Don't Just Ask the Database Directly

Generative AI assistants embedded directly into fundraising CRMs are being promoted as transformative technology: a way for fundraisers to bypass technical barriers by asking natural language questions and receiving instant insights.

In vendor messaging, these tools promise streamlined workflows, smarter donor engagement, and faster decisions. Here are a few of the promises they make:

• Retrieve contact information and recent gift history via natural language prompts such as “Find John Doe’s contact information and latest gift.”
• Generate draft communications, suggest engagement strategies, and summarize donor histories through simple chat queries.
• Deliver next-best-action recommendations across a donor base.
• Automatically draft personalized emails, engagement plans, and reminders for development officers and major gift officers.
• Generate personalized content such as emails and call scripts tailored to an organization’s voice and audience.
• Assist in drafting donor outreach and engagement plans to save time on communication.

But fundraising databases are not clean systems of record like inventory counts or financial transactions; they are accretive, relational, and policy-dependent objects. They are manageable, but inherently messy.

And we know what happens when the data feeding AI tools is messy, incomplete, or contextually ambiguous: garbage in, garbage out…at scale.