Why AI can move fundraising from mass communication to individual engagement
IN BRIEF
For more than two decades, digital fundraising has been organized around the campaign. A nonprofit identifies a need, develops an appeal, selects an audience, and sends the same message to thousands – sometimes hundreds of thousands – of people on the same day.
The model works. Email remains one of the most productive digital channels in nonprofit fundraising, and its revenue continues to grow. But the same benchmarks that confirm email’s value also show how hard relevance is to achieve at scale: the average fundraising email now generates a direct response from roughly one recipient in two thousand.1
The sector’s answer has been segmentation – dividing supporters into ever more specific groups. Segmentation improves relevance, but every new segment adds another query, another content variation, another approval cycle. Eventually organizations hit a practical ceiling: no fundraising team can make an individual communication decision for every supporter, every day.
The constraint has shaped digital fundraising more deeply than we usually acknowledge. Campaigns, list, and segments are not just marketing conventions. They are mechanisms for managing the limits of human capacity.
Artificial Intelligence is beginning to change that constraint. For the first time, technology can help a fundraising team weigh very large numbers of individual supporter relationships at once – creating communications, interpreting behavioral signals, judging which message suits which person, deciding when to communicate and when to stay silent, and flagging the moments when a human fundraiser should step in personally.
The implication goes well beyond using generative AI to write better emails. If the cost of individualized decisions keeps falling, the fundamental unit of digital fundraising could shift from the campaign to the individual relationship.
This paper examines the evidence for that proposition, the limits of the campaign-centric model, the current state of nonprofit Al adoption, the concerns organizations are right to raise, and the conditions under which AI could strengthen – rather than erode – the human relationships on which fundraising depends.
1. Email Works. The Question Is Relevance.
Any serious argument for changing digital fundraising should begin by acknowledging that the existing system works.
According to M+R’s 2026 Benchmarks, email revenue grew 16% in 2025 and accounted for 11% of all online revenue, while annual revenue per subscriber rose from $1.87 to $2.40.1 The channel is not failing; it is one of the most reliable tools nonprofits have.
The same research shows what it costs to keep it working. The average subscriber received 50 nonprofit emails during the year, 31 of them fundraising appeals. The average appeal produced a click-through rate of 0.59%, a response rate of 0.05%, and $54 in revenue per 1,000 messages sent.1
An organization emailing 100,000 people at the benchmark response rate can expect roughly 50 direct responses. The other 99,950 messages are not wasted — email reinforces awareness, supports other channels, and shapes future giving in ways that resist attribution. But the gap between messages sent and responses received raises a strategic question: could fundraising technology become dramatically better at identifying the people for whom a particular story, request, or opportunity is actually relevant?
That question leads past the performance of any single email and toward the structure of digital fundraising itself.
2. The Campaign Was a Rational Answer to a Capacity Problem
The campaign is so familiar that it is easy to mistake it for an inevitable feature of fundraising rather than a product of operational constraints.
The sequence is always the same. A program needs funding, an emergency strikes, year-end approaches. The team develops the proposition, creates the communications, and then asks which members of its audience should receive them. Digital technology supercharged this model by collapsing the cost of distribution: once an email exists, sending it to 100,000 people is scarcely harder than sending it to 1,000.
Scale came at the price of relevance, and segmentation emerged as the fix. Rather than treating the whole database identically, fundraisers divided supporters into groups: donors and non-donors, monthly and one-time, active and lapsed, higher- and lower-value, issue by issue.
Each layer of segmentation improves relevance – and each adds work: another query, another message variation, another approval, another QA pass. The point at which further personalization becomes impractical is set less by technology than by the human effort available to manage it.
The capacity constraint is easy to document. The 2026 Charity Digital Skills Report found that 51% of charities cited lack of headspace and organizational capacity as a major barrier to digital progress, and 38% cited a lack of technical expertise or anyone able to lead digital work.2
Escalator’s Founding 50 research shows how the constraint plays out inside fundraising teams. In nine structured interviews with nonprofit digital and development leaders, only two organizations were doing substantial behavioral segmentation; the rest relied on basic suppressions, a single donor tier, or little meaningful segmentation at all.3 These teams had not rejected personalization – most said they should be doing more. What stopped them was practical: bandwidth, technical knowledge, CRM integration, and the time it takes to approve and test every additional variation.
The sector, in other words, does not lack interest in personalization. It lacks the capacity to deliver it.
3. Supporters Are Not Segments
Consider two supporters who each gave $100 last year. A traditional model may reasonably place them in the same audience – yet their relationships with the organization may have almost nothing in common. One has given repeatedly, attended an event, signed a petition, and returns again and again to content about a single program. The other made one spontaneous gift and has ignored nearly everything since.
From a database perspective they share the trait matters: both are $100 donors. From a relationship perspective, they are strangers to each other.
This is not an argument against segmentation. Segments made large supporters files manageable – groups a human team can understand, communicate with, and evaluate. The mistake is treating the segment as though it were the relationship. Segments are an operational compromise, adopted because managing every supporter individually was impossible.
AI raises the possibility that it no longer is.
4. When Individual Decisions Become Economically Practical
Imagine a nonprofit with 50,000 supporters. In principle, the fundraising team could ask 50,000 questions every morning: Should we contact this person today? Which of our current campaigns is most relevant to her? Should we ask for money, share an update, invite an action, or simply say thank you? Which version, at what time? And in some cases – would the best decision be to send nothing at all?
No team can answer those questions one person at a time. Machine-learning systems can process signals and produce recommendations at exactly that scale. That does not mean AI understands every donor, and it does not remove the need for human rules, judgment, and oversight. It means the economics of individualized decision-making are changing.
Which puts a question in front of the sector: what happens when making 50,000 individualized communication decisions a day becomes operationally feasible?
We do not yet have the sector-wide evidence to answer conclusively. The hypothesis has to be tested against fundraising outcomes, donor retention, staff workload, and the quality of supporter relationships. But the underlying technology is developing quickly – and nonprofit adoption has already moved past experimentation.
5. AI Is Already Here. The Next Phase Is Decisioning.
6. From Campaign-Centric to Supporter-Centric
The traditional model begins with the organization’s needs: create a campaigns, select an audience, distribute the message. An AI-first model can reverse the sequence — begin with the supporter, and ask which of the organization’s current communications is most appropriate for that individual, at that moment.
Campaigns do not disappear under this model. Organizations will always have stories to tell, money to raise, and moments that demand collective attention. What changes is the distribution logic: Instead of “Who should receive our spring appeal?”, the operating question becomes “What should this supporter hear from us next?”
It sounds like a subtle shift. It changes the unit around which fundraising is organized. The campaign still supplies the content, the objectives, and the organizational priorities. The supporter becomes the unit of decision.
7. Personalization Is a Relationship, Not a Salutation
“Personalization” has spent years describing modest variations in presentation: a first name in the greeting, an image swapped by interest category, an ask amount adjusted to previous giving. These techniques are useful — but they are not a personalized relationship.
Relationship-level personalization asks harder questions. Should a first-time donor get the same email as someone who has given for ten years? Should a supporter devoted to one program received stories about another? Should a monthly donor be solicited again simply because the calendar says today is appeal day? And what should happen when a supporter’s behavior signals that the relationship itself has changed? A person who suddenly opens everything, engages deeply with one issue, and increases her giving may not need another automated email. She may need a phone call.
There is also a harder discipline hiding inside true personalization: knowing when to stop. Email’s marginal cost of sending is near zero, and near-zero cost naturally encourages distribution. But the cost to the supporter is not zero. Every message claims attention; every irrelevant one spends credibility that future messages will need. A campaign-centric model asks how widely the campaign should go. An individual-decisioning model admits a possibility the campaign model cannot: for some supporters, the best communication today is silence.
That reframes the objective itself — from maximizing the reach of a campaign to optimizing the long-term quality of a relationship. It may prove more valuable than any improvement in the emails themselves.
8. AI Could Make Fundraising More Human, Not Less
The most understandable fear about AI is that automation will drain the humanity out of fundraising. The concern is widespread: Pew Research Center found that 57% of U.S. adults are highly concerned that AI will lead to less connection between people.6 The risk is real — badly designed automation would mean more messages, more generic content, and more distance between organizations and the people who support them.
But the current model deserves scrutiny on the same grounds. Sending an identical appeal to 80,000 people at the same moment is extraordinarily efficient. It is not intimate. Efficiency and intimacy are different things, and mass communication has always traded one for the other.
AI makes a different division of labor possible: machines doing what machines are strong at — processing information, finding patterns, managing complexity — while humans spend more time on the work that remains irreducibly human: judgment, empathy, creativity, conversation.
The Founding 50 interviews added an important design requirement. One participant imagined a system that becomes excellent at understanding and engaging a donor — while the human fundraiser responsible for that relationship remains unaware of what the system has learned. His worry was that digital engagement could improve while the human relationship quietly atrophies. His proposed solution: the system must report back. When engagement shifts, when a supporter’s interests come into focus, when personal outreach is warranted — the fundraiser should know.3
Responsible AI in fundraising, in other words, should do more than keep a human in the approval loop. It should keep the human in the relationship.
9. Fundraising Superpowers: Strength and Judgment
One way to grasp the opportunity is to imagine a fundraiser acquiring two capabilities that were previously impossible.
The first is strength. A small team suddenly does the work of a much larger one. An idea becomes a campaign; campaign material becomes appeals, updates, advocacy communications, and landing pages; multiple appropriate versions exist without anyone starting from a blank page. The fundraiser still supplies the purpose, knowledge, and judgment. The technology supplies the executional capacity.
The second is judgment at scale. The team weighs thousands of individual relationships simultaneously, using behavioral and historical signals to decide who should hear a story, who should not, when to communicate — and when to step aside because the relationship deserves a human.
Greek mythology offers the shorthand: Heracles for strength beyond ordinary capacity, Athena for wisdom and strategic intelligence. The point of the analogy is not that AI becomes a mythical autonomous fundraiser. It is that technology can give the existing fundraiser capabilities that were impossible to exercise at scale until now. Not replacement — augmentation: more strength to execute, and better information with which to judge.
10. Trust Is Not an Obstacle to Adoption. It Is a Requirement.
AI’s spread makes trust more important, not less. A 2026 Google/Ipsos study of roughly 21,000 adults across 21 countries found that 66% had used an AI application in the previous year — up 18 percentage points from 2024.7 Stanford’s 2026 AI Index reports that 58% of employees globally now use AI at work with some regularity — and that while 59% of people believe AI;s benefits outweigh its drawbacks, 52% say AI products make them nervous.8
Those findings are not contradictory. They describe a public growing familiar with a technology whose usefulness and risks are both becoming more visible. Charities mirror the tension exactly: 63% are excited by AI’s potential to improve efficiency and impact, while citing skills gaps, training, data privacy, accuracy, and plain lack of trust as barriers to going further.4
Trust should therefore not be treated as an irrational obstacle for vendors to market their way past. For organizations that hold donor relationships, reputations, and sensitive data, skepticism is appropriate. In practice, that means AI-first fundraising platforms must be able to explain how donor data is used, what information drives each decision, which controls remain with the organization, and how the organization keeps visibility over everything the system recommends and sends.
The burden is on technology providers to build systems worthy of that trust.
11. Capacity May Be AI's Most Democratic Contribution
The biggest consequences may land on small and mid-sized organizations. Large institutions already field specialist teams — strategists, analysts, copywriters, designers, CRM specialists, agencies. Smaller organizations cannot reproduce that structure at any level of ambition, and the pressure is structural: 63% of charities cite financial pressures as a barrier to digital progress.2
The 2026 Charity Digital Skills Report documents charities using AI to do work they previously could not do at all, for lack of staff or specialist expertise. The Founding 50 participants used strikingly similar language: asked to describe the product in their own words, they compared it to an email platform, an agency, or a person their organization could not afford to hire. One put the constraint plainly — her organization would not give her another employee, so she needed another way to create capacity.3
This may be the most important dimension of AI adoption in fundraising: not doing existing tasks a little faster, but making previously uneconomic work possible. A three-person fundraising department will not become a thirty-person one. It may, for the first time, exercise some of the capabilities that used to require one.
12. Doubt Is Rational. Certainty Is More Dangerous.
Voltaire is commonly translated as writing that “doubt is an uncomfortable condition, but certainty is a ridiculous one.” The line fits this moment in AI unusually well.
Nonprofits should approach AI with doubt: question its accuracy, examine its data practices, test its recommendations, establish governance, demand evidence of impact. But the skepticism should run in both directions. Assuming that the constraints digital fundraising was designed around will remain fixed is also a bet — and the adoption evidence is mounting against it. AI use among the public and the workforce is climbing steeply, while nonprofit AI remains concentrated in content production rather than decisioning.4,7,8 The first wave was about production. The next will be about decisions.
No one knows how fast that transition will run or how far it will go. The evidence does not justify claiming that individualized, AI-mediated fundraising will inevitably displace the campaign model. But assuming the existing model will remain essentially unchanged is a prediction too — and it is increasingly difficult to call it the conservative one.
13. The End of the Campaign?
The title of this paper is deliberately provocative, because campaigns are not about to disappear. Nonprofits will keep having priorities, stories, emergencies, and moments that demand collective action — and campaigns will remain the way that work is organized.
What may end is the assumption that the campaign determines the communication.
Picture an organization running ten campaigns at once. One supporter hears about Campaign A today because it matches her demonstrated interests. Another gets Campaign D tomorrow. A third receives an update instead of an ask. A fourth receives nothing, because recent contact has been sufficient. A fifth is flagged for a fundraiser, because her engagement suggests a conversation would do more than another automated message.
The campaigns still exist. They still embody the organization’s priorities and supply the stories from which communications are built. What changes is the decision about which campaign matters to which person.
This is a different organizing principle for fundraising.
Conclusion: A Different Unit of Fundraising
For decades, digital fundraising has operated inside one unavoidable constraint: too many supporters, too few fundraisers to treat each relationship individually. Campaigns, lists, segments, and automation all grew up inside that reality, and they have served the sector well.
That constraint is starting to move. AI can increasingly create content, process information, recognize patterns, and support decisions at a scale no human team can reproduce. This does not yet prove that individualized engagement will outperform campaign-centric fundraising — the evidence on revenue, retention, and relationship quality still has to be built. It does justify a serious question:
If individualized decision-making becomes economically practical, should the campaign remain the fundamental unit of digital fundraising?
The organizations that answer it best may not be the ones that automate the most. They may be the ones that drew the clearest line between what technology should do and what must remain human. Machines bring scale, pattern recognition, and capacity. Humans bring purpose, empathy, judment, and relationships.
Used well, AI does not make fundraising less human. It creates the capacity to treat supporters less like entries in a database and more like people. Fundraisers have always aspired to give every supporter the right relationship; they have rarely had the resources. AI may finally make that ambition operational.
One person at a time, at scale.
About the Escalator Founding 50 Research
Escalator commissioned qualitative product research in July 2026 with nine nonprofit digital and development professionals, representing nine organizations across multiple cause areas, in roles spanning digital fundraising, development, database management, engagement, and executive leadership. Each participant viewed the same recorded product demonstration and answered the same structured questions before and after it.
The research was designed as qualitative product and market discovery, not a statistically representative survey of the nonprofit sector. Its finding are used in this paper to illustrate how sector-wide challenges appear inside real fundraising workflows.
References
- M+R. M+R Benchmarks 2026: Email Messaging. 2026. Nonprofit email volume, fundraising response, revenue, and list-growth benchmarks.
- Charity Digital Skills Report. Priorities and Barriers for the Year Ahead. 2026. Research on financial pressure, organizational capacity, technical expertise, digital fundraising, and CRM limitations within the charity sector.
- Stratovation Partners / Escalator. Escalator Founding 50: Demo and User Feedback. July 2026. Nine structured qualitative interviews with nonprofit digital and development professionals.
- Charity Digital Skills Report. Artificial Intelligence. 2026. Research on charity AI adoption, use cases, benefits, organizational readiness, governance, skills, and barriers.
- Blackbaud Institute. The Status of Fundraising in the AI Era. 2025. Survey research examining fundraising performance, digital maturity, generative AI, and predictive AI adoption.
- Pew Research Center. How the U.S. Public and AI Experts View Artificial Intelligence, McClain, C., Kennedy, B., Gottfried, J., Anderson, M & Pasquini, G. April 3, 2025.
- Google / Ipsos. Multi-Country AI Survey 2026. January 15, 2026. Research involving approximately 21,000 adults across 21 countries.
- Stanford Institute for Human-Centered Artificial Intelligence. Al Index Report 2026: Public Opinion. Stanford University, 2026.
Why AI can move fundraising from mass communication to individual engagement
For more than two decades, digital fundraising has been organized around the campaign. A nonprofit identifies a need, develops an appeal, selects an audience, and sends the same message to thousands — sometimes hundreds of thousands — of people on the same day.
The model works. Email remains one of the most productive digital channels in nonprofit fundraising, and its revenue continues to grow. But the same benchmarks that confirm email’s value also show how hard relevance is to achieve at scale: the average fundraising email now generates a direct response from roughly one recipient in two thousand [1].
“The sector, in other words, does not lack interest in personalization. It lacks the capacity to deliver it.”
When Individual Decisions Become Economically Practical
Imagine a nonprofit with 50,000 supporters. In principle, the fundraising team could ask 50,000 questions every morning: Should we contact this person today? Which of our current campaigns is most relevant to her? Should we ask for money, share an update, invite an action, or simply say thank you? Which version, at what time? And in some cases — would the best decision be to send nothing at all?
No team can answer those questions one person at a time. Machine-learning systems can process signals and produce recommendations at exactly that scale.
It means the economics of individualized decision-making are changing.
The Next Phase Is Decisioning
The first phase of nonprofit AI has been about helping people create things faster. The next phase is about helping them make better decisions — and that transition, from creation to decisioning, is where the campaign model begins to come under pressure.
Content generation answers the question “What should this email say?” Decisioning asks a different set: Who should receive this communication? Which communication suits this supporter? When should it arrive? When should the supporter be left alone?
Generation improves production; decisioning shapes the management of the relationship itself.
From Campaign-Centric to Supporter-Centric
An AI-first model can reverse the sequence — begin with the supporter, and ask which of the organization’s current communications is most appropriate for that individual, at that moment.
Campaigns do not disappear under this model. Organizations will always have stories to tell, money to raise, and moments that demand collective attention. What changes is the distribution logic: instead of “Who should receive our spring appeal?”, the operating question becomes “What should this supporter hear from us next?”
It sounds like a subtle shift. It changes the unit around which fundraising is organized. The campaign still supplies the content, the objectives, and the organizational priorities. The supporter becomes the unit of decision.
AI Could Make Fundraising More Human, Not Less
The most understandable fear about AI is that automation will drain the humanity out of fundraising.
But the current model deserves scrutiny on the same grounds. Sending an identical appeal to 80,000 people at the same moment is extraordinarily efficient. It is not intimate. Efficiency and intimacy are different things, and mass communication has always traded one for the other.
AI makes a different division of labor possible: machines doing what machines are strong at — processing information, finding patterns, managing complexity — while humans spend more time on the work that remains irreducibly human: judgment, empathy, creativity, conversation.
The End of the Campaign?
The title of this paper is deliberately provocative, because campaigns are not about to disappear.
What may end is the assumption that the campaign determines the communication.
The transition is not from campaigns to no campaigns. It is from asking “What do we want to send?” to asking “What is the most appropriate thing for this supporter to hear from us next?”
That is a different organizing principle for fundraising.
The organizations that answer it best may not be the ones that automate the most. They may be the ones that draw the clearest line between what technology should do and what must remain human.
Machines bring scale, pattern recognition, and capacity. Humans bring purpose, empathy, judgment, and relationships.
Used well, AI does not make fundraising less human. It creates the capacity to treat supporters less like entries in a database and more like people. Fundraisers have always aspired to give every supporter the right relationship; they have rarely had the resources. AI may finally make that ambition operational.
One person at a time, at scale.
Dowload and read the full report
Download the full paper
11 pages
Want to read more like this?
Be the first to know about new features, fundraising tips, and what’s next for Escalator.
