organisations, the task is not simply to be found. It is to remain felt.