INMA outlines seven steps for newsroom AI
- INMA published a seven-step newsroom AI adoption guide on May 13, drawing on recommendations from Newsroom Robots host and Canadian AI Journalism Network co-founder Nikita Roy. (inma.org) - Roy’s framework starts with workflow audits rather than training and warns against adding “yet another tool” outside Google Docs, CMSs or Slack. (inma.org) - INMA posted the guide on its Newsroom Innovation Initiative blog on May 13, where Sonali Verma summarized Roy’s April presentation. (inma.org)
INMA published a seven-step guide for newsroom AI adoption on May 13, framing the advice around a presentation by Newsroom Robots host Nikita Roy at an April meeting of the Canadian AI Journalism Network. The article, written by INMA Newsroom Innovation Initiative lead Sonali Verma from Toronto, said Roy’s checklist was aimed at getting newsrooms to use AI tools “happily.” Roy’s recommendations focused less on model selection than on workflow design, peer buy-in and integrating tools into systems journalists already use. (inma.org) ### Why did Roy say newsrooms should start with audits instead of training? Roy’s first recommendation was to begin with workflow audits, not AI training, according to INMA’s May 13 article. (inma.org) She said managers should map how work moves from idea generation through publication and distribution before deciding where AI belongs. Roy said the point of that exercise is to establish a baseline and identify where work actually slows down. INMA quoted her asking, “What does today’s baseline look like? How does work actually happen?” and describing the need to trace a story from idea to website and social platforms. (inma.org) ### What kinds of newsroom problems was the framework built to surface? INMA said Roy’s second step was to surface pain points and what she called “unlimited capacity” opportunities. The article said the audit should reveal where AI could help with real frustrations and ambitions rather than abstract experiments. (inma.org) Roy told the Canadian AI Journalism Network meeting that editors should ask what teams would do differently if they had unlimited capacity. INMA said that question helps staff themselves identify opportunities worth solving and reduces the risk of outside teams imposing tools that feel unnatural in editorial settings. (inma.org) ### How does the guide propose winning over skeptical journalists? Roy’s third step was to provide role-specific, evidence-based use cases, INMA said. The article said skeptical staff are more likely to respond to examples from colleagues doing similar jobs than to pitches from outside departments. (inma.org) INMA wrote that Roy urged editorial staff to explain to peers how AI was helping them do their work better. The article said journalists tend to trust people who understand their routines and concerns, which made peer testimony more useful than generic training sessions. ### Why did Roy warn against “yet another tool”? (inma.org) Roy’s fourth step was to integrate AI into existing tools and workflows, not force staff into separate systems, according to INMA. The article cited Google Docs, content management systems and Slack as examples of places where AI features could be embedded. (inma.org) Roy said, “The biggest issue I’ve seen with AI adoption has been: I don’t want to go to another tool,” INMA reported. She said the question for managers was not only how to build a new product, but how to fit AI into the workflows people already follow. ### Where does newsletter drafting fit into the broader AI push? (inma.org) HeyNews said on May 12 that it had opened public access to a voice-trained AI newsletter platform after more than a year of internal use producing nearly 600 issues across 10 formats. The company said its system learns from a publisher’s archive and can run separate writers for separate newsletter formats, including different editions for different days. (inma.org) That product description tracks with one of the operational questions raised around newsroom AI adoption: whether speed gains can be preserved without flattening editorial voice. (inma.org) HeyNews co-founder Eren Daskesen said, “A newsletter’s voice is not a prompt,” describing a workflow built around publication-specific drafting rather than one generic output style. ### Who is Roy in the newsroom AI ecosystem? Reuters Institute said in a May 9, 2025 profile that Roy is a data scientist, journalist and AI expert who has been explaining newsroom AI use cases through her Newsroom Robots podcast since 2023. (markets.businessinsider.com) Her Substack describes the publication as a weekly forum for conversations with AI experts working at the front edge of journalism. INMA’s May 13 post said Roy delivered the seven-step framework at a Canadian AI Journalism Network meeting in April. The guide remains available on INMA’s Newsroom Innovation Initiative blog under Verma’s byline. (inma.org) (reutersinstitute.politics.ox.ac.uk) (markets.businessinsider.com)