Waste Didn't Go Away. It Got a Corner Office.

Let's talk about something nobody wants to say out loud in the middle of their AI transformation…you did not eliminate waste when you adopted your shiny new toolchain. You automated your way into spending more money on it.

Waste in knowledge work has been the same problem for decades. Wasted time is wasted time. It does not matter if that time is a developer staring at a Jira board in 2009 or a developer waiting on a 4000-line AI-generated pull request to get reviewed in 2026. Idle work is idle work. And idle work costs money. The clock does not pause because your pipeline is slow (or blocked).

What has changed is the unit economics. When humans were writing every line of code, your daily output capacity had a natural ceiling. Now, with AI-assisted development, that ceiling is gone. Developers are creating more code, faster, in higher volumes than ever before. Which sounds great until you realize that the same broken flow conditions that slowed everything down before are now slowing down a much bigger piles of work. The waste is multiplied. And the cost is multiplied with it.

Idle Work Is Still the Tell

If you want to find waste in a knowledge work system, you do not start by looking at what is moving. You look at what is sitting still.

Idle work is the single biggest indicator of a flow problem. Work that is in progress but not progressing is cost accumulating with nothing to show for it. It is the difference between a car sitting in a repair bay for a week because the part has not arrived and a car that got fixed in two hours. Both show up as "in progress." Only one of them is actually in progress.

Idle work shows up in three ways, and they have not changed one bit just because we are all supposedly living in the future now.

Stalled work is work that cannot move because something is blocked or something is waiting. Waiting on a review. Waiting on an approval. Waiting on another team to finish something upstream. Waiting on clarification from someone who is in three other meetings. The work exists, the work is "in progress," and the work is going absolutely nowhere.

Task and context switching is what happens when someone pauses active work to focus on something else. This one is sneaky because it often gets dressed up as responsiveness or flexibility. What it actually is, is a flow interruption. Every time someone stops working on one thing to address another, the thing they set down does not sit still in some pristine state waiting for them to return. It ages. It accumulates complexity. The mental context that made it easy to work on starts to dissolve. Context switching is not multitasking. It is serially halting your own progress.

New shiny priorities is the one that should make every senior leader a little uncomfortable. This is when something new gets pulled into the queue not because the current work is done, but because something newer got louder. The team stops working on what was agreed to be important and starts working on what is suddenly urgent. Everything that was in flight stalls. Priorities get shuffled. Delivery dates quietly move to the right. Rinse and repeat every two weeks.

None of these patterns are new. They are just more expensive now.

How to Spot Waste in the Age of AI

Here is the eternal problem with waste: it is invisible until you make it visible. And most teams are not making it visible. They are looking at velocity metrics, story points, and deployment frequency and feeling pretty good about how much is getting shipped. But shipping is not the same as delivering value, and motion is not the same as flow.

In an AI-assisted development environment, the pull request is your unit of work. It is where intent becomes reality, where code transitions from individual effort to organizational output. Which means the health of your PR flow is the health of your delivery system. Want to see where your waste lives? Start there.

Age of pull requests is the most honest metric most teams are not tracking. The age of a PR is how long it has been open. That number tells you how long work has been idle. A PR that sits open for four days is not in progress. It is stalled. It is waiting on something, or competing with something, or nobody has gotten to it yet because five other things got prioritized. An old PR is a symptom. If your average PR age is climbing, your flow is degrading, and cost is accumulating in the gap.

Volume of pull requests tells you something different but equally important. AI-assisted development dramatically increases the rate at which code gets written, which means the volume of PRs a team produces can spike significantly without anyone really noticing until the review queue looks like a small city. Volume is not a problem in itself. Volume that outpaces your team's capacity to review, integrate, and ship is a problem. When PRs pile up faster than they move through, you have a queue that is actively growing older. See above.

Alignment of PRs to priorities is where things get genuinely interesting and genuinely uncomfortable. If you can look at the PRs your team is actively working on and the priorities your organization has stated, and they match, you have alignment. If they do not match, you have drift. Work is getting done. It is just not the work that matters most right now. In a world where AI makes it easy to generate a lot of code fast, drift gets expensive fast. You can spend a sprint producing a high volume of technically impressive output that does not move the needle on what you said you were going to deliver.

Put those three signals together and you have something useful: a picture of where your flow is actually healthy and where idle work is silently inflating your cost of delivery. Age tells you about stalls. Volume tells you about capacity constraints. Alignment tells you whether the work you are doing is the work you should be doing.

Make It Painfully Obvious

The promise of AI-assisted development was not that it would eliminate the need for good flow management. It was that it would amplify what your teams could deliver when flow was working well. That is a meaningful distinction.

If you have always had a culture of context switching, you now have a culture of context switching with a higher throughput of interrupted work. If you have always had a habit of pulling in new priorities mid-sprint, you now have a habit of pulling in new priorities mid-sprint while carrying more in-flight inventory. If reviews have always been slow, they are now slow against a bigger backlog.

Waste did not get disrupted by AI. Waste got a bigger budget.

The teams that will actually benefit from this technology shift are the ones that treat flow as a discipline and not an afterthought. They are the ones watching PR age, managing PR volume against review capacity, and asking hard questions about whether what is in motion is actually aligned to what matters. They are the ones who already know that idle work is cost, and that the answer to that problem is not more throughput. It is less tolerance for the conditions that create stalls in the first place.

‍