The intern test
Automation pays back when it removes a repetitive, low-judgment task. It does not when the task needs real judgment, runs too rarely to cover the setup, or needs a person to check every output anyway.
Would you trust a part-time intern with it and a one-page guide? If yes, automation probably can do it. If no, it will make a confident mess at scale.
What to pilot and what to skip
| Worth piloting | Not worth the meeting |
|---|---|
| Intake and triage of forms, emails and applications | Regulated decisions where a mistake costs more than the time saved |
| Pulling data from PDFs, receipts and grant reports | Workflows you run five times a year |
| First-draft donor thank-yous a person edits | Process problems dressed up as tool problems |
| EN and ES translation and alt-text drafts | Donor-facing writing that should sound like a person |
| Internal Q&A over your policies and past reports | Anything you cannot check before it ships |
Data pulling is the classic first pilot. Fort Mason reruns one n8n pipeline for listing data instead of cleaning a new export by hand.
What it costs
Build prices from my current list. Tool subscriptions usually add $20 to $200 a month.
| Offer | Price | What you get |
|---|---|---|
| Workflow automation | From $1,500 | One manual process, running on its own. |
| AI chatbot or agent | From $2,000 | Answers from your own documents, handed off to a person when needed. |
| Audit and scoping | From $500 | A written plan and fixed quote. Credited to the build. |
A custom agent pays back when you have real volume: hundreds of questions a week, support in two languages, after-hours coverage. As an experiment, it rarely does.
Scoping the pilot
Pick one to three workflows from the left column and run a four to six week pilot with before and after numbers. The full rollout (inventory, scoring, policy, scale) is in the AI strategy guide, and the 50 Questions Before Your Nonprofit Uses AI (opens in a new tab) checklist covers the vendor questions to ask first.
Common questions
- What about privacy for our donors and clients?
- Use business plans with a data-processing agreement, or EU-hosted or self-hosted options. Never put donor or client personal data into a free consumer tool. Write the AI policy before you deploy.
- Will this replace staff?
- On small teams it almost never does. It removes drudge work so staff spend more time on the mission. If the plan is to cut headcount with AI, you are measuring the wrong thing.
- How do we know the pilot worked?
- Pick two or three mission metrics before it starts: people served per staff hour, grants submitted per quarter, donor contacts per month. Measure at the start and at 90 days. If they did not move, the pilot did not work.
By Andrés Sepúlveda Morales. Top Independent on Contra, rated 4.99 (opens in a new tab) by clients and collaborators, and a speaker on AI in English and Spanish.