The AI-Ready Non-Profit Back Office · Part 03
The Maker's Revolution: From Concept to Custom Tool Without the Bloat
For years, a custom software problem had two answers: force-fit an expensive tool or absorb the manual workaround. Both were genuine options because there weren't better ones. Three things I've built recently — a bespoke CMS, an internal operations tool that replaced a $500/month subscription, and a two-hour build for my father — put some numbers on how much that has changed.
For most of the time I've worked in operations and technology, a custom software problem had exactly two answers. The first was to find the closest SaaS product that fit, accept that it solved about 60% of what you needed, build workarounds for the rest, and pay a monthly invoice for the privilege. The second was to skip the tool entirely and absorb the manual process — knowing it was inefficient, knowing it would still be there next year, and not having a better option.
Both of those were genuine answers for a long time. I'm not describing the 1990s. I'm describing 2022.
What has changed since then is not incremental. AI-assisted development has made it possible for someone with deep domain knowledge — who understands the problem, the constraints, and the edge cases — to build custom solutions without a development team. The translation from expertise to working software no longer requires a team of people with complementary technical skills. The domain knowledge remains entirely human. The translation process is now shared.
Three projects I've built in the past year put some numbers on what that shift actually looks like.
The New Economics of Building
The cost barrier has collapsed. The domain-knowledge barrier has not.
Chapter 01 of 03
Skip chapter introBefore the specific projects, it's worth naming the macro shift precisely, because the numbers are more striking than the abstract framing.
The force multiplier here is not just speed. It is the elimination of entire cost categories. When AI handles the technical translation between intent and code, a person with domain knowledge can build solutions that previously required a team. The specialist knowledge — understanding what the solution needs to do, what constraints it must respect, what edge cases matter — remains entirely human. The translation to working software is what AI accelerates.
This has direct implications for the non-profit sector, which has been told for decades that purpose-built infrastructure is a luxury it cannot afford. The economics of building have changed faster than most organizations' assumptions about what is possible. The question is no longer whether you can afford to build something custom. The question is whether you have someone who understands the problem well enough to describe it clearly.
AI-assisted development uses AI coding agents to translate intent into working code. The human provides domain knowledge — what the tool needs to do, what constraints matter, what edge cases exist. The AI handles the technical translation. Neither can do the job alone.
Three Projects, Three Case Studies
The macro shift, made concrete. Specific builds, specific numbers, specific outcomes.
Chapter 02 of 03
Skip chapter intro
Escaping the Template Trap
Every organization that has tried to build a professional web presence using a commercial website builder has hit the same wall. The templates look great in the demo. The moment you need a specific data structure — a custom content management system, a dynamic form with conditional logic, a blog with rich visual components — you discover that the platform's flexibility ends exactly where your requirements begin. For the PF TECH website, I refused to accept that limitation. Using AI as an architect and planner, I mapped out the exact content requirements, data structures, and user flows we needed before writing a line of code. That context then let AI coding agents accelerate development of a fully custom content management system with built-in analytics, compliance features (PIPEDA-aligned cookie consent and data collection), and SEO optimization built into the data model from day one. The result is a platform I can extend in minutes, not days. I own the code and the data entirely. We are not renting our digital presence from a vendor with the power to change pricing, deprecate features, or sunset the platform. And the economics were significant: a comparable custom build through a digital agency would have cost somewhere north of $10,000 in design and development fees. Our AI compute costs were under $100. The website you are reading this post on was built this way — including the blog component that displays this post, the visual component library that renders the interactive elements, and the Supabase-backed CMS that publishes new content. Built by one person, in weeks, for less than the cost of a single agency meeting.
Sunsetting the $500/Month Monolith
The second project has the clearest financial case, because the before and after are both measurable in dollars. Our operations required managing client schedules across a legacy ticketing system that cost over $500 per month and included hundreds of features we never touched. The core operational problem — generating dynamic planning calendars that integrated scheduling data across multiple clients and timeframes — was not something the ticketing system handled well. Our Operations Lead was spending days each month manually building those calendars. Conservative estimate: $2,000 per month in staff time absorbed by a process that should have been automated. We built a replacement. The first goal was straightforward: solve the dynamic calendar problem. Using an automation workflow engine and AI-assisted development, I built a custom tool that generated the calendars automatically. Eight hours of development time. Less than $40 in AI compute costs. Once the calendar logic was working, the natural next question was what else the ticketing system actually did that we used. The honest answer, examined carefully, was very little. We replicated the core workflow in the same custom environment. The legacy system is gone. The $500/month subscription is gone. The $2,000/month labour overhead is gone. What we have instead is a lean, purpose-built tool that fits our actual workflow — no unused features, no workarounds, no monthly invoice for functionality we do not use. The build took one person less than a day. The ongoing savings are real and recurring.
The Two-Hour Build
The third project is the most personal — and, I think, the most instructive about what this shift actually means for people who are not founders or operations leads. My father creates videos about travel and history. The visual production is where he excels. Narrating in English, his second language, was a persistent friction point that caused projects to stall — complete in every other way, waiting on the voiceover. Commercial text-to-speech tools were either too expensive, too low-quality, or too restrictive in their licensing for his use case. The problem was specific enough that no startup had decided it was worth solving. It was enormous for the person experiencing it. In two hours, I built a custom AI assistant that does exactly two things: it helps him refine and polish his scripts with conversational feedback, and it generates high-quality audio narration via an AI voice API. No subscription fees. No confusing interface. No onboarding. It runs on pennies per session. He can now complete his projects. The creative work that was stalling is moving again. The part of this that tends to get lost in the productivity discourse is that the shift isn't only about operational efficiency or cost reduction. It is also about solving problems that the market has decided are too small to address — because the market works at scale, and your specific problem is not on any vendor's roadmap. Those problems are now buildable, by the person who understands them best.
What This Means for the Non-Profit Sector
The constraint is no longer technical skill — it is domain knowledge and the willingness to build.
Chapter 03 of 03
Skip chapter introThe implications for non-profits are direct and significant.
Every sector-specific operational problem is now a reasonable candidate for a custom build rather than a permanent workaround:
- Grant reconciliation workflows that do not quite fit any existing tool
- Reporting formats required by a specific funder
- Donor communication systems that need to integrate with a legacy database
The constraint is no longer technical skill. It is the domain knowledge to describe the problem clearly, and the willingness to spend the hours building rather than tolerating.
The Mission Multiplier Program exists, in part, to help close this gap. If you are a non-profit practitioner who has accepted that technology decisions belong to the IT department or the board's technology committee, that assumption is worth revisiting. The tools have moved. The opportunity to build solutions that fit your specific operational reality is real, accessible, and closer than it has ever been.
Build in public
Monthly notes from a builder, not a pitch deck
What we're building, how long it actually takes, what it costs, what breaks. Written from inside the work.
By subscribing you join the PF TECH mailing list to receive Mission Multiplied posts and occasional related PF TECH updates. Monthly cadence. Unsubscribe anytime from any email. See our privacy policy at purposeforwardtech.com/privacy for how we handle your data.
Personal projects are professional preparation. The reps you build solving problems that matter to you personally make you faster and more capable at the operational problems that matter organizationally. Start with something you actually care about solving. The skills transfer directly.
Learn to build for your organization
The Mission Multiplier Program is a small-group coaching cohort for non-profit practitioners who want to build real competency with AI tools in a context that matches their work. Monthly 90-minute workshops, 15–20 participants per group, evolving curriculum, and first access to TERN capabilities as they launch. If you have been waiting for the moment to start building — this is it.
How did this land?
About the author
Greg Zatulovsky
Founder & CEO, PF TECH
Greg founded PF TECH to multiply the operational capacity of purpose-driven organizations. CPA with fifteen-plus years in non-profit finance, operations, and technology. Writes from inside the work — practitioner voice, not pitch deck.
More reading
Your Accounting System Needs a Bouncer, Not a Butler
Why the automation story needs a second chapter
There's a version of the AI-in-non-profits story being told right now that goes like this: AI will automate the tedious back-office work, free up your staff, and let you focus on your mission. That version is true — I believe it deeply and I'm building the tools to make it real. But there's a part of the story that isn't being told, and the gap between those two versions is where a lot of organizations are going to get hurt.
Resilience is a Verb: What the CPA Ontario Not-for-Profit Conference Revealed
Field Note · CPA Ontario NFP Conference 2025
In November 2025 I hosted the CPA Ontario Not-for-Profit Conference. The theme was Resilient Futures. Seven speakers, eight hours, topics ranging from AI adoption to board governance to front-line service delivery. What I took away was more urgent than the theme: the sector's biggest obstacle to building resilience is not funding or technical skills. It is the calculus that makes staying stuck feel safer than taking the next step.
Building a Better Engine: The PF TECH Origin Story
Origin Story · PF TECH
PF TECH didn't start at a whiteboard. It started in a filing cabinet — in 2013, watching a colleague print every email he received and file the paper copies in rows of cabinets along the wall. Nearly two decades later, that pattern is still the sector's central problem. Here's what I built because of it.