Most small nonprofits aren't behind on AI because they're unwilling. They're behind because no one has sat down with their staff, learned what the work actually is, and built something that fits it. This is a proposal to do exactly that — one person, one organization, one prototype at a time.
Every one of these is built for a specific person on your team. Keep going to see who gets what — and how it's made.
It's a term from the software industry: someone a technology company sends to embed directly with a customer — not to sell a product, but to sit beside the people using it, learn the work, and tune the thing until it fits.
Today the big AI labs — OpenAI among them — field these teams, but almost only for the giants. This proposal brings the same model to your size, independently: I get your team onto the right tools (say, a five-seat Claude Team plan), learn each person's work, then build the tools for you — or teach you to build your own.
When AI doesn't land at a nonprofit, it's rarely the software. It's the distance between a generic tool and the specific way your team actually works. Four things we hear:
We don't really know what's possible.
Nobody's shown us what it does for our intake forms, our grant calendar, our volunteer list.
We're afraid it's about replacing people.
Our staff are stretched thin and loyal. We won't make anyone feel like the goal is doing their job without them.
We can't afford a tech team.
No budget line for a developer — and the off-the-shelf platforms cost more than they save.
The generic tools don't fit how we work.
We've tried the all-in-one systems. We just end up bending our work to the software.
The missing piece isn't another tool. It's a person who understands both the tools and your work — and who builds the gap closed.
The “forward deployed engineer” emerged in live, high-stakes, constantly-changing software environments — the kind where slideware fails and only working software counts.
Not a product sold across a table. Someone who embeds in your team, learns the work from the people doing it, and builds and tunes real software against your real problems — in the room, in real time.
And it travels far past its corporate origins. A food bank, a land trust, a literacy program — each is a deeply specific operation with its expertise locked in its people and almost none of it written down. Exactly what the role was built for.
Every tool is built to help a specific named person do the part of their job only they can do — and delete the part a computer should've been doing all along. The director still picks the funders and tells the story. The coordinator still runs the program. Only the drudgery goes.
The multiplier up top is real but honest — task by task, not a blanket claim. What compounds is the hours handed back every week. Nobody becomes redundant; everybody becomes harder to replace.
And replacement is really a big-company story — it happens where there's slack to cut: bloated headcounts, redundant layers, work nobody can quite explain.
A small foundation or grant agency is the opposite case. A handful of people each do the work of five, wearing ten hats, with no slack at all — so AI has nothing to subtract. It just hands the overstretched their time back. The fear of being replaced is borrowed from a kind of workplace that isn't yours.
Same start: I sit with your team and build something real. What happens next is your call — and most teams want some of both.
You point at the work; I ship the tools — the fastest path to hours back, when no one on staff has time to learn the machinery right now.
Same build — then I open the hood: what the AI handed me, how I reshaped it into something worth trusting, the best practice for that specific kind of build. Your staff leave able to make the next one.
Most engagements do both: I build the first few for you, then teach you to carry it on. The arc is dependence → independence.
A real sequence, not a flowchart on a wall — it ends with you owning everything that gets built.
A one-on-one with every staff member: what's your week, what eats it, what do you dread? No jargon, no quiz — just the work, in their words.
Week 1The handful of places where today's AI genuinely helps — and an honest word where it doesn't. A short, ranked list, each tied to a person and the hours it saves.
Week 1–2Real, working tools in days — not decks. Put each in the hands of the person it's for, watch how they use it, reshape around what really happens.
Week 2–6Tools your staff can run themselves, documented. No lock-in, no dependence on me — if I vanished, everything keeps working. That's the point.
OngoingComposite portraits of the people inside a typical nonprofit — the kind I build for each real staff member. Read the week, then the build.
Rewrites the same three success stories into the format each funder demands. Loses days to boilerplate, deadline tracking, and digging up last year's numbers.
A proposal assistant that drafts a first cut from your past winning applications and a funder's own guidelines — plus a living grants calendar that surfaces what's due and what you've already said.
Messy intake spreadsheets, numbers hand-copied between five files, and a quarterly impact report assembled by brute force the night before the board meeting.
A pipeline that cleans intake data on the way in and a one-click report where the memo is the dashboard — charts, trends, and plain-language findings, generated from the live numbers.
One good story has to become a newsletter, three social posts, a donor appeal, and a press note — each rewritten by hand, on brand, on deadline, every single week.
A repurposing tool that turns one source story into every channel at once, in your voice — plus on-brand graphics generated in the styles you actually like, no designer required.
Matching volunteers to shifts and skills by memory and email chains. Onboarding repeated by hand for every new face. Thank-yous that arrive late, if at all.
A lightweight matching and scheduling tool, plus an onboarding flow that runs itself and warm, personalized thank-yous drafted the moment a shift is logged.
Hours of case notes after every client. A resource binder that's always out of date. The same eligibility questions answered from memory, one client at a time.
A private note-taker that turns a conversation into a structured record, and a searchable resource finder that matches a client's situation to the right program or referral in seconds.
Carrying the whole organization in her head. Board memos rebuilt from stale decks and spreadsheets. No single place that shows what's working across every program at once.
A living map of the organization — programs, grants, and outcomes pulled into one view a busy ED can actually hold — and board memos that update themselves from the real data.
None of these are off-the-shelf. Each is a small, sharp tool shaped to one person's week and handed to them to keep. That's the whole job.
Not theory — each one below has already shipped in a real project, and maps straight onto a nonprofit's day.
Data-driven apps and dashboards that open in a browser — no install, no platform fee — and can be shared with a link.
Shipped: an interactive map of a 1,700-plant property from a messy spreadsheet; a five-difficulty bird-sound training game with real audio.
Turning the messy real-world spreadsheets and exports an organization already has into clean, queryable, trustworthy data.
Shipped: a platform analyzing 84,000+ field observations with bias-corrected statistics and an automated daily refresh.
Claude used not as a chatbot but as an engine for drafting, scoring, sorting, and verifying — inside tools built for one job.
Shipped: a personalized digest that scouts live sources, ranks relevance with an LLM, and remixes it into a clean briefing.
Polished, on-brand reports, briefs, and pages generated straight from live data — and verified against the source.
Shipped: a pipeline that drafts, fact-checks against primary sources, and narrates long-form essays end to end.
Scheduled jobs that fetch, update, and publish without anyone touching them — quietly, every day, for free.
Shipped: automated daily data updates and recurring research digests for a working lab, no servers to babysit.
Every tool made to feel considered and human — this proposal is a fair sample — never like default "AI slop."
Shipped: bespoke sites, interactive study guides, and data visualizations with real typographic care.
The Phase 1 conversation isn't improvised — I've built a working practice around it: profile a person by what they do and how comfortable they are with tech, then map genuinely useful, no-code tools onto their real workflow.
I've done it for a foundation director, a literary scholar, a commercial producer, a publishing marketer, and a biotech attorney — each non-technical, each handed a short, specific set of things AI could build for them. A nonprofit's staff is the same exercise across one team — with working prototypes at the end, not recommendations.
The bias throughout: small and opinionated over big and generic. Cheap-to-run over subscription-heavy. Local and private when the data is sensitive. Built to be handed off, not held hostage.
Tools draft, sort, and suggest. A human always decides. Nothing goes out the door, or into a record, without a person who chose it.
For client records, donor details, or anything confidential, tools can run on your own machines and never send private data to an outside service.
Every tool, its code, and its documentation are yours to keep, run, and change. No lock-in, no platform you're forced to keep paying.
Where AI is unreliable for your task, I'll say so and we won't ship it. Numbers tied to your mission get verified against the source before anyone sees them.
The premise is augmentation. Every build targets the drudgery, never the judgment, relationships, or care that only your staff provide.
Default to free and low-cost tooling. The goal is to give hours and capacity back to the mission, not to add a recurring expense.
Three commitments, low first — and each runs in either mode: built for you, or taught to your team.
I sit with one team, learn the work, and ship one or two working prototypes. A low-risk way to feel the difference before committing to anything larger.
Best for: a wary board that wants proof, not promises.The full deployment: a conversation with every staff member, a mapped and ranked plan, and a set of prototypes built, tested in real use, and handed off across the organization.
Best for: an organization ready to change how the work feels.After the embed, a standing line to extend tools as needs shift, adopt new capabilities as they arrive, and keep everything running — without rebuilding the relationship each time.
Best for: keeping the tools alive as the work evolves.No commitment, no sales deck — just an hour to hear what your team does and where the days get lost. If there's something worth building, I'll show you what. If not, I'll tell you that too.