Move past static spreadsheet lists. See how AI builds living donor segments from giving, engagement and affinity signals, and what to do.
Where LYBUNT stops working
A gift officer pulls the LYBUNT list on a Monday, the donors who gave last year but not this, and sends the appeal on Thursday. In between, one of the names on it opened three emails, clicked a campaign page and started a peer fundraiser. The spreadsheet never knew, because LYBUNT and SYBUNT describe one variable, the recency of a gift, and freeze it the moment the list is exported. They have anchored nonprofit segmentation for decades and they still earn their place; they just cannot see anything that happened after Monday.
We think real segmentation needs the whole shape of a relationship: giving history, channel preference, event attendance, content affinity, and how someone engages between asks. AI donor segmentation reads all of those signals off the donor record at once and keeps the segment current as behaviour changes, so the list you send on Thursday is the list as it stands on Thursday. That is the shift this guide is about, from a static list rebuilt every quarter to a living audience that updates itself overnight.
The signals on the record
Three families of signal matter. Giving signals are amount, frequency, recency, recurring status and whether a gift was restricted or in someone's honour. Engagement signals are opens and clicks, page visits, event RSVPs and how recently any of that happened. Affinity signals are which programmes a donor reads about, which appeals she answers and which causes she shares. Held together on one unified donor profile, they turn a name-and-amount row into a portrait of intent.
AI is good at this work precisely because the patterns are noisy and there are too many of them for a person to hold. A model can notice that a mid-level donor who has started opening every newsletter and visited the legacy page looks ready for a planned-giving conversation long before anyone scanning a report would. Surfaced through donor analytics, that gives the team segments by likelihood to upgrade, churn risk and best next channel rather than one date column, and since industry estimates suggest most donor data goes unused, the gain is real.
What each segment needs
A segment is only useful if it triggers a different action. A first-time donor needs a fast, warm welcome that says where her money went, not a general year-end ask three months later. A loyal recurring donor needs stewardship and the occasional gentle upgrade, not a run of emergency appeals that trains her to ignore you. A lapsing donor whose engagement is rising is the highest-value reactivation you have, because the relationship is warming even though the giving paused; the cadence for that is in our guide to keeping donors with AI.
The principle is to match the message to the moment. High-affinity readers of one programme should hear from that programme, not from the organisation at large. A major-gift prospect flagged by upgrade signals should reach a person, not a sequence. And because the segments stay live they can be wired straight into delivery: an AI-built audience becomes the recipient list for an email, a text or a personalised ask, and refreshes before each send instead of ageing in a file.
Keeping segments live
A segment is worthless if it lives in one tool while the gift officers work in another. Whitelabel layers on top of the systems you already run, so there is no replatforming and no second database to reconcile, and two-way CRM sync keeps Salesforce, HubSpot and Klaviyo aligned with the live record. An updated churn score or affinity tag is visible to the colleague making the call rather than buried in an export nobody opened, and the segment and the source of truth stay the same object.
It also has to be safe, because segmentation models touch all of the donor data there is. The platform is PCI DSS Level 1, SOC 2 and HIPAA compliant through a Vanta-powered trust centre, pricing is 3.5 percent platform plus 1.1 percent processing with no monthly fee and no contract, and if you would rather build your own logic on top, the developer API exposes the same data the dashboard shows. The list the gift officer pulls on Monday is finally the same list she sends on Thursday.
Frequently asked questions
What is the difference between LYBUNT, SYBUNT and AI segmentation?
LYBUNT (Last Year But Unfortunately Not This) and SYBUNT (Some Year But Unfortunately Not This) sort donors by a single variable: when they last gave. AI segmentation reads giving, engagement and affinity signals together and updates as donor behavior changes. So instead of one date column, you get living segments like likely-to-upgrade or rising churn risk.
Do we need a data scientist to use AI donor segmentation?
No. The point of platform-based segmentation is that the modeling runs for you and surfaces plain-language segments in the dashboard, like new donors, loyal recurring givers or warming lapsed donors. Your team acts on the segments rather than building the models. A developer API is there if you want to extend the logic, but it is optional.
Will AI segmentation replace our existing CRM?
No. Whitelabel layers on top of your current stack with no replatforming, and keeps a two-way sync with Salesforce, HubSpot and Klaviyo. The live donor record and the segments built from it stay aligned with your CRM, so the colleague making a call sees the same updated scores and tags your dashboard shows.