Insights
Thinking on engineering transformation, flow measurement, and the organisational design questions nobody wants to touch.
Why Your Team Stopped Reporting Problems
Fewer escalations. Cleaner dashboards. Lower aging on the board. Before you interpret that as progress, ask whether your metrics system taught people that honesty is risky.
Your AI Strategy Is Incoherent — Here’s How to Tell
Your CEO declares AI-first. Your CISO blocks every tool request. Legal takes 90 days to approve a $500/month licence. You haven’t picked a lane — and the inconsistency is costing you credibility, talent, and time.
Copilot Made Your Velocity Number Meaningless. Now What?
Velocity was never a great metric, but it was directionally stable — until AI tools compressed the coding phase and broke the assumption underneath it. Here's what to measure instead.
The Stability Test Your Forecasts Are Failing
The forecast methodology was correct. The confidence intervals were calculated properly. The data was unstable, so all of it was wrong.
The Intervention Order Nobody Gets Right
Cycle time rises and leadership responds with hiring, tooling, and framework changes. The correct first move is a policy change that costs nothing. Applying the expensive fix before the cheap one is how transformation programmes waste budget without changing outcomes.
You Don’t Have an Agile Problem. You Have a Power Problem.
Transformation programmes fail because they redesign process and leave the power structure untouched. Capital allocation, decision rights, incentive structures — the variables nobody touches are the only ones that matter.
Your Team Isn't Slow. Your System Is.
Every developer is allocated. The utilisation report shows 95%. Delivery is still late. The problem is not effort — it's what high utilisation does to every queue in the system.
Five Stages Your Delivery Process Passes Through Before It Breaks
Delivery systems don't break suddenly. They pass through five identifiable stages. Most teams are in stage four before they act — and stage four is the dangerous one, because it still looks fine.
The Metrics That Protect Careers and Break Delivery Systems
Velocity, RAG status, and burndown charts have been wrong for years. Everyone knows it. They keep running them. This is not irrationality — it is rational behaviour in a system that punishes accurate reporting.
Why Frameworks Fail
SAFe, LeSS, Spotify — all installed, none changed measurable outcomes. The missing variable isn’t the framework. It’s the structural reform that nobody wants to make.
A Date Is Not a Forecast
A date without a confidence level is not a forecast. It is a hope with a deadline attached. Here is the difference, and why it matters for both sides of the conversation.
When Teams Game Metrics: What Anomaly Detection Reveals About Culture
Identity swapping. Timestamp tampering. Bulk status changes at sprint boundaries. These aren’t data quality problems. They’re diagnostic signals about incentives, psychological safety, and power.
When Fixing One Bottleneck Creates Another
The bottleneck moved after your last intervention. Most teams read that as failure. It isn't — it's the clearest signal that the intervention worked.
The Delegation Problem: Why AI Made Intent the New Bottleneck
AI made production cheap. But it made intent, validation, and accountability the new constraints. That’s an organisational design challenge, not a tooling one.
AI Generates Faster Than You Can Review. That's Your Real Constraint.
AI generates code faster than humans can review it. Review is now the investment-worthy constraint on most AI-assisted teams — and it scales differently than generation.
When Your Historical Data Stops Describing Your Current System
The forecast looked credible. It was built from six months of real data. The problem was that the system it described had stopped existing three months ago.
The Dashboard Delusion
Your delivery metrics report green. Your teams say they’re on track. Meanwhile, cycle times are rising, WIP is 3x capacity, and nobody can explain why. Dashboards don’t lie — they just show you what you asked for.
Aggregation Lies: What Your Portfolio Dashboard Isn't Telling You
Portfolio average cycle time: 12 days. Half your teams are at 5 days, half at 22 days. The average is technically correct and completely useless.
Specification-Driven Delivery: Governing What AI Builds
Five governance layers for AI-assisted delivery: context, enforcement, specification, proof, and review. How to maintain quality and accountability when code generation is no longer the bottleneck.
AI Coding Tools Moved the Bottleneck. Most Teams Haven't Noticed.
Faster code generation is real. But the bottleneck didn't disappear — it migrated downstream to review, testing, and approval. Most teams are optimising the wrong stage.
Resistance Is Relational, Not Just Personal
Resistance in Agile change efforts isn’t a personality flaw or a knowledge gap. It’s a relational signal — produced in the space between people, not inside one of them.
The Seven Failure Modes of a Stuck Delivery System
'Slow delivery' isn't a diagnosis. These seven patterns are — each with a distinct data signature and a distinct first move.
Process Behaviour Charts for People Who Hate Statistics
When your cycle time jumps from 14 to 18 days, is that a problem or just noise? Process behaviour charts answer this question with four simple rules. No statistics degree required.
The One Equation That Explains Most Delivery Slowdowns
When delivery slows, the instinct is to start more work. Little's Law says this is precisely the wrong move. WIP is the only variable you directly control — and reducing it is usually the fastest path to faster delivery.
Percentiles, Not Averages: Why Your Forecasts Are Wrong
Your average cycle time is 14 days. This number is meaningless. Percentile-based forecasting replaces the fiction of the average with a conversation about probability that leaders can actually use.
Your Improvement Worked. Or Did It?
Most teams declare an intervention successful because delivery feels better. Feeling better is not evidence. Here's the three-part test that is.
The Promise You Can Actually Keep
Most delivery commitments are negotiated guesses dressed up as plans. A Service Level Expectation derived from your own cycle time data is a different kind of promise — one grounded in evidence, trackable, and designed to rebuild trust rather than erode it.
Your Delivery System Is Running On Debt
The gap between your proxy-based cycle time and your actual cycle time grows wider every quarter. Flow debt is governance theatre, quantified — and most organisations don’t know they’re accumulating it.
Brownfield First: Why You Can’t Just Let AI Rewrite Your Codebase
Before AI rewrites anything, you need to recover current behaviour. Characterisation tests, boundary mapping, incremental delegation — the scaffolding most organisations skip and discover the cost later.
Portfolio Flow: Why Your Fastest Teams Are Hiding a Shared Problem
Your highest-performing team has the worst cycle time in the portfolio. This is not a coincidence — it is a structural guarantee. Here's why.
AI Is an Amplifier — Your Engineering Culture Determines Your ROI
The orgs getting 10x returns from AI already had strong engineering practices before AI arrived. If your foundations are broken, AI amplifies the dysfunction. A diagnostic framework for what to fix first.
Want to know when new articles drop?