The five questions every development director asked us — and what their answers revealed.

Twenty-seven conversations, one hour each, no demo, no slides. We asked people to walk us through their last submitted application from first contact to final upload, in order, out loud. When they skipped a step we made them go back.

We did it that way because the interview question people expect — what features do you want? — produces answers about tools they have already seen. The narrative question produces answers about what actually happened, including the parts nobody would think to request.

Five questions came back nearly every time. Where does the narrative live? Who checks the budget math? How do we know a renewal is coming? What happens when the person who wrote it leaves? And: can I trust this thing not to make things up?

Notice what is not on that list. Nobody asked for more funder leads. Nobody asked for a better search filter. Two people mentioned discovery unprompted, and both framed it as a problem of having too many options rather than too few. The demand was overwhelmingly for continuity, verification, and memory.

On the first question — where does the narrative live — the honest answer in twenty-two of twenty-seven organizations was a folder structure that only one person understood, plus a set of documents named things like final_v3_USE_THIS.docx. Several development directors could not, during the interview, locate the most recent version of their own organizational capacity statement. This is not a competence failure. It is what happens when the artifact is produced under deadline and then abandoned until the next deadline.

On budget math: in nineteen organizations, nobody checked it independently. The person who built the budget was the person who reviewed the budget. Three interviewees described discovering arithmetic errors after submission; one described discovering one after an award, which required an amended budget and a conversation they described as the worst professional week of their career.

On renewals: nine of twenty-seven had missed or nearly missed a renewal in the previous twenty-four months. The near-misses were caught by accident — a program officer emailing to ask if everything was alright, a board member seeing something at a conference. No organization we spoke to had a system that would have caught it on purpose.

The fourth question — what happens when the person leaves — was the emotional center of most interviews, and several people got quiet when it came up. Institutional grant knowledge lives in one inbox and one person's head. Turnover in this role runs high, partly because it is a job that produces burnout by design. Four organizations described losing a multi-year funder relationship outright because the person who held it left and nobody knew the cadence, the contact, or the history.

One development director described inheriting a portfolio and spending her first four months doing archaeology: reading three years of email to reconstruct which funders had said what, and to whom. She won nothing that quarter. Not because she was slow, but because the organization's memory had been deleted and she was rebuilding it from logs.

The fifth question — trust — came up defensively almost every time, usually phrased as a challenge. These are people who have been pitched AI tools that confidently produced fabricated statistics about their own programs. The bar for credibility was not enthusiasm. It was whether we could show the source of every claim, and whether the tool would say 'I do not have this' instead of guessing.

There were counter-findings too, and they mattered. Six organizations told us that any tool requiring a full data migration before producing value would be dead on arrival — not because migration is hard, but because it requires a quarter with no deadlines, and there is no such quarter. That constraint shaped the product more than any feature request: value has to arrive before the library is complete.

Two interviewees pushed back on the premise entirely. Their view: the bottleneck is not process, it is that the organization is chronically underfunded and the development office is understaffed by two people. Software cannot fix that. We think they are partly right, and it is worth saying plainly — a tool that gives back fifteen hours a week does not create the two hires. It just makes the existing capacity less likely to be lost to logistics.

The shared library, the pipeline history, the funder relationship timeline, the deadline calendar covering reports and renewals rather than just submissions, and the citation-first drafting model all exist because of these conversations. Not one of them is on the roadmap because it demos well. Most of them do not demo well at all.

We are running the same protocol again in 2026 with a different cohort, including organizations outside the Southeast and outside the HBCU-adjacent world, because we would like to know which of these findings are structural and which are ours.