Human-First AI: Why Your Smartest Agent Still Needs You
AI has moved quickly from emerging trend to operational reality across industries, including fundraising. While much of the talk around AI is focused on generative AI products such as ChatGPT, agentic AI (also referred to as “AI agents”) is growing in prominence. Reactions to agentic AI vary widely, from skepticism to the assumption that agents can be set free “right out of the box.”
A nuanced, human-first approach is a more balanced way to address the topic. The tools are powerful, but relationships and trust are the cornerstones of fundraising success. This is exactly why a human-first approach to AI is essential.
Free Resource
How AI agents support fundraising teams
As Adriana Dolgetta Cosgriff, founder and chief executive officer of Capacita Consulting, states, “Capacity is a tool, and trust is a choice you make. It’s a practice that goes at the speed of human relationships.”
Fundraising teams face constant pressure on their time and cognitive load, but AI agents can help expand capacity by:
- Reaching more donors with personalized outreach that feels timely and relevant
- Moving routine cultivation work forward based on donor context and relationship history
- Taking on follow-up and coordination work so fundraisers can spend more time where human judgment matters most
While this can help teams move faster, fundraising isn’t just about speed. You constantly make informed decisions in a way that preserves trust. Vered Siegel, principal of Seagull Partners and director of prospect management and research at American Friends of the Hebrew University, writes: “One of the biggest shifts generative AI has introduced in our industry is that information is no longer the scarce resource. Judgment is.”
We can now generate lists, drafts, summaries, and signals faster than ever, but that doesn’t automatically make our decisions better. One key aspect of being a strategic partner right now means helping the room slow down just enough to ask the right questions. The benefit of agentic AI is that it gives humans more time to focus on key judgement and relationship moments.
Why your agent still needs you
It helps to picture your AI agent as a brand-new team member. I think about myself, when I took my first fundraising job in my mid-20s: I had lots of energy, I was eager to get involved with everything—and I thought I knew a whole lot more than I actually did. Fortunately, I had a great supervisor who could teach, encourage, and redirect me when needed.
In the same way, your new AI agent is ready to help, but it needs you to provide guidance. Without guidance, errors can occur, just as they can with human employees. A new employee may come up with a very confident but completely inaccurate solution (based on their knowledge). Likewise, an AI agent can get things wrong if it doesn’t have the right context or training data.
An amusing example of this occurred recently in my home, when my spouse was using AI to research replacement hardware for window blinds. Instead of returning retail listings for the hardware, the AI tool misinterpreted the photo and insisted, “This is a laser-cut model of the Silverstone racecourse in England.” While incidents like this can be a little funny, they can also erode trust, and erosion of trust can have very serious consequences in fundraising.
Used well, AI agents can help fundraising teams expand capacity, move faster on donor opportunities, and keep routine work moving. But, like any new colleague, guidance, training, context, and oversight are key to maintaining trust and getting an agent’s best work.
Where to keep humans in the loop
Protect your brand voice
Your organization has a tone, a mission, and a consistent way of speaking to supporters. You probably don’t want your AI agent telling donors, “This organization is lit!” (or whatever the hot slang is now). But you also may not want something so formal it sounds like a legal notice. You want your AI agent’s communication to accurately reflect your organization’s brand voice in the same way that a human voice would communicate.
Just as you would make sure a new assistant had plenty of reading material to get to know your organization, make sure your AI agent is trained on this same material—your mission statement, strategic plan, website copy, donor-facing FAQs, event calendar, and regular donor-facing correspondence (gift acknowledgments, stewardship documents, etc.).
Have a human review your agent’s early drafts and provide detailed feedback until the messages accurately and consistently represent your brand voice.
Define moments for human escalation
One of the first lessons I learned as a new fundraiser was which questions not to answer. For example, if someone called and asked for the contact information for an alumnus/alumna, a faculty or staff member, or a student, I was instructed not to provide that information; rather, I would take the caller’s contact information and give it to the person that they were trying to get in touch with.
Likewise, decide in advance when your AI agent should route a conversation to a person. This protects supporters, protects your staff, and keeps the agent from guessing in high-stakes moments. Some examples of when to hand off to a human include:
- Sensitive requests such as “What Social Security Number do you have on file for me?” or anything involving banking details, identity verification, or confidential account data.
- Explicit preference, like when the supporter asks to speak with a “real human,” requests a call, or indicates discomfort.
- Exceptions and complaints, for example, disputes about tax receipts, chargebacks, pledge adjustments, event accessibility concerns, or dissatisfaction that needs empathy and judgment.
- Anything outside the knowledge base provided to the AI agent. If the answer isn’t in approved materials, the agent should not improvise.
Maintain accountability and oversight
To support responsible use, AI decisions should be:
- Connected to real-time data
- Visible to fundraising teams
- Owned by your organization—not external tools.
The architecture behind an AI agent matters as much as its capabilities. Strong performance depends on combining human guidance and thoughtful design: the fundraiser, acting as the agent’s manager, provides direction and context, while built-in guardrails help keep outputs aligned with approved standards.
Native tools (the kind that operate within your system of record) ensure that the AI agent is acting on timely data, as well as ensuring that human oversight remains within your organization and accountable to your team. Native AI also facilitates integration of multiple agents and workflows, rather than relying on a collection of disconnected tools and vendors.
Agentic AI can be a meaningful force-multiplier for fundraising—but it works best when it’s treated like what it is: a capable team member whose impact depends on human guidance and accountability. If you lead with a human-first approach that directs AI toward the work that matters most, you’ll expand what’s possible for your team without sacrificing the trust-built relationships that make fundraising work.
Want to learn more about what it means to take a human‑first approach to agentic AI in the philanthropic sector? Check out the webinar, Smarter Together: A Human First Approach to Agentic AI
