You learned to prompt, and your organization learned to spend. Unfortunately, few organizations have learned to connect the two. That is where the AI4Agile online course comes in.
An AI operating capability exists when delegated work can be reproduced without its original creator, meets an explicit quality standard, follows a defensible execution path, has a named owner, and is inspected often enough to detect drift. The AI4Agile Online Course V3 teaches practitioners how to build one, independent of a particular AI model, and still, there is no coding required. The course is in English. 🇬🇧
TL; DR: Dangerous Agile Myths — Food for Agile Thought #553
Welcome to the 553rd edition of the Food for Agile Thought newsletter, shared with 35,462 peers. This week, Henrik Mårtensson dismantles seven dangerous Agile myths, showing that fat-tailed cycle-time data invalidates the use of story points. Teresa Torres and Petra Wille question whether support tickets can replace story-based interviews, while Roman Pichler pushes visions beyond feature lists toward purpose. Turning to AI, Laura Summers finds LLM-assisted coding replaces building satisfaction with supervision fatigue, Benedict Evans sees foundation models becoming commodities, and Satya Nadella urges firms to own their learning loops before providers capture proprietary knowledge.
Next, John Cutler reframes software assets through a portfolio lens, asking whether AI makes you faster or moves you faster in the wrong direction. George Sivulka and Arvind Narayanan both place the bottleneck in management, not model capability. On the human side, Sean Goedecke redefines engineering politics as knowing who holds power and making contributions visible, while Steven Sinofsky compares Chicago Law School’s AI ban to Harvard’s 1982 computer ban, arguing such restrictions never last.
Lastly, Pavel Samsonov argues that product empathy rings hollow without respect, a gap LLMs deepen by pushing error correction onto users. Thomas Squeo and Matt Kamelman trace enterprise AI failure to missing governance, not weak models. Susan MacKenty Brady, Stuart Kliman, and Leslie Smith name four leadership traps quietly eroding trust. Finally, Dave Rooney rethinks story slicing when AI handles large tasks, and Tristan Kromer notes AI accelerates experiments but cannot pick the right question.
Your team already has rules for using AI. Some live in templates, some in habits, exceptions, and one person’s memory. The AI Working Agreement puts the decisions that matter in one place: what the team delegates to AI, what stays human, what must be reviewed, what never enters a model, who owns which workflow, and how the agreement changes. Write it, and a new colleague can read your team’s AI decisions on their first day, while the decisions stay when someone leaves.
Thesis: Team-level AI governance fails more from uncodified judgment than from missing policies. The AI Working Agreement turns scattered AI decisions into one inspectable artifact, so a team can onboard people, survive departures, and challenge its own habits before those habits harden into risk.
TL; DR: AI Creates Jobs? — Food for Agile Thought #552
Welcome to the 552nd edition of the Food for Agile Thought newsletter, shared with 35,468 peers. This week, Ramp Economics Lab and Revelio Labs report that heavy AI adopters grew headcount by 10%, yet Charity Majors insists only honest feedback loops turn adoption into results. Alex Karp questions the economics entirely, calling token pricing fundamentally broken. Pavel Samsonov and Jerry Colonna both argue that speed without trust or judgment produces waste, while Janna Bastow reminds us that roadmap dates are comfort objects that mask the need for outcomes.
Next, Jeff Gothelf proposes that when AI makes building nearly free, teams should prioritize learning value and reversibility over effort. Kyle Poyar believes the resulting cost crisis is self-inflicted and offers a five-step spending fix. Yanli Liu warns that even working tools like Claude Skills silently rot without maintenance, while Addy Osmani suggests engineers must own accountability as agents handle execution. Also, John Cutler recommends that overthinkers disconnect self-worth from work entirely.
Lastly, Paweł Huryn frames the 2026 AI PM roadmap around whether agents run on your work or inside your product, while Alberto Romero raises a stranger question: why do AI models keep inventing their own languages? Fabian Metzeler and McKinsey colleagues distill seven truths from 15 AI-native companies, yet Cris Beswick warns most transformations stall when organizations skip differentiated innovation. Finally, Ant Murphy proposes a two-question test to tell actionable metrics from noise.
You moved your routine AI work to a cheaper model, so you think the cost question is handled; however, often, that is not the case. The decision lives in one person’s head and produces nothing that the person accountable for the invoices can read. Worse, it is an architectural choice nobody documented. The AI Routing Policy is the missing artifact of Stage 2 of the Delegation Lifecycle: it records which execution path, from a cheaper model to a frontier model to plain code, handles each class of work, what counts as good enough output to meet the AI Definition of Done, and who owns the call. The skill it needs to work is one you already have: You write acceptance criteria.
Thesis: An AI routing policy is not about picking a cheaper model at the moment of executing an AI task. It is a written, repeatable team decision that assigns each task class to the cheapest sufficient execution path: a model, human review, deterministic code, or no automation. Paired with a minimal routing log, it creates the spend-by-task-class record that your finance team will eventually request. You can draft the first three lines in twenty minutes.
TL; DR: AI Confidence Theater — Food for Agile Thought #551
Welcome to the 551st edition of the Food for Agile Thought newsletter, shared with 35,473 peers. This week, Elena Verna calls out AI confidence theater and asks teams to show real workflows, which pairs well with Teresa Torres and Petra Wille’s advice to start AI adoption with one messy to-do item. Also, Janna Bastow brings the same discipline to alignment meetings: clarify decisions before vague input becomes commitment. Anthropic frames Fable 5’s return as governance, while Alberto Romero questions the safety bargain, and Mike Cohn redirects failed Agile blame toward broken conditions.
Next, Aakash Gupta and Shubham Saboo treat PM work like code, while Hamel Husain extends that discipline to AI evaluation: track changes, show provenance, and make review paths obvious, and Tomasz Tunguz adds the cost pressure that will force selective adoption. Joost Minnaar reminds teams that rituals without shared power rot into theater, and Anthropic frames Claude Fable 5 as a teammate needing clearer boundaries.
Lastly, Ethan Mollick sees AI work shifting toward agent management, while Peter Yang expects model portfolios to reshape software economics. Charity Majors argues that leaders must support learning rather than demand unpaid adaptation, and Gergely Orosz reminds us that reinvention beats nostalgia. Finally, Abraham Thomas ties lasting progress to data quality that delivers real business outcomes by matching fitness for purpose with measurable value rather than relying solely on checklists.
