David Geilhufe Has Spent 30 Years Making Nonprofit Technology Less of a Chore

In the first days after Hurricane Katrina, the internet had too much information and not enough answers. Missing-person notices were scattered across Craigslist, Yahoo, Red Cross pages, message boards and improvised databases. A family could check 20 sites and miss the useful entry on the twenty-first. David Geilhufe, then part of the open-source civic-technology world, saw the shape of the problem: the web did not need one more place to post. It needed a way for all those places to talk.

Geilhufe began calling and emailing people. Developers started scraping structured sites. Volunteers took on the slower job of reading messages and typing names, ages, locations and descriptions into consistent fields. With Ka-Ping Yee and Jon Plax, he helped create the PeopleFinder Interchange Format, a compact standard that let separate survivor registries exchange records. The Salesforce Foundation supplied database infrastructure. Open-source organizers supplied code, coordination and a working habit of trust.

The project moved with the tempo of an emergency. By the morning of September 6, 2005, more than 2,100 volunteers had created over 68,000 records. The wider effort eventually involved roughly 3,000 volunteers, accumulated 649,015 entries and served more than one million searches. An academic case study recorded the blunt economics of it: four days, zero dollars. The more important achievement was architectural. A temporary network of people and organizations agreed to make their work legible to one another.

Two decades later, Geilhufe still describes technology’s public value in the language of usefulness. “Technology’s potential for good isn’t abstract,” he wrote while marking PeopleFinder’s twentieth anniversary. “It’s deeply practical.” The software is different now. The working question is almost identical: how do you turn fragmented information into help without handing an already stretched organization another assignment?

I. The first principleBuild around the burden

Geilhufe’s path into nonprofit technology began before the term became a startup category. His public record reaches back to community-computing and digital-inclusion work, including Eastmont Computer Center in Oakland. He studied at UC Santa Cruz’s Cowell College and later at American University. In Washington, DC, he volunteered as a mentor for at-risk youth and encountered what nonprofit workers call the starvation cycle: funders want program results, but resist paying for the operating capacity that makes those results possible.

That experience gave him a durable suspicion of software sold as transformation without the time, training or organizational support required to use it. Nonprofits do not operate in a frictionless market. Funding can be restricted. Staff turnover can erase institutional knowledge. Technology refresh cycles can stretch for years. A cheap tool can become expensive if it quietly demands hours that nobody has.

In the early 2000s, this way of thinking took an open-source form. Geilhufe helped create CiviCRM, a constituent relationship management system designed for advocacy groups and nonprofits. At CivicSpace, where he served as managing partner, the aim was to package organizing tools so groups could manage contacts, donations, events and communications without stitching every component together themselves. The work was technical, but the premise was organizational: a mission needs plumbing.

In 2007, NetSuite co-founder Evan Goldberg recruited him to lead NetSuite.org. The program grew from a corporate philanthropy effort into a nonprofit division serving organizations with cloud-based financial and operating software. By 2019, Geilhufe could point to 1,300 nonprofit and social-enterprise users and six years of Hackathon4Good events. Those hackathons paired NetSuite employees with charities to solve defined operational problems. The useful part was not the theater of a demo. It was bringing technical attention to work that normally sat below a product roadmap.

He also served on the board of NTEN from 2014 to 2018, a role that placed him inside a network of practitioners who had watched waves of nonprofit software arrive, consolidate and disappear. Colleagues describe him as a systems thinker and natural teacher. One public recommendation adds a detail that fits his writing: dry wit, paired with an instinct to forecast second-order consequences.

II. The recurring cycleEvery tool inherits a business model

Geilhufe talks about nonprofit-software history with the pattern recognition of someone who has watched the same movie in several formats. Open-source projects solve an ignored need. SaaS companies make the experience easier. Growth capital enters. Products are acquired, merged or sunset. Customers, whose data and workflows may have accumulated over a decade, discover that their technology decision was also an ownership decision.

He does not treat this as evidence that one funding model is pure and another corrupt. His sharper observation is that incentives eventually become product behavior. A public company, a private-equity owner, a founder, a cooperative and an open-source community each have different pressures. Nonprofits feel those pressures through prices, support, migration work and the fate of features built for their peculiar needs.

This is why capacity, in Geilhufe’s telling, is concrete. It is not a vague synonym for goodness. It is the ability to act: staff time, usable systems, reliable partners, financial room and knowledge that survives a departure. Good software can expand that ability. Poorly fitted software consumes it.

III. A different AI postureGuardrails as product features

Geilhufe became Resilia’s CEO in 2024, joining a company founded by Sevetri Wilson and built around nonprofit capacity. Resilia connects nonprofits with coaching, peer learning and operating tools while giving foundations a way to support and understand a portfolio of grantees. The company says its network reaches more than 24,000 nonprofits. The job put Geilhufe back at a familiar seam: translating information held across many organizations into action that respects each organization’s context.

Artificial intelligence makes that seam more powerful and more dangerous. Grant applications, reports and program-officer notes contain sensitive information. A generic chatbot can summarize them, but it can also send private material into the wrong system, flatten local nuance or produce a claim nobody can trace. Geilhufe framed the tension directly: “I wonder if we are setting up the social sector for failure with AI.”

His answer is neither a pause until every harm disappears nor adoption without conditions. At Resilia, he has argued for a harm-aware approach organized around four product requirements.

Resilia’s Grantmaker Intelligence product, launched in 2025 and powered by an AI system called Murdoc, applies that framework to reports, shared drives and financial information. Its proposed value is institutional memory: a foundation can ask a question across grant cycles, find patterns in a portfolio and trace an answer back to the underlying material. The model retrieves and compresses. People retain the authority to decide.

That distinction carries an old PeopleFinder lesson. Shared infrastructure becomes useful when it gives dispersed participants a common view. It becomes trustworthy when they can understand how the view was assembled. In 2005, the mechanism was a simple data standard and an army of volunteer typists. In 2025, it is retrieval, models and governance. Both require someone to care about provenance.

IV. From teaching to doingThe to-do list gets the final vote

The most revealing change at Resilia did not begin with a model. It began with users pushing back on a familiar capacity-building ritual. A nonprofit attends training, receives templates and returns to work with a longer list. Resilia’s users asked why the company could not help them complete the task instead of explaining how to complete it.

Geilhufe took the complaint seriously. During 2025, Resilia moved toward what he calls a “let’s get it done together” approach. Its CapacityNow offering combines human guidance with AI assistance and real-world practice. He said the company listened to 7,800 nonprofits while shaping the model. The shift changes the unit of value from content consumed to progress made.

In April 2026, that model gained another piece when Resilia acquired Lightful’s intellectual property, including BRIDGE. The six-month learning program had helped more than 3,000 nonprofits across more than 80 countries build digital fundraising, storytelling and audience skills. For Resilia, the acquisition brings a tested curriculum and community into a platform increasingly focused on guided execution.

There is an aspiration hiding inside all this operational language. Geilhufe wants technology to preserve the care and judgment of the people using it, then relieve them of work that does not require either. He wants Resilia to be “practical, values-driven, and meeting changemakers where and when they are.” That is a restrained ambition for AI, and a demanding one. It asks the product to know when to help, when to show its work and when to get out of the way.

The thread from Katrina to Resilia is not a heroic story about software saving people. It is a story about coordination. People had the knowledge. Organizations had pieces of infrastructure. Families had urgent questions. The useful invention was a structure that allowed the pieces to work together.

Thirty years into his career, Geilhufe is still working on that structure. The interfaces are more polished, the models are more capable and the vocabulary has migrated from open source to SaaS to AI. The standard remains pleasantly stubborn: did the system help someone do the work? If the answer is buried under another login, another training deck and another dashboard, it has not passed yet.