Social servicesExample project · Fictional nonprofit
How far can you take a one-person nonprofit?
A small association spends too much time coordinating volunteers, handling communications and preparing paperwork. The first goal focuses on just one part: making volunteer scheduling take fewer hours and fewer messages.
- Nonprofit
- Companionship association for older people
- Led by
- [TBD]
- Repository
- Opens with the project
Automatic
prepared and logged on its own
Approve
prepared on its own, she says yes
Human
she really steps in
- Thank a supporter for their help
- Sign up a new volunteer
- Grant deadline reminder
- Volunteer drops out · fill the shift
- Impact report for the city council
- Draft a grant application
- Reply to a recurring supporter
- Supplier invoice
- Move resources between programs
- Publish the annual report
- Conflict between two volunteers
- A family at risk
- Internal complaint
Today, everything lands on the same person.
The problem.
The problem, in the nonprofit's words"On Mondays I spend the morning finding someone to cover the visits that fell through. There are two of us for eighty volunteers, and every hour in the group chat is an hour I'm not with anyone."
27,962
organizations in Spain's social action third sector
Barómetro del Tercer Sector de Acción Social 2024 · Plataforma de ONG de Acción Social (Spain)
48.7%
had income under €30,000 in 2023
Barómetro del Tercer Sector de Acción Social 2024 · Plataforma de ONG de Acción Social (Spain)
Monday, 9 a.m. Three volunteers wrote over the weekend to say they can't make it this week. One of them was going to take someone to a medical appointment on Tuesday. The coordinator spends the morning looking for replacements in the group chat, checking who knows each older person and updating the spreadsheet. By noon she still has one visit uncovered, and the grant application due Friday hasn't been started.
Pick a task.
Three tasks, each with what you learn. Open one to see the steps and requirements.
- 01
DefineM
Interview the team and document their process
You learn: How to go from a conversation to a documented process with AI without losing what the person said.
Understand how shifts are arranged today and where time is lost.
What you do
- 1.Two or three 45-minute interviews.
- 2.Map the process, from availability to the visit.
- 3.Mark the three points where most time is lost.
Accepted when
- The coordinator confirms the process is accurate.
- Each friction point has a real example.
- No data identifies anyone.
- You deliver
- A two-page document with the process and the friction points.
- What credit it gives
- Your reviewed contribution record. If it's your first on this project, its badge too.Technique you show: Defining with AI: from interview to process.
- 02
DesignM
Design the availability screen
You learn: How to prototype fast with AI and validate it with the person who'll use it before writing code.
Let the coordinator assign visits and cover cancellations in a few steps.
What you do
- 1.Design the weekly availability view.
- 2.Design assigning a visit and covering a cancellation.
- 3.Test the prototype with the coordinator for 30 minutes.
Accepted when
- She assigns and covers a cancellation without help.
- Works on phone and computer.
- Uses the association's own words.
- You deliver
- A clickable prototype and the test notes.
- What credit it gives
- Your reviewed contribution record. If it's your first on this project, its badge too.Technique you show: AI prototyping validated with the user.
- 03
BuildL
Connect to their tools
You learn: How to give a coding agent just the right context so it integrates without making things up, and how to review it.
Read and write assignments in the spreadsheet they already use.
What you do
- 1.Read availability from their spreadsheet.
- 2.Write assignments without breaking the format.
- 3.Notify the volunteer through their usual channel.
Accepted when
- An assignment shows up in the sheet in the usual format.
- A cancellation triggers a notice and is logged.
- No credentials in the repository.
- You deliver
- Code with instructions and automated tests.
- What credit it gives
- Your reviewed contribution record. If it's your first on this project, its badge too.Technique you show: A coding agent with context and tests.
When it's done.
The coordinator spends fewer hours a week matching shifts and covering cancellations, and fewer visits are lost for lack of a substitute. The association itself would measure it: hours before and after, and visits lost per month.
- Collect volunteers' availability in one place.
- Suggest visit assignments that the coordinator reviews and confirms.
- Flag a cancellation and suggest replacements who already know the older person.
- Work with what they already use: their spreadsheet and their messaging channel.
Weekly coordination hours and visits lost per month, before and at 90 days.
Risks and sensitive data
- That it becomes one more app nobody installs: it has to live in the spreadsheet and channel they already use.
- That it suggests substitutes the older person doesn't know: the coordinator always confirms.
- No health or personal-situation data about the elderly people: only neighborhood, schedule and who knows them. Nothing that identifies a volunteer leaves the project.
Data and license
Apache-2.0. Design and testing use invented data shaped like the real thing; the association's data would only enter the final test, with its permission and without leaving its systems.