In January 2020, CREDO Action was shut down. I'd spent the previous eight years there, most of them as Director of Operations. The pandemic arrived a few weeks later, and the "what's next" question stayed open longer than I'd planned.
I started taking on consulting work later that year and named the practice CampaignHelp in 2021. The first clients were progressive nonprofits with the kinds of operations problems I'd spent the previous decade solving: email deliverability, hiring rebuilds, technology cleanup. Then around 2025 every client conversation started ending in the same question: what should we be doing about AI?
So I did the slow thing. I tried it on my own work first. I shipped small tools, ran small experiments, started a newsletter (Mission Control) about what I was learning. By the beginning of 2026 the AI work was becoming a much larger share of every CampaignHelp engagement. It was its own practice in everything but name.
The audience widened. The work didn't.
Grounded AI gives that work its own brand. The audience widened from progressive advocacy nonprofits (CampaignHelp's lane) to mid-sized nonprofits and membership associations more broadly. The orgs that don't have a Chief AI Officer because they're too small, but are too big to leave AI to whoever happened to read about it on the bus. That's the lane. I named it Grounded because that's the posture: feet on the floor, slow when slow is right, willing to say this isn't the tool you need.
What the practice is actually for
Most AI consulting in the nonprofit space comes from one of two places: tech consultancies who pivoted to AI and don't understand mission-driven orgs, or vendors with a tool to sell. The third option, a peer who's held the operations seat and now does this full-time, barely exists yet. That's the gap.
My clients are 50–500-staff service nonprofits, statewide associations, and the occasional foundation. The leaders I work with are EDs, COOs, CTOs (when they exist), and program directors who've been told to "figure out AI" without anyone telling them what figured-out looks like.
What I'm not
I'm not a software engineer by training. I ship production code with coding agents doing the heavy lifting, and I review every change before it ships. I don't have a PhD in machine learning and I'm not interested in pretending I do. The credential that matters here is I sat in your seat, and the honesty to know what's outside the lane.