Gurus and detailed lore plus an unshakable conviction that the way the world works can now be understood, and reshaped, by those who’ve mastered the new tool.   You’ve seen the discourse. You know the feeling: real excitement, and a nagging suspicion that more than a few people are performing understanding rather than demonstrating it.

AI?  Nope, that was a description from Harper’s magazine in 1984 about the spreadsheet.  The idea was that the world could be captured in rows, columns, and formulas. The excitement was justified but not for the reason people thought.

The obvious value was faster calculations, more data processed, numbers updated in real time. Useful, but incremental.  It was the word processor compared to the typewriter.  The real shift is what the spreadsheet made possible to ask, for those curious enough to be asking.

When calculation becomes cheap enough, questions that were previously impractical  or impossible become routine. The spreadsheet didn’t just make answers faster, it made new questions legitimate.

That’s a different order of change.   But tools don’t just expand what can be asked, they define what counts as a reasonable question in the first place. Once a way of thinking becomes easy to model, easy to defend, and easy to argue in a room, it starts to crowd out everything that doesn’t fit.

This plays out in any system that over-optimizes what it can measure.  The dominant theory embedded in fundraising systems is simple: the donor file is a response variable to be maximized. Budgets, reporting, evaluation and decision making are organized around campaigns. It’s all clean and defensible and it’s all structurally blind to the thing that determines whether the system grows or decays: the state of the relationship between donor and organization.

That relationship has no column.

A donor’s sense of why they give. Whether they see the mission as part of who they are. Whether the last interaction left them feeling respected or processed. These aren’t abstract ideas as they show up in behavior but they don’t show up cleanly in campaign reporting, so they keep losing every allocation decision.

Year after year, the system reallocates toward what can be measured and justified – i.e. more campaigns.  And like any system built this way, it tends to look fine until the underlying asset is too degraded to ignore. At that point the decline appears sudden, even though it has been compounding for years.

Now layer AI on top of that system.  The “do it faster” version is already here: Summarize, generate, personalize, automate.  This is fine but it’s not a transformation. Running the same system more efficiently produces the same outcome, just faster.

The more important shift is the same one the spreadsheet introduced, changing what questions are operational.

For fundraising, that means something very specific. The question the current system cannot ask in a usable way is this: what is the actual state of this relationship, for this individual, right now?

That includes things the current architecture was never built to represent: commitment, identity alignment, the quality of recent experience, sensitivity to frequency, responsiveness to different frames.  Passive behavioral signals alone get you marginally better than chance, the step change comes when you stop inferring the relationship and start measuring it directly.

But these shifts only happens if the questions change.  If AI is used as a confirmer, it will reinforce the existing system. Faster content, better timing, more efficient execution of the same underlying theory. The Volume Machine becomes more efficient and the underlying problem compounds more quickly.

If it is used as a questioner, it does something else. It allows you to test the assumptions that the current system treats as fixed. It allows you to run the counterfactuals and expose where the model is substituting what is easy to measure for what actually matters.

Every tool pushes in one of two directions. It either helps you extend the system you already believe in, or it helps you see where that system breaks.  The spreadsheet did both and most people chose the first path. AI will likely follow the same pattern.

Kevin