What AI is worth paying for in an NGO
Three jobs, in order of safety: drafting from your own records, reading what arrives, and answering staff from your own documents. And one to skip.

Much of my work has been for NGOs and associations, and the same thing shows up in nearly all of them: capable people spending their days retyping. Applications re-entered into a database, reports assembled from five spreadsheets, the same question about a policy answered by the one person who remembers. That is where AI is worth paying for, and in roughly that order of safety.
1. Drafting from your own records
The first draft of a donor report, a letter to members or a meeting summary, written from records you already keep. A member of staff edits it and sends it. This is the safest place to start, because nothing reaches anyone until a person has read it, and the saving is easy to measure: how long the first draft used to take.
My own platform works this way for couples. Its writing tool drafts speeches, thank-you notes and invitations from what the couple has already told it, and the couple decides what to keep.
2. Reading what arrives
Applications, forms, invoices and CVs, read into the fields of your system instead of typed in by hand. The AI fills the fields, a person checks them against the document, and only then is the record saved. How often it reads your documents correctly is something nobody can promise in advance, so this one starts with a pilot on your real paperwork: scanned, handwritten, crooked, all of it.
3. Answering staff from your own documents
Policies, past board decisions, the procurement manual, last year's grant agreements. A member of staff asks a question and gets an answer drawn from your documents, together with the document it came from, so they can check it. It reads and never changes anything.
The assistant on my own platform is built on the same principle. It answers from the couple's own plan rather than from what weddings are generally like, and it can only propose changes, never make them.
The one to skip
A chatbot that answers the public on your behalf, unsupervised. It is the one people ask for most and the hardest to get right, because every mistake it makes is made in your name, to a member, a donor or a beneficiary, with nobody checking first. If you want one anyway, start it on the questions your staff already answer from a written script, and keep a person reading what it says.
What to settle before anything is built
Three things, in writing, whoever builds it. First, where your data goes, and whether the provider of the model may learn from it. Personal data under the Data Privacy Act needs a written agreement before anyone outside your organisation handles it, and some of it, such as health information, I would keep away from the AI entirely. My own platform never sends guests' dietary needs to its assistant, on a lawyer's advice.
Second, what one use costs and the monthly ceiling, so a busy month cannot become a budget problem. Third, who confirms what the AI produces. The answer to that one cannot be nobody.
Enrique Lacambra III builds internal business systems from Cainta, Rizal.
Tell me what is broken and I will tell you if it is worth building. A 20-minute call, no charge, and I will say so if you do not need me.
