Anyone building AI into a CRM needs to know what they are building

AI can take over or prepare a growing number of tasks in a CRM: putting together a selection for a mailing from a request in plain language, looking up figures, drafting a thank-you message. With Claudeforce, the partnership between Salesforce and Anthropic, that option is coming within reach of Salesforce users as well. Add everything else already on the market and there is a great deal you can do.

At Fundflow, we are therefore thinking about how AI can strengthen the CRM. We are deliberately taking our time and want to think through every application before we build it.

Everything seems easy to build these days

With AI, almost anyone can have a working application generated in a short time, even without any development knowledge. That is a big step forward for anyone who wants to try an idea quickly.

A CRM that holds donor data is a different matter. If you do not know what sits behind an application, you cannot check whether it does what it promises. You do not know where the data goes, what happens when something goes wrong or who can fix it. You end up using an application without knowing how it works.

Why we think before we build

What matters to us is building something we know works. That takes a little more time. For each application, we look at what impact it has once it is built into the CRM, how we keep donor data safe and how we monitor the quality of the results.

AI results look convincing, even when they are wrong. A selection that looks right at first glance but contains the wrong group of donors ends up as a mailing to people it was never meant for. Donor data is personal data: it must be clear in advance where it goes, who processes it and what happens with the result.

The risk depends on the type of task. Looking up an answer has few consequences if it is wrong. Sending a mailing does. The further AI goes into the work, the more important it becomes that a fundraiser can see what happened and step in before anything goes out.

The sector is already in the middle of this. Research among fundraisers shows that 86 percent use AI, while fewer than half follow internal guidelines. Use is running ahead of the rules. That is why we add capabilities one at a time and test each application before it reaches users.

What we want to hear from fundraisers

Besides control, we need input. How would you like AI to help you as a fundraiser today, inside the CRM or outside it? That could be selections for a mailing, getting figures without having to build a report, preparing a conversation with a major donor or checking data. It could also be something I have not thought of yet, and that is the answer that interests me most.

What you need helps determine what we develop. A developer knows what is technically possible. A fundraiser knows what takes time in daily work and what makes a difference.

Let me know

Send me an email at [email protected] or schedule a meeting to talk it through. One sentence is enough, for example about a task that costs you time every week or something you would like to hand over to AI. I read everything and take it into account in what we work out next.