TL; DR: AI Watermarks — Food for Agile Thought #557
Welcome to the 557th edition of the Food for Agile Thought newsletter, shared with 35,378 peers. This week, Anthropic details the AI watermarks and C2PA metadata Claude now attaches under the EU AI Act, while Dror Poleg shows machines spot each other through word statistics that every writer carries. Detection remains messy, and so does judgment: Marty Cagan revisits Benedict Evans on problems AI never solves; John Cutler swaps prioritization frameworks for 12 tension prompts, and Vasuman suggests counting automated work rather than AI adoption based on vanity metrics. Also, Nigel Thurlow traces delay to variation rather than to people.
Next, Johanna Rothman counts running tested features, not activity, and pushes teams to finish aging work before inventory eats money. Ant Murphy goes further: value appears only after delivery, so prioritization interrogates confidence; remember ‘thinking in bets?’ Then, Itamar Gilad warns AI produces work nobody can judge, blurring roles and inflating certainty; Tim O’Reilly unpacks Drew Breunig’s prompt debt, which traps teams on old models, while Sunil Pai’s Cassandra agent watches Slack and speaks only when consensus looks wrong and rocking the boat seems appropriate.
Lastly, Matthew Hodgson notices agents spend by the second while budgets renew yearly, so he wants a per-agent cap now, while Meryem Arik finds similar waste in inference bills. Mark Levison keeps the human ledger: GenAI-triggered human skill loss is a choice, so pick which skills must stay sharp. Finally, Sebastian Ankargren, Joel Persson, and Mårten Schultzberg show LLMs replace A/B test users only under unverifiable assumptions.
Disclaimer: I belong to those who read Charniak/McDermott’s book on “Artificial Intelligence” decades ago. Of course, I make use of AI, for example, for proofreading, challenging story arcs and article structures, or summarization. It is a tool, and a powerful one… 🤷♂️
🎓 🇬🇧 The A3 Delegation System Founding Workshop — September 28-29, 2026
Your team already delegates work to AI: reports, research, customer feedback analysis, stakeholder communication, or parts of operational workflows.
But can you answer these questions without improvising?
- What may AI decide, and what must remain a human decision?
- What does “good enough” mean for this particular work?
- Who verifies the result before somebody acts on it?
- Who checks whether the delegation still works after the model or workflow changes?
If those answers live in one person’s head, or nowhere, your problem is no longer prompting. You have a delegation problem.
The A3 Delegation System gives you a practical way to decide what AI may do, hand over the work clearly, define acceptable results, and inspect the delegation over time.
During two hands-on sessions, you will apply the system to a workflow. You will leave with a clear understanding of how to apply the A3 Delegation System to your workflows so that team members or stakeholders can understand, challenge, and continue your AI delegation work. Everything you learn is directly applicable to your situation the next day. The class is in English.
👉 Join the Workshop Now — $199: The A3 Delegation System Founding Workshop — September 28-29, 2026
Did you miss the previous Food for Agile Thought issue 556?
🗞 Shall I notify you about articles like this one? Awesome! You can sign up here for the ‘Food for Agile Thought’ newsletter and join 35,000-plus subscribers.
🎓 Join Stefan in one of his upcoming training classes!
🏆 The Tip of the Week: AI Watermarks
(via Anthropic): How Claude marks AI-generated content
Anthropic explains how Claude now marks AI-generated content under the EU AI Act. Models launched from August 2, 2026 embed invisible watermarks in text and attach signed C2PA provenance metadata to files. A detected mark does not prove authorship, and a missing mark proves nothing either.
🎯 Product
(via Silicon Valley Product Group): A Fresh Definition of The Product Role
Marty Cagan borrows Benedict Evans's framing: spotting the general problem behind the pain, finding a solution customers value, and making it work for the business. AI lowers thresholds, not problems.
: Tension-Based Prioritization (A Non-Framework)
John Cutler suggests that prioritization frameworks mostly produce agreeable narratives, while the relevant decisions lie in recurring tensions that teams keep avoiding. He offers 12 fill-in-the-blank prompts to surface them before any prioritization framework is introduced.
: Nobody’s Actually Prioritising ‘Value’
Ant Murphy suggests nothing is inherently valuable. Your vision and strategy set the context, and since real value only appears after delivery, prioritization is about interrogating confidence in your bets.
: The AI Competence/Judgement Gap
Itamar Gilad suggests AI now lets us produce work we cannot judge, pushing everyone into a danger zone where roles blur, overconfidence grows, and colleagues critique each other through chatbots.
🧠 Artificial Intelligence
: AI Can See Your Watermark
Dror Poleg explains how statistical word biases enable machines to identify AI-generated text and one another, including Anthropic's invisible watermark, and points out that every human carries a similar linguistic signature.
(via O’Reilly Media): Prompt Debt and 'Fighting the Weights'
Tim O'Reilly unpacks Drew Breunig's idea of prompt debt: the patches teams pile up fighting a model's defaults, which slow iteration, block collaboration, and trap them on outdated models.
: AI Adoption is a Myth
Vasuman finds AI use in large organizations is dumbbell-shaped: a few power users, most barely touching it. He suggests stopping adoption reporting and instead reporting how much work is actually automated.
: every company needs a cassandra: an ai agent for the socially expensive work of organizational dissent
Sunil Pai proposes Cassandra, an AI agent that absorbs the social cost of dissent, watching Slack quietly, keeping receipts, and speaking only when consensus looks wrong, and the decision is expensive to reverse.
(via Zen Ex Machina): Your AI agents are spending in real time and your funding governance is annual
Matthew Hodgson points out that AI agents commit spend by the second while budgets are approved annually. Typically, there is no running total, no cap, no clear owner, and he suggests introducing a per-agent threshold now.
🖥 💯 🇬🇧 AI4Agile BootCamp #8, August 27 – September 17, 2026
The job market’s shifting. Agile roles are under pressure. AI tools are everywhere. But here’s the truth: the Agile professionals who learn how to work with AI, not against it, will be the ones leading the next wave of high-impact teams. Therefore, Stefan created the AI4Agile BootCamp.
So, become the professional recruiters‘ first call for „AI‑powered Agile.“ Be among the first to master practical AI applications for Scrum Masters, Agile Coaches, Product Owners, Product Managers, and Project Managers. The AI4Agile BootCamp is in English.
Learn more: 🖥 💯 🇬🇧 AI4Agile BootCamp #8, August 27 – September 17, 2026.
Customer Voice: “Last week, I finished the 𝗔𝗜 𝗳𝗼𝗿 𝗔𝗴𝗶𝗹𝗲 𝗣𝗿𝗮𝗰𝘁𝗶𝘁𝗶𝗼𝗻𝗲𝗿𝘀 course. And I’m mutating… It started on the train. I was scrolling through my messages, half-distracted, when a newsletter from Stefan Wolpers popped up. Stefan, a deep thinker with a hands-on attitude, was launching a new course. A pilot cohort. The mission: explore how AI can actually support us as agile practitioners. I couldn’t resist. I tapped: “𝘚𝘪𝘨𝘯 𝘶𝘱”. What followed were four bi-weekly sessions. Four intense afternoons. Full of exploration, experimentation, and practice. […] At the beginning, Stefan said that 𝘫𝘶𝘴𝘵 𝘴𝘪𝘨𝘯𝘪𝘯𝘨 𝘶𝘱 𝘢𝘭𝘳𝘦𝘢𝘥𝘺 𝘱𝘶𝘵𝘴 𝘶𝘴 𝘢𝘩𝘦𝘢𝘥 𝘰𝘧 𝘮𝘢𝘯𝘺 𝘱𝘳𝘢𝘤𝘵𝘪𝘵𝘪𝘰𝘯𝘦𝘳𝘴. That sounded like a big statement. But somewhere along the way, I noticed a shift… an emerging superpower in how I approach my tasks with AI.⚡And now, as my AI-mutation continues, I catch myself wondering: 💭 𝘏𝘰𝘸 𝘥𝘰 𝘐 𝘶𝘴𝘦 𝘈𝘐 𝘵𝘰 𝘴𝘢𝘷𝘦 𝘵𝘩𝘦 𝘢𝘨𝘪𝘭𝘦 𝘸𝘰𝘳𝘭𝘥?” (Ilya Zaytsev, Leading Agility at HUGO BOSS.)
➿ Agile & Leadership
: The Hidden Reason Work Keeps Taking Longer
Nigel Thurlow suggests delay is rarely a people problem. Unnoticed variation creates queues and rework, while averages and longer horizons hide it. Better forecasts will not fix an unstable system.
: Skill Loss and GenAI – Make It a Choice, Not an Accident
Mark Levison suggests GenAI skill loss is not inevitable. Rather, it is a choice, since the tools are right often enough to lull you into it. Therefore, decide deliberately which skills to drop and which to keep sharp.
: The Same Side of the Table
The author proposes one cardinal rule for managers: never leave your team's side of the table in a meeting, and offers graceful ways to intervene when someone is visibly struggling.
📈 Survey Result: From Agile to a Product Operating Model
Between August 2 and August 10, 2026, 48 practitioners participated in my Agile to Product Operating Model (POM) survey, which tries to shed light on what is actually changing.
Let me summarize the answers for you: the reported transformations change decision-making less than the Cagan framework suggests. Where respondents report improvements, they appear in delivery and collaboration rather than in business results. Unfortunately, the human side of the transition is the least encouraging part of the answers.
Thesis: Product operating model transformations mostly change vocabulary and organizational structure while leaving the decision system, who decides what gets built, on what evidence, at what speed, largely untouched. AI is changing product decisions independently of POM transformations.
Learn more: From Agile to the Product Operating Model: What Practitioners Say Is Actually Changing.
🛠 Concepts, Practices, Tools & Measuring
: How Flow Exposes the Reality and Costs of Progress, Part 2
Johanna Rothman suggests progress means running tested features, not activity metrics. Tackle aging work first, cut WIP, and decide faster, because inventory is no asset and old work costs money.
(via InfoQ): 📺 Producing the World's Cheapest Tokens: A How-to Guide
Meryem Arik suggests most organizations overpay for inference because their stack ignores the use case. Instead, trading latency for cost through hardware choices, bigger batches, and smarter queuing cuts token bills by 10x.
(via Spotify Engineering): When Can LLMs Replace Humans in A/B Tests?
Sebastian Ankargren, Joel Persson, and Mårten Schultzberg show that LLMs can substitute for users in A/B tests only under assumptions you cannot verify, and practically never when testing genuinely new ideas.
📅 Scrum Training & Event Schedule
You can secure your seat for Scrum training classes, workshops, and meetups directly by following the corresponding link in the table below:
| Date | Class and Language | City | Price |
|---|---|---|---|
| 🖥 💯 🇬🇧 August 27-September 17, 2026 | GUARANTEED: AI4Agile BootCamp #8, August 27 – September 17, 2026 (English; Live Virtual Cohort) | Live Virtual Cohort | €499 incl. 19% VAT (If applicable.) |
| 🖥 💯 🇬🇧 Sep 28-29, 2026 | GUARANTEED: A3 Delegation System Founding Workshop (English; Live Virtual Class) | Live Virtual Class | $199 incl. 19% VAT (If applicable.) |
| 🖥 🇩🇪 Sep 30-Oct 1, 2026 | Professional Scrum Product Owner Training (PSPO I; German; Live Virtual Class) | Live Virtual Class | €999 incl. 19% VAT (If applicable.) |
| 🖥 🇬🇧 October 15-November 5, 2026 | AI4Agile BootCamp #9 (English; Live Virtual Cohort) | Live Virtual Cohort | €499 incl. 19% VAT (If applicable.) |
See all upcoming classes here.
You can book your seat for the training directly by following the corresponding links to the ticket shop. If the procurement process of your organization requires a different purchasing process, please contact Berlin Product People GmbH directly.
📺 Join 6,000-plus Agile Peers on Youtube
Now available on the Age-of-Product YouTube channel to improve learning, for example, about AI Watermarks:
- Stop Writing Prompts. Let AI Do It for You — Hack #01, AI4Agile Online Course v2.
- Socratic Prompting — Hack #10, AI4Agile Online Course v2.
- Check Your AI’s Plan Before — Hack #7, AI4Agile Online Course v2.
- From Product Requirements to Experiments to Learnings — Supported by Generative AI.
- Never Accept an LLM’s First Offer — Improve GenAI’s Usefulness w/ Feedback Loops and Challenges.
✋ Do Not Miss Out: Learn more about AI Watermarks — Join the 20,000-plus Strong ‘Hands-on Agile’ Slack Community
I invite you to join the “Hands-on Agile” Slack Community and enjoy the benefits of a fast-growing, vibrant community of agile practitioners from around the world.
If you would like to join, all you have to do now is provide your credentials via this Google form, and I will sign you up. By the way, it’s free.
Help your team to learn about how AI Intensifies Work by pointing them to the free Scrum Anti-Patterns Guide:
