Blog
A collection of topics that keep me thinking.
26.10.2026
Cheap prototypes do not make product decisions cheap
AI makes prototypes faster, not product decisions easier. Use a learning contract to test assumptions before polished ideas enter the delivery queue.
21.10.2026
Your Work Is Waiting for a Decision Nobody Owns
Trace decision latency in software delivery by making ownership, evidence, handoffs, wait age, and escalation rules visible for one delayed item.
20.10.2026
Smaller Batches, Better Delivery? Find the Feedback Boundary
Use local flow and quality signals to find a useful software delivery batch size without imposing arbitrary story-point or ticket-size limits.
09.10.2026
Moving People Is Not Moving Capacity
Evaluate developer reassignment through team familiarity, transfer work, and delivery continuity before moving experienced people between initiatives.
07.10.2026
Does your approval still cover the software you deploy?
Check whether a software change approval still covers the deployed artifact and its conditions, without replacing judgment or required controls.
02.10.2026
How often should you review software delivery metrics?
Choose a useful review cadence for software delivery metrics by separating data collection, interpretation readiness, and urgent operational alerts.
05.10.2026
Your Postmortem Is Not Finished When the Document Is
Turn postmortem action items into testable risk-reduction work with an owner, review date, verification signal, and explicit closure decision.
23.09.2026
The delivery decision your portfolio keeps avoiding
A practical, evidence-informed way to review low-value software initiatives, examine escalating commitment, and make explicit pause or stop decisions.
22.09.2026
Your Jira board can show where time accumulates. It cannot tell you why.
Use Jira or Azure DevOps state history to locate long elapsed states, validate candidate delays with teams, and improve flow without individual targets.
16.09.2026
Your roadmap needs a budget for unplanned work
Production incidents distort roadmap forecasts when they stay invisible. Learn how to measure unplanned demand and set explicit displacement rules.
11.09.2026
Another Agile Transformation Is Not a Delivery Strategy
Agile transformation delivery performance cannot be inferred from a label. Test one constraint, one mechanism, and one management decision.
14.09.2026
What Should Your Agile Coach Be Accountable For?
Assess Agile coach accountability through delivery mechanisms, observable signals, and capability transfer rather than ceremony activity alone.
09.09.2026
Defining the Product Owner Is Not Enough. Define the Decisions.
Keep the shared Product Owner definition, then trace local decision ownership through real work to inspect gaps, overlaps, and escalation paths.
04.09.2026
Why incident recovery stalls between teams
Inspect ownership, authority and expertise handoffs in incident recovery without assuming one team model will lower MTTR.
04.09.2026
Your Delivery Forecast Is Missing the Work That Is Getting Old
Work-item age can focus a forecast review before a commitment fails. Learn how to inspect aging work without blaming people or trusting crude thresholds.
31.08.2026
AI coding saved time. Where did your attention go?
AI coding may reduce implementation effort while moving scarce human attention into verification, integration, coordination, and rework.
04.09.2026
The Hidden Recovery Delay: When Nobody Owns the Next Decision
Production incident recovery slows when decisions lack clear owners. Use a recovery contract to clarify authority, escalation, and verification.
31.08.2026
AI-assisted development can move the queue into review
A useful first question for an AI-assisted delivery system is: where did the queue move? Consider a hypothetical team. Pull requests begin to arrive faster after it adopts an authoring agent. The review column grows while the number of reviewed, integrated changes barely moves. A chart of PRs opened per week looks better. Confidence in the next release does not.
31.08.2026
AI-assisted development can move the queue into review
A useful first question for an AI-assisted delivery system is: where did the queue move? Consider a hypothetical team. Pull requests begin to arrive faster after it adopts an authoring agent. The review column grows while the number of reviewed, integrated changes barely moves. A chart of PRs opened per week looks better. Confidence in the next release does not.
31.08.2026
Your engineering dashboard is not a diagnosis
Learn how to combine delivery metrics, developer-experience signals, and workflow data into a testable engineering diagnosis instead of one health score.
31.08.2026
A faster coding tool can leave delivery unchanged
The AI adoption review starts with a convincing demo. A developer turns a short prompt into a working implementation. Another shows how an agent handled a routine change across several files. The room sees less typing, quicker first drafts, and fewer minutes spent searching for syntax.
31.08.2026
The queue behind a full capacity plan
Open the capacity plan. Every project has a name next to every role, nothing is unstaffed, and every line reconciles. Then sit in three standups on the same morning. The first team is waiting for the data engineer. The second is waiting for a security review. The third is waiting for the one person who knows how the billing service behaves under load. It is the same person, and often the same week.
31.08.2026
Your team is not slow. The work needs too many people.
Take the last change your team shipped. Ignore the code. Follow the route. A product decision. A schema review from a team in another department. A security check. A platform ticket for a configuration change the team is not allowed to make. A legal sign-off, because one field turned out to be personal data. A vendor contact who answers on Thursdays. A release slot.
31.08.2026
Every team hit its target. The product still arrived late.
The quarterly report says analysis finished on schedule. Development kept utilization high. Security review stayed inside its SLA. Operations protected stability and took no unplanned outage. The change reached the customer six weeks late. Nobody there is incompetent. Nobody is gaming numbers. Each function did what it was measured on, and the measurement was accurate. The problem sits one level up.
31.08.2026
Your AI agents can be right and your delivery system can still be wrong
When AI tools produce locally correct work faster, that speed does not automatically become system throughput. This piece separates local correctness from system correctness, walks the evidence from socio-technical congruence to recent AI productivity studies, and gives a Monday-morning flow check to find where the hidden integration work piles up.
31.08.2026
Your platform is the queue it removes
Most organizations now run an internal developer platform, but that says nothing about whether it helped. This piece reframes the platform as the queue it removes and gives a four-question flow audit: waiting, cognitive load, feedback, and safe self-service. Grounded in DORA 2025, a 2026 Frontiers review, and Google developer-productivity research, with the evidence caveats kept visible.
31.08.2026
Your developer tools are part of the delivery system
Developer tooling, dependencies, and CI/CD are part of your delivery system, not background IT. Unmanaged tool trust behaves like hidden work in progress, so a small trust queue keeps delivery moving without quietly accumulating risk.
31.08.2026
Scrum is shrinking. The work did not go away.
Agile Coach and Scrum Master job titles are losing market legitimacy, but the delivery work behind them did not disappear. This piece argues for defending the evidence-backed mechanisms instead of the label: responsiveness, stakeholder concern, continuous improvement, autonomy, and management support, drawing on Verwijs & Russo (2023), Tam et al. (2020), and two large-scale agile reviews.
31.08.2026
Your team already has an AI working agreement
Every software team using AI already runs on an unwritten agreement: who can paste what into a model, whether generated code gets normal review, and what an autonomous agent may read or trigger. This piece argues for writing that agreement down one workflow at a time and tying it to a delivery signal, drawing on Russo (2024), Song et al. (2024), and Alami & Ernst (2025).
31.08.2026
Your reliability target is a product decision
SLOs and error budgets turn reliability from an operations metric into an explicit product and delivery decision: agree per service how much unreliability you are willing to spend, and decide in advance what changes when the budget is gone.
31.08.2026
Your bonus system is part of your delivery system
In interdependent software delivery, individual bonuses and targets can quietly pull people toward their own local proxies and away from finishing shared work. Before redesigning boards or ceremonies, it is worth auditing whether your reward system matches the unit of work your delivery depends on. The compensation research is solid, but the direct software-flow evidence for bonus design is thin.
31.08.2026
Hidden work in progress: why busy teams still deliver slowly
One ticket in progress can mean five unfinished problem states in someone's head. Hidden WIP rarely appears in Jira or Azure DevOps, but it decides whether work moves. What interruption research really says, and a practical fragmentation audit for Delivery Managers and Agile Coaches.
31.08.2026
The help that makes teams helpless
The most useful person on a team can quietly become its bottleneck: when every blocker, decision, and escalation routes through one helper, the team never learns to clear the next one without them. This piece separates support that builds capability from support that only removes the immediate problem, with the research on agile role transfer and a flow audit you can run this week.
31.08.2026
The conflict you avoid becomes WIP
Avoided conflict does not disappear. It behaves like hidden WIP: deferred decisions, reopened tickets, quiet stakeholders. What the team conflict research supports, and how to build conflict-processing capacity into delivery work.
31.08.2026
Your AI pilot worked. Your delivery system did not.
AI pilots often look successful while delivery outcomes stay flat. The gap is compatibility debt: the distance between what an AI tool can do on its own and what your delivery system can absorb. A four-week diagnostic tests it before you scale.
31.08.2026
Your retrospective isn't broken. Your improvement system is.
Teams that run retrospectives every sprint and still don't improve usually do not have a meeting problem. The peer-reviewed evidence points downstream, to the system that converts action items into measured change.
31.08.2026
What is actually cutting the junior pipeline
Large-N payroll data shows young workers in AI-exposed jobs losing ground, but only where AI automates rather than augments. That fork is a delivery decision, not a procurement one, and a Delivery Manager can read it in a week.
31.08.2026
Merged is not shipped. The AI rework tax your dashboard isn't reading.
Recent empirical work shows that a meaningful share of AI-generated code is merged, counts on the dashboard, then quietly removed or left as a maintenance load. This article gives a Delivery Manager 5 numbers from existing Git history and static-analysis output to read the half of the delivery system that throughput hides.
31.08.2026
Conversation debt is slowing your delivery system
Most teams do not lack communication. They carry conversation debt: the accumulated cost of meetings and threads that never surface the information the work needed, showing up later as reopened tickets, late objections, and rework.
31.08.2026
Your backlog is full. Where is the value proof?
Many teams order the backlog but never close the value loop: what was learned from the last increment rarely returns into the next prioritization decision. A look at the evidence, plus a two-hour audit you can run on your last 10 delivered items.
31.08.2026
Code got cheaper. Decisions didn't.
AI made code cheap, but the queue of decisions in front of the work did not get faster. This article shows how to measure decision-wait in Jira or ADO and five moves a Delivery Manager can make this sprint.
31.08.2026
Your team does not need fewer meetings. It needs better designed ones.
Cutting meetings is the wrong lever. A 2025 field study shows meeting load follows an inverted-U, and the real failure is meetings that never close a decision. Two low-effort design moves the evidence supports.
31.08.2026
Your AI-assisted team got faster. Why did the forecast get worse?
AI-assisted teams ship code faster, but the delivery forecast often gets worse. Predictability lives in the shape of the cycle-time distribution, not in the estimate, and that is where a Delivery Manager has levers this sprint.
31.08.2026
The capability cost behind AI coding assistants
A randomised trial found that developers using AI on an unfamiliar library scored lower on comprehension and debugging, with no significant average efficiency gain. The capability cost of AI coding assistants is set by how you route the work, not by the license.
31.08.2026
They didn't cut Scrum. They cut what Scrum became.
In 2023 Capital One cut 1,100 "agile" roles and said the delivery practices belonged inside core engineering. This looks at what Agile work still produces measurable value, and how a Delivery Manager or Agile Coach should sell that work in 2026.
31.08.2026
Your AI bill is a flow problem, not a finance problem
Many teams treat a runaway AI invoice as a finance problem. It behaves more like a flow variable, and the levers you already use for cycle time and WIP work on the bill.
31.08.2026
AI did not remove your bottleneck. It fed it.
AI coding tools raise the arrival rate of work into your delivery system. If review, integration, or release is the real bottleneck, faster generation grows the queue instead of shipping more. The first move is to limit work in progress near the constraint, not to add more dashboards.
31.08.2026
Code review is not a checkpoint. It is how the team keeps up.
AI makes it cheaper to generate code, not cheaper to understand it. That gap lands on code review, which works less as a defect gate and more as a flow-control system for shared understanding.
31.08.2026
Ticket farming starts when activity replaces flow
Busy boards are not the same as delivery. A look at why ticket movement gets mistaken for progress, with a short diagnostic for delivery leaders who want to manage flow instead of farming tickets.
31.08.2026
The AI Bottleneck Moved to Code Review
AI coding tools solved typing speed and exposed a quieter constraint. When code generation gets faster but review and verification stay flat, the bottleneck moves into pull requests. Here is how delivery leaders can measure and manage it.
31.08.2026
Your sprint did not slip because of bad estimates
Sprint rollover and forecast misses are usually blamed on bad estimates, but they are more often symptoms of dependency-coordination mismatch. This post traces the mechanism from Cataldo, Herbsleb and Carley's congruence research to queueing economics, and offers a concrete Monday-morning diagnostic.
31.08.2026
Your staging environment is a queue with bad lighting
Under AI-assisted code velocity, the step between merge and production behaves like a shared queue that bends sharply near saturation. This article shows what the dashboard hides and three moves a delivery manager can make this quarter, grounded in peer-reviewed empirical work.
31.08.2026
Your conversations are a delivery system
In a delivery system, decision-bearing conversations are infrastructure, not soft skills. Manage them as part of flow, with a simple audit you can run on Monday and two cheap design rules.
31.08.2026
Your delivery metric is lying to you
Every delivery metric is a proxy, and any proxy used as a target eventually bends. Here is how velocity, burndown, DORA and coverage get gamed, what the research actually supports, and one question to ask in your next planning session.
31.08.2026
Your agile transformation is not real until the work changes
An agile transformation can look complete while the work barely changed. This article separates adoption from normalisation and offers a practical, evidence-aware audit a delivery manager or agile coach can run before accepting a green transformation report.
31.08.2026
Your portfolio is full. Your teams are not the bottleneck.
Teams get blamed for weak focus while the real queue often sits above them, in the portfolio. A look at portfolio WIP through product-development research, IT portfolio management, dependency reviews, and Ecocycle Planning, plus what to measure before the next prioritization workshop.