About this casebook.
This casebook documents what six small and mid-size newsrooms built, tested and learned during their strategic artificial intelligence (AI) adoption journey. This includes three newsrooms in Latin America, and three in East Africa:
- Revista Anfibia (Argentina)
- Chequeado (Argentina)
- La Diaria (Uruguay)
- Lake Region Bulletin (Kenya)
- Radio Domus (Kenya)
- Orkonerei FM (Tanzania)
All six took part in Advancing Ethical and Strategic AI Adoption in Newsrooms, a programme run by the Thomson Reuters Foundation (TRF) in partnership with the Patrick J. McGovern Foundation. The partnership equips newsrooms to integrate AI responsibly into their editorial and business operations. The programme combines in-person training and mentoring on two tracks, (1) AI solutions and (2) policies that govern them.
The cases lay out what problems these newsrooms were trying to solve, their AI solutions, and guardrails they put in place for the ethical and responsible adoption of the technology. The objective of this casebook is to help other newsrooms weigh what AI could do for their team, and potentially build on what the six newsrooms already found, rather than starting from scratch.
How to use this casebook
The casebook comes in five parts:
- Part I is organised by theme, opening with a featured case that walks through editorial or operational problems each newsroom wanted AI to solve, resulting AI solutions and how they were built, results (including what did not work), insights from real use, and cost in both money and staff time. Shorter spotlight sections cover additional examples under the same theme.
- Part II synthesises how all six newsrooms are governing AI, from the risks they identified to the guardrails they built.
- Part III retraces how seven of the prototypes were built, one page per build.
- Part IV synthesises the key takeaways from the casebook.
- Part V includes a glossary of key terms.
How we built this casebook
This casebook is built from interviews conducted by TRF with the six newsrooms in June 2026, and their written AI policies. Where attribution could breach a newsroom’s privacy, examples are anonymised.
Interviews were conducted on Microsoft Teams and transcribed automatically. AI tools were also used to help extract, cross-check and draft the casebook. Every claim was verified against the source interviews and documents, and all content was reviewed, edited and approved by the TRF team.
What the cases show
A pattern holds across the six newsrooms. AI earned its place on narrow, repetitive tasks, like turning a weather bulletin into an article or reading thousands of comments. The affordable route was almost never building from scratch, but configuring an existing tool. And every tool that stayed useful had both a human checking its output and a named person responsible for keeping it running. While Part IV includes the full synthesis, Part I shows how these lessons were learned, including what didn’t work, what each solution cost, and what each newsroom would do differently.