AI Is Saving Nonprofits Time. But Is It Helping Them Change Lives?

Artificial intelligence is spreading through the nonprofit sector at breathtaking speed. Yet behind the impressive adoption figures lies a troubling question: what if the greatest obstacle to the AI revolution has nothing to do with technology?

The complete interview with Remy Reya, Director of AI & Thought Leadership at Compass Pro Bono, is available to download via the link at the end of this article.

The management of a nonprofit organisation had a problem. Its employees were not embracing artificial intelligence quickly enough.

The technology promised to reduce administrative work, improve efficiency and free up precious resources. The benefits seemed obvious. Yet staff were apparently reluctant to move forward.

The organisation called in Compass Pro Bono, a nonprofit that helps other organisations strengthen their capabilities. What its advisers discovered was rather uncomfortable.

The problem was not simply the employees’ attitude towards AI. It was management itself.

Staff felt pressured, judged and overwhelmed by the insistence that they adopt the technology faster. Leadership’s enthusiasm had become an obstacle to the very transformation it was supposed to encourage.

Instead of prescribing more software, Compass Pro Bono organised a workshop to improve communication between management and employees. When the organisation was surveyed again, AI adoption had increased significantly, accompanied by much greater enthusiasm among staff.

The technology had not changed. The people had been given the opportunity to understand it, question it and participate in decisions about its use.

This episode, recounted to me by Remy Reya, Director of AI & Thought Leadership at Compass Pro Bono, reveals something profoundly important about the current enthusiasm for artificial intelligence.

We may be experiencing the most extraordinary technological transformation in decades. But organisations are discovering that acquiring powerful tools and becoming more capable are two very different things.

And the evidence is becoming difficult to ignore.

The extraordinary numbers — and the uncomfortable truth behind them

In August 2026, research commissioned by Momentive Software found that 91% of surveyed nonprofit and educational organisations were already using artificial intelligence in some official capacity.

Ninety-one per cent.

By almost any measure, this represents extraordinary technological adoption.

But research published by The Bridgespan Group and NTEN revealed another side of the story.

Only 4% of respondents indicated that AI was meaningfully embedded in their organisation’s everyday work. Just 8% reported having a formal AI roadmap, while 70% believed their organisations were missing significant opportunities to use the technology.

The studies surveyed different populations, so their figures cannot be compared as if they came from a single experiment. Nevertheless, their findings point towards the same disturbing contradiction.

Artificial intelligence is spreading much faster than organisations are learning how to make it genuinely useful.

And nowhere is this contradiction more consequential than in the nonprofit sector.

These organisations exist to tackle problems that conventional economic structures and public institutions have often failed to resolve adequately. They provide assistance, support vulnerable communities, address inequalities and undertake work whose value cannot always be expressed in financial terms.

Yet many operate under severe financial pressure, with limited staffing and growing demand for their services.

AI appears to offer precisely what they need: greater capacity without equivalent increases in expenditure.

But what happens when the technology intended to solve a resource problem requires resources that an organisation simply does not possess?

That was one of the central questions I wanted to investigate when Abby Forbes of On The Marc Media approached me and invited me to cover this story.

Through her introduction, I interviewed Remy Reya, whose work at Compass Pro Bono places him directly at the intersection of AI, organisational strategy and nonprofit operations.

I was interested in something more fundamental than the familiar promises of efficiency and automation.

I wanted to understand whether artificial intelligence was actually changing what organisations could accomplish — or merely allowing them to do the same things faster.

When the cure demands more than the patient can afford

Reya describes generative AI as a general-purpose technology, capable of assisting with a remarkable variety of tasks.

Paradoxically, that versatility makes adoption more complicated.

A tool designed for a single purpose is relatively easy to introduce. AI requires employees and managers to reconsider numerous activities, sometimes questioning working methods that have remained unchanged for years.

Reya explains that implementation often follows what he calls a J-Curve.

Just as a newly recruited employee initially requires training and supervision before becoming productive, AI adoption may temporarily slow an organisation down.

People need time to experiment, make mistakes, develop judgement and discover where technology actually helps.

For a well-funded company, this period may be an acceptable investment.

For an overstretched charity struggling to meet immediate demands, it can be a serious obstacle.

And here lies a cruel irony: the organisations that could benefit most from greater operational capacity may be those least able to endure the investment required to achieve it.

Reya argues that implementation must begin with an entirely different question.

“But the starting point should be the task, not the tool.”

It is a principle that sounds elementary until one considers how frequently organisations do precisely the opposite.

They acquire technology, announce an innovation initiative and then attempt to discover what the new systems should actually do.

Compass Pro Bono instead encourages organisations to examine work through a framework it calls Head/Hands/Heart.

Activities requiring human expertise and judgement may benefit from AI assistance. Repetitive manual activities may be suitable for automation. Work that depends on genuine empathy, personal presence and relationships should remain fundamentally human.

This is a more intelligent approach than simply asking how many operations can be automated.

The distinction becomes especially important in an environment where the people being served are not abstract customers or entries in a database, but individuals whose circumstances may require understanding, sensitivity and trust.

Ten hours saved. But what was actually gained?

One of the most interesting aspects of my interview concerned Compass Pro Bono’s own experience.

The organisation introduced a formal AI policy and team-wide access to AI tools in late 2024.

According to internal employee surveys, by January 2025 staff were reclaiming an estimated four to five hours per person each week. By June, that figure had reportedly increased to ten to twelve hours.

Consider the implications.

An extra working day, or more, potentially available to every employee each week.

In theory, the effect could be substantial.

Administrative tasks might require less attention. Employees could spend more time with beneficiaries, cultivate donor relationships, develop programmes or undertake strategic work that would otherwise be postponed indefinitely.

But I wanted to know something else.

How had those time savings been measured? And, more importantly, what evidence existed that the additional capacity had translated into better outcomes?

Reya’s response was particularly revealing.

The figures came from internal surveys in which employees estimated the time they had reclaimed and described how their work had changed. They were therefore self-reported estimates, not independent measurements of productivity.

More importantly, Reya made a distinction that is too often absent from discussions about AI:

“But we don’t equate time saved with impact.”

Some employees reported spending their additional time on valuable activities, including collaboration, strategic planning and relationships with volunteers and nonprofit partners.

Others found themselves using the reclaimed hours to work through existing backlogs, rather than pursuing the more valuable activities they would have preferred.

Compass identified what Reya describes as a values gap: the distance between how employees wanted to use their newly available time and how that time was actually being spent.

This deserves far more attention than another statistic about productivity.

An organisation can become dramatically more efficient without becoming significantly more effective.

It can process more documents, prepare more communications and complete more administrative procedures while making remarkably little progress towards its underlying objectives.

The uncomfortable question is whether AI is liberating employees from unnecessary work or merely enabling organisations to produce more of it.

Saving ten hours a week is impressive.

But if those hours disappear into an endless queue of additional tasks, what has really changed?

The human relationship that cannot be automated

Fundraising provides a particularly illuminating example.

AI can analyse donor information, identify prospective contributors, prepare communications and reduce the administrative burden associated with seeking financial support.

These are valuable capabilities, especially for organisations whose survival depends on maintaining a reliable flow of donations.

But fundraising is not simply a process of identifying people with money and persuading them to provide it.

It depends on credibility, confidence and relationships developed over time.

I asked Reya where automation might undermine the very trust that nonprofits depend upon.

His response was unequivocal:

“Personalization can’t offer a substitute for real relationships; a donor should never feel like they are receiving an automated version of a personal exchange.”

The distinction is crucial.

AI can help a fundraiser prepare for a meaningful conversation. It cannot make that conversation meaningful merely by producing convincing language.

Indeed, the ease with which AI generates personalised communications introduces a potential danger: the industrialisation of apparent sincerity.

An organisation may believe it is developing closer relationships because its messages sound more personal, even when genuine human engagement is diminishing.

The same problem extends beyond fundraising.

Where people are dealing with illness, poverty, exclusion, displacement or personal difficulties, human attention cannot be reduced to the production of plausible responses.

The most valuable application of AI may therefore be not replacing human interaction, but creating more time for it.

That is an important distinction between automation and progress.

The risks nobody can afford to ignore

There is another complication.

While managers debate strategies and policies, employees may already be using AI tools independently.

The phenomenon is known as shadow AI: the use of artificial intelligence outside established organisational guidance or oversight.

For nonprofits handling donor information, beneficiary records and potentially sensitive personal data, the consequences can be serious.

Reya does not believe blanket prohibitions provide an effective solution. Instead, he advocates clear rules, staff education, appropriate technological safeguards and an environment in which employees can openly discuss their needs.

This approach recognises an uncomfortable organisational reality: banning a useful technology does not necessarily prevent people from using it.

Nevertheless, making AI available through approved systems is only part of the answer. Privacy protections, data management, accuracy, bias and accountability remain essential considerations.

And these responsibilities create another possible source of inequality.

Larger nonprofits may possess the budgets, specialist employees and infrastructure required to implement sophisticated AI systems responsibly.

Smaller organisations may struggle to afford even basic training or governance.

Reya acknowledges this danger, warning of an AI capacity gap developing alongside existing inequalities in resources.

Yet he also believes smaller nonprofits could be among the principal beneficiaries of AI, provided they receive appropriate support.

Both possibilities are real.

Technology can reduce the disadvantages of being small. But it can also magnify the advantages of already possessing money, expertise and organisational discipline.

Access to an AI application does not necessarily mean access to the knowledge required to use it effectively.

The revolution that has yet to happen

Artificial intelligence will undoubtedly continue to develop. Its capabilities will expand, its applications will multiply and many activities currently performed by human workers will become increasingly automated.

But the future of an organisation cannot be determined by the sophistication of its software alone.

Reya’s responses repeatedly return to questions that are fundamentally managerial rather than technological.

What are we trying to achieve? Which activities genuinely create value? How should we measure improvement? What must remain in human hands? And how can employees participate meaningfully in changes that affect their work?

These questions are not new.

They belong to the history of organisational management, long before anyone imagined generative artificial intelligence.

What is new is the extraordinary speed at which technology is forcing institutions to confront them.

There is also a broader lesson in the nonprofit experience.

If an organisation devoted to helping people measures success primarily through the number of hours saved, documents processed or communications generated, it risks confusing its internal machinery with its actual purpose.

The same mistake can be made by businesses, governments, universities and virtually any other institution.

The real value of AI cannot ultimately be determined by counting how often it is used.

It must be judged by what becomes possible because it is used.

Does a nonprofit reach people it could not previously reach? Does it provide better assistance? Does it improve decisions, strengthen relationships or produce outcomes that genuinely matter?

Does the technology make the organisation better at fulfilling its purpose?

Or does it simply make existing inefficiencies run faster?

These are the questions that should determine whether the AI revolution deserves its name.

For now, the tools have arrived. The enthusiasm is everywhere. The adoption figures are extraordinary.

But the revolution that matters most is not the one taking place inside the machines.

It is the one that still needs to happen inside organisations.

---

About the interview

For this article, I interviewed Remy Reya, Director of AI & Thought Leadership at Compass Pro Bono, exploring ten questions concerning AI adoption, organisational change, productivity, fundraising, governance and the future of the nonprofit sector.

Download the complete interview →

I would like to thank Abby Forbes and her colleagues at On The Marc Media for inviting me to explore this story and facilitating the interview. My thanks also go to Remy Reya for his time, thoughtful responses and willingness to engage with the difficult questions surrounding AI’s real impact.

Bob Mazzei is an author and business and technology professional with extensive experience in organisational processes, management and information technology.