You’re reading The Steady Beat, a weekly pulse of must-reads for anyone orchestrating teams, people, and agents across the modern digital workplace – whether you’re managing sprints, driving roadmaps, leading departments, or just making sure the right work gets done. Curated by the team at Steady.
Safety by proxy
People who don’t feel safe raising a concern at work have found a new route: ask an AI to make the case and just forward the output. Amy Edmondson and 3M’s Jayshree Seth call it “safety by proxy.” The concern gets heard, but the person doesn’t have to own it and accept a reputational risk. One board director explains the appeal: “If it’s AI that says it, then it’s out there, without being tied to a particular director.” It’s the first of four ways the authors see people using AI’s apparent neutrality to say what they wouldn’t say themselves. People also use AI to phrase a concern outside their expertise. Or they test an argument with AI before making it in person. And leaders ask AI for the strongest case against their own decision, because nobody who reports to them will actually give it. The author’s main advice to managers: when a concern reaches you through AI instead of a person, ask why nobody felt safe saying it directly. And ask AI the hard questions before you announce a decision, not after.
— Harvard Business Review, 15m, #leadership, #ai, #management
On-site AI
Many companies are running two expensive programs at once: getting people back into the office and getting them to use AI. Phil Kirschner went looking for a company that uses its back-to-office move to speed up AI adoption on purpose and couldn’t find one. But the evidence that it could work is out there. Gensler is seeing its first significant rise in in-office learning in about 20 years, and its explanation is that with AI, “there’s no manual” so you learn by watching colleagues. Cisco’s numbers fit too: employees in the office three or more days a week are more likely to use AI. Cisco also found its most active AI users trusted their teams less than occasional users, and concluded that AI “can unintentionally create isolation when adopted individually rather than collectively.” Kirschner’s read is that the difference is whether people learn AI alone or together.
— The Workline, 7m, #ai, #transformation, #leadership
Look at your feet
Thorsten Ball posted a list of what he believes about the future of software development, and it blew up. Code review is dead, because humans won’t find bugs in model-written code in any reasonable time. Unit tests might be next. Writing code by hand becomes a niche craft: “Yes, there are still Italian shoe makers around. But look at your feet.” The craft of building software matters more than ever, which he defines as knowing how to solve business problems with it, when to ship, and how to get feedback. Most bugs won’t be coding bugs. They’ll be “you asked for the wrong thing” bugs. The PM, design, and engineering triad stops making sense, and so does Agile. Engineers who act as “meat proxies,” passing tickets to agents and reporting back, lose their value. “Good” code may not matter, because that idea rests on humans needing to work with it. And it will all take a generation to play out. People still get paid to maintain websites built on Microsoft’s late-1990s ASP framework, and people will still get paid to write code in 10 years. His question is whether you want that job.
— Thorsten Ball, 6m, #engineering, #ai, #strategy
The overhang
Ethan Mollick’s point is that the debate over future models misses what current ones already do. He had GPT-6 Astra turn the 1977 text adventure Zork into a playable 3D game, and Fable 5.1 reconstruct Umberto Eco’s 30,000-book library in 3D from videos, photographs, and two catalogs, placing about 5,000 books it could identify and marking each one certain, guess, or unknown. Asked for a trailer for his upcoming book, GPT-6 opened Blender, built an animated scene, wrote a script, generated voices and music, and delivered a film 45 minutes later. Each of these would have taken weeks of human work. He calls the gap between what the models can do and what almost anyone does with them the capability overhang, and he argues that closing it depends on four human advantages. Deep knowledge, so you can see where the AI went wrong. Wide knowledge, so you know what to ask for. Taste, because making things is now cheap and selecting among them is the scarce skill. And agency, the willingness to try things when nobody knows yet what AI can do.
— One Useful Thing, 10m, #ai, #productivity, #leadership
Judgement lapses
Maggie Appleton builds AI prototypes at GitHub Next, and she no longer reads the code her agents write. Instead, she writes a detailed spec, including how the agent should check its own work, and judges the result. She also names two ways working with agents goes wrong. The first is planning fatigue: “An agent grills you with a set of choice A, B, or C questions a hundred times over. By question 20, you’re quite tired and your brain starts shutting down.” So you start accepting whatever it recommends. The second is what she calls “capability gaslighting.” A model nails a task one day, convinces you it’s an expert, and fails the same task the next. As agents take the code off your plate, relentless judgement and verification becomes paramount.
— The Pragmatic Engineer, 87m podcast, #engineering, #ai, #productivity
Echo of the week
Echoes are AI agents in Steady that automatically gather and deliver work context to teams on a schedule – answering recurring questions about progress, capacity, and coordination so you stop burning hours assembling the same information manually.
The conversation you’re avoiding. Every Friday at 3:00 PM, this Echo reads the week’s check-ins, blockers, and goal updates and finds the conversations you’ve been putting off: a teammate who has gone quiet, a blocker nobody owns, a goal drifting without comment. It names the one that matters most and suggests how to open it.
The human-agent teamwork OS
Teams rely on two coordination loops to function: a big-picture loop connecting plans to progress, and a ground-level loop keeping teammates in sync.
Running those loops was already a marathon of meetings, chat threads, dashboards, and manual toil. Pile on flatter orgs, exponential output, and AI agents shipping 24/7 — the old way can’t keep up.
Steady removes the coordination bottleneck by running both loops for you. Working in the background, Steady distills updates and activity into targeted context for everyone on the team — human and agent alike. Full visibility, tight alignment, zero overhead.
The outcome: high-performing teams that deliver the right work, not just more of it.
Learn more at runsteady.com.