Context
A civic initiative working with veterans — rehabilitation programs, legal support, documentation for benefit claims. A 12-person team: lawyers, psychologists, coordinators. They won a grant from an international organization — ~$80K over 12 months to build an internal CRM system that would help them manage their beneficiaries more effectively and generate reporting for donors. There are no technical people on the team. The deadline to launch the system is 9 months.
The pain
The project coordinator received a 40-page requirements template from the donor, plus proposals from 6 IT companies. Prices ranged from $35K to $120K. Timelines from 4 to 14 months. Each proposal describes “what’s included,” but she can’t compare them — every team uses its own terms, its own architecture, its own guarantees. One vendor proposes a “no-code” option on Bitrix24, another “custom development on React+Node,” a third “an adaptation of Salesforce.” The coordinator is afraid of getting it wrong: the grant doesn’t allow “let’s try again with a different vendor in a year” if the first one fails.
Approach
Six weeks. Weeks 1–2 — translating what the team actually does into the language of requirements. 6 internal interviews (a lawyer, a psychologist, 2 coordinators, the director, the accountant) — what hurts right now, how they keep their documents, how much time reporting eats up. This produced 23 concrete scenarios the future system has to fit. Weeks 3–4 — comparing approaches: 3 architectural options that genuinely suit the team (no-code on Bitrix, low-code on Airtable+Make, custom development). For each — a real estimate of what it would cost to launch plus 3 years of upkeep (not just the first year). No-code turned out cheaper to launch but more expensive over 3 years. Custom — the opposite. Weeks 5–6 — we rewrote the requirements around the recommended approach (low-code), and helped the coordinator understand how to talk to vendors: 9 right questions, 5 red flags. We met with the three shortlisted vendors alongside her as a “technical friend.”
Result
The team chose a vendor, and the project budget came to $42K instead of the initial $80K (the donor allowed the remainder to be reallocated to user training and learning materials). The vendor — a low-code team from Ukraine, on a contract with clear acceptance criteria and an exit procedure (credentials in the client’s hands from day one). The 3-year TCO forecast — $58K, including platform subscriptions. The coordinator got a document she can use every month to check whether the project is heading in the right direction.
What it taught
The biggest losses in grant-funded projects aren’t a “bad vendor” — they’re choosing the wrong architecture at the outset. It hurts NGO teams especially, because they don’t have the right to a second attempt a year later. Six weeks of discovery before signing the contract pay off many times over — not only in money, but in reputation preserved with the donor.
This is an illustrative composite pattern from 8 years of work in IT — not a specific client. AG is in a validation phase.