An AI tool to increase the Productivity of understaffed nonprofits

Project Brief

Humanitas AI is a company building technology to improve the social care ecosystem. This early stage start-up aims to use AI to strengthen the nation’s social safety net and help people around the world get the help they need.

Timeline

February - January 2023

Team

Margot Field | Product Designer
Ara Ravisankar | Engineer
Brandolon Barnett | Product Manager

GOAL

Create a tool to improve nonprofit professionals' workflows that utilizes AI

The Humanitas partner platform is meant to be a multi-faceted tool for nonprofit professionals. The AI Assistant was the first feature to launch while arrangements for other integral aspects were made. This case study demonstrates our efforts to increase usability and functionality of the product.

PROBLEM

Small nonprofits are understaffed and overworked. With existing staff experiencing burnout at high rates, nonprofits are struggling to acheive their goals.

SOLUTION

A customizable AI Assistant specialized to nonprofit needs to boost capacity

If done correctly, AI presents an opportunity to give these organizations the capacity to reduce their workload without the need to hire additional personnel. We utilized ChatGPT API to create a chatbot specialized for nonprofits.

Solution breakdown

Our final product

1. Simple chat interface with clear language

2. Capacity to upload and edit different examples of their work

3. Ample help content

TESTING

Moderated MVP Usability Testing

Moderated Usability Testing

Users were tasked to:

Create an account
Use the assistant to create 3 types of content
Extra task: Users must upload data to complete #2

Our TESTERS

5 participants, all women
Average age: 35
Average organization size: 25 people
Most had very limited chat gpt experience

“This is so much more custom than ChatGPT. It’s so much easier when you have a product that’s built for you, that you can manipulate and save the data into”

Tester PH

RESULTS

While generally positively received, there were clear areas for improvement

Average SUS score 75 (Good)
Raw scores: 87.5, 87.5, 87.5, 62.5, 50

Above: I recorded participant challenges and pain points and organized them into related categories.

Theme	Who owns this problem?	How urgent is this?
1. Confusion in parts of the data upload process	Design	Very
2. Users don’t know how to interact with assistant	Design, Product	Very
3. Users want more data flexibility	Design, Product	Very
4. Some assistant content is bad or wrong	Engineering	Very
5. Wants more data integration	Product	Less

THEME #1: Users don't know how to interact with the assistant

How might we encourage prompt exploration?

1. Added prompt ideas section

2. Created an ongoing email marketing strategy

THEME #2: Users want more data flexibility

How might we give users more ways to upload data?

1. Created an open category to add more diverse data options

2. Added a webpage connection option

THEME #3: Confusion in the Data connection PRocess

How might we increase users' System Knowledge?

1. Added more feedback to the upload and analysis process

2. Blocked off Assistant until analysis is complete

3. Added a tutorial module to homepage

Bonus iteration

Prompt Builder

I drew some sketches and came up with a v1 of an AI-Assistant prompt builder that walks the user through creating a prompt that would elicit a high quality response.

Constraint

The team decided not to market this product, limiting our impact measurement capacity

While exciting and useful to some, for most nonprofit professionals, this kind of tool is more of a “nice-to-have” product as opposed to a “need-to-have” product.
Given the budget limitations of nonprofits and high costs of OpenAI API use, there’s no clear path to monetization for this product without additional features added.
The co-founders’ experience with a past free product indicated to them that they needed to determine the next direction before investing in marketing.
As a result, our ability to track impacts was lacking.

Reflection

How I would have Measured Design Impact

Initiative	KPI	How to track
1. Added more feedback to upload process	Task success rate	Embedded feedback survey
2. Created an open category to add more data upload options	Adoption rate	Analytics software
3. Block off the assistant until data analysis is complete	Task success rate	Embedded feedback survey
4. Add a tutorial module to the homepage telling users what they need to do	Task success rate	Embedded feedback survey
5. Add Prompt ideas section	Adoption rate	Analytics software
6. Create an email marketing strategy	Email Click-rate, Time between actions	Analytics software

REFLECTION

How I've grown from this project

Going through the full process of product strategy to launch taught me so much. It was massively challenging to be a founding designer in a rapidly evolving space. I gained experience in the following topics:

Design Strategy
UX Writing
Design Systems

Small Team Collaboration
Logo and Branding Design
Video and Motion Design

Being a Solo Designer
Designing with AI
Accessibility