In July 2023, the American Journalism Project launched its Product & AI Studio to explore the smart application of generative AI and other technology within local news. This Studio, made possible by OpenAI with additional support from the Patrick J. McGovern Foundation, is sharing regular updates of its learnings. Here, we examine insights from an experiment within our cohort to innovate development operations.
Learning from Open Campus: An AI assistant for communications
“It’s amazing.”
This was the sentiment shared among a small group of leaders from American Journalism Project portfolio outlets who recently got a sneak peek at the AI-powered assistant built by Scott Smallwood.
Smallwood is CEO and co-founder of Open Campus, and he is part of our 2025 AI grant cohort of six newsrooms working together to find common solutions to shared revenue challenges. There is an opportunity to leverage AI to reduce the amount of time drafting reporting that conveys impact, progress and alignment.
Identifying an opportunity
At Open Campus, a nonprofit news organization that places higher education reporters in local communities, two people on staff manage the relationships across their wide range of supporters. Similar dynamics play out in other AJP portfolio newsrooms.
Smallwood often found himself spending evenings perfecting communications funder updates to better highlight their work, progress and impact. He suspected AI could be useful with these updates but hadn’t quite defined how.
A breakthrough came during lunch with a donor when she asked Smallwood why so many of her grant recipients struggled to give meaningful updates. She said what she needed most is for grantees to help her “connect the dots.”
“I’m trying to give people money, and I need to talk to my colleagues, but the updates don’t always make it easy,” he remembers her saying.
Smallwood realized he could build an AI assistant to generate the first draft of donor communications by layering donor-specific details with Open Campus’ updates.
Building a working prototype
Additional inspiration came from other grantees in AJP’s Revenue Experimentation cohort. As donor communications emerged as a major theme amongst grantees, AJP built a prototype of an AI tool to support grant applications. This sparked ideas about how a tool’s interface might function, and workflows it could support. Scott saw opportunities to broaden the job this tool could do.
After a few weekends and late-night coding sessions, Smallwood had a working version of a tool with the potential to transform how these reports are assembled and customized. Together with Director of Revenue Operations Paula Matias, as well as Editor-in-Chief and Co-founder Sara Hebel, they built a system that captures the genuine alignment between the outlet’s journalism and the priorities of their supporters.
The tool works by connecting the team’s structured inputs with organizational data and communications, and it includes these features:
- A user can select the type of communication they’re preparing, such as an update, report or proposal.
- There is a selection of writing styles to match different contexts.
- Interactive content refinement means adjustments can be made without regenerating from scratch.
- It entails a comprehensive donor management system.
- It stores organizational messaging and automatically pulls in new strategic communications.
- It matches impact stories and organizational messaging to specific funder priorities.
“What makes this different is that we’re not dumping in grant guidelines and hoping for the best,” Smallwood said. “We’re documenting in our own words how we’re actually aligned with each donor. The AI helps us surface connections consistently across all our communications.”
The result: outputs that are authentically from the organization because they’re built on the team’s insights. Matias can now generate a tailored first draft in minutes rather than hours, and although it’s early days, the Open Campus team hopes their supporters get clearer insights into how their support drives impact.
As his prototype took shape, Smallwood said, “I could feel we were close to having something that would actually be useful. Not just for us, but potentially for others facing the same challenge.”
For Hebel, the tool’s most valuable feature might be what it doesn’t do: “It doesn’t sound like a robot wrote it. It sounds like us.”
Here’s a short video explaining how it works.
Learning from real-world experiments: CPI’s journey with AI
In Puerto Rico, Centro de Periodismo Investigativo (CPI) took a different approach to their own set of donor approaches. Executive Director and Editor-in-Chief Carla Minet and Development Director Annette Ramírez Díaz launched four experiments to answer a critical question: How can AI deepen insights about donors, personalize outreach to donors, and streamline fundraising efforts?
The four experiments were:
- Donor database analysis: Using AI for segmentation, churn analysis and behavior prediction,
- Email subscriber analysis: Using AI to identify locations and donor behaviors versus non-donor behaviors,
- Fundraising messaging optimization: Using AI to craft a donor communications guide, and
- Research: Using AI to find high-net-worth individuals with ties to Puerto Rico and to then create an outreach plan.
CPI’s experiments revealed both the promise and pitfalls of AI for accurately examining data and providing insights that will educate their revenue strategies.
Through trial and error, the outlet discovered that ChatGPT needed more guidance than anticipated to analyze more than 6,000 donor records.
“We initially assumed ChatGPT would correctly identify what each column represented. That wasn’t always the case,” Ramírez said. They ran five different prompt refinements and learned to ask ChatGPT for help improving the prompts to get the desired results.
In another example, CPI used ChatGPT to examine data to form a list of potential high-net-worth Puerto Rican donors. When it included actor Pedro Pascal, who is Chilean, not Puerto Rican, the inaccuracy underscored the importance of verification at every step.
CPI’s strategic approach changes its fundraising strategy
Eventually, AI began to surface valuable insights within the data, such as:
- Major donor retention rates are as high as 76%.
- CPI’s small-donor base experiences the highest churn rate at 34–45%.
These and other insights shifted retention strategy to prioritize personal touches for mid-level and major donors while testing upgrade campaigns for smaller ones.
Document everything
Perhaps CPI’s most valuable discovery was the importance of documentation. After each test, Ramírez asked ChatGPT to summarize what went right and wrong, building an institutional memory of the AI learning journey.
“It helps us avoid repeating mistakes and serves as a record of what’s been done,” Ramírez said.
These experiments also yielded practical automation solutions, like sending anniversary notes to donors that mark their original donations. And the outlet created a communications guidebook to ensure CPI’s authentic voice is always prioritized in revenue messaging and ChatGPT prompts.
What’s next for the cohort
Both organizations confirmed AI isn’t about replacing human effort — it’s about enhancing it. For Open Campus and CPI, these tools act as “force multipliers,” handling time-consuming tasks to allow more time to nurture relationships.
Other organizations in the cohort are already personalizing these tools for their own teams, and as they continue to experiment, we aim to share revenue results and other developments.
