Poynter: AI advice from journalists who stopped talking and started building

Alex Mahadevan at Poynter reports from the Hacks/Hackers and Poynter AIxJournalism Day at SXSW 2026, where newsroom practitioners — not researchers or consultants — shared what actually worked when they stopped planning AI adoption and started building.

Five lessons emerged from organizations that had moved past discussion into production. First, start with a specific pain point. Pew Research Center’s audience editor identified that her team spent 95% of its time writing formulaic social media posts and only 5% engaging with audiences. They built a WordPress plugin that auto-drafts those posts, freeing time for the work that required human judgment. The Upasna Gautam formulation: “Look at stuff you do every day. Where is the mundane work that you’re sick of doing?”

Second, define what AI is not for. Nebraska Public Media stopped using AI for editorial writing after quality proved insufficient, but kept it as a research and brainstorming tool. The boundary is explicit: corporate writing (grant applications, marketing copy) is acceptable; editorial writing is not.

Third, be specific about where the human stays in the loop. KQED’s editor-in-chief named the exact decision point where human oversight was non-negotiable — choosing which moments in a radio piece are most important — and built the workflow around that boundary rather than around a general principle.

Fourth, use AI to surface what audiences actually want. Texas Tribune trained a chatbot on their school voucher coverage. When readers asked questions the coverage hadn’t answered, reporters treated those questions as story leads.

Fifth, non-engineers can build real tools now. Mahadevan, not a software engineer, built a working fact-checking prototype using AI coding tools during a weekend. The observation from the session: “The gatekeeping vanished and people could just build.”

Relevant for journalists, editors, and newsroom managers evaluating where to begin with AI — particularly those whose teams have discussed the topic without yet making structural changes to how work gets done.