BONUS: Your Developers Got 20x Faster — Now Watch Your Product Managers' Heads Explode
Clarke Ching is "The Bottleneck Guy" — and he just spotted the bottleneck that AI is about to create in every software organization. It's not in the code. It's inside the heads of the people who decide what gets built. In this conversation, Vasco and Clarke unpack why speeding up developers with AI tools pushes the real constraint upstream — onto product managers, designers, and leaders — and what to do before cognitive overload crushes the people your organization depends on most.
Every Business Has a Bottleneck — Most Are in the Wrong Place
"Every single client I have is a detective puzzle. We're looking for this quiet killer sitting inside their business, siphoning off money. And if you look at them without the idea of going 'where's the bottleneck?' — you mistake the busyness for productivity."
Â
Clarke approaches Theory of Constraints like a detective story, not a physics lecture. Every business has a bottleneck — the narrowest point that chokes throughput. The question isn't whether you have one, it's whether it's in the right place. In software development, Clarke argues, the bottleneck should almost always be the developers. Not because they're slow, but because they're the pacing resource — like the aircraft carrier in a naval fleet that sets the speed for everything else. When developers are the bottleneck, the people upstream (product managers, designers, architects) have time to curate high-quality, high-value inputs. The people downstream (testers, ops) can deliver fast feedback. Everything flows. But when the bottleneck drifts somewhere else — and nobody notices — everyone gets busy, nothing flows, and the organization mistakes that busyness for productivity. Clarke's latest book, The Speed Book, lays out how to find where your bottleneck actually is and move it to where it belongs.
AI Just Moved the Bottleneck — And Nobody's Talking About It
"Just imagine one person trying to feed 100 developers. It's ridiculous. Everyone goes, 'oh, that's just crazy.' But that's kind of going to be what it's like."
Â
Here's the problem: AI coding tools — Claude Code, Cursor, Copilot — are making developers dramatically faster. If a team of 5 developers becomes 20x more productive, that's the equivalent of 100 developers. But you still have one product manager feeding them. The bottleneck hasn't disappeared — it's moved upstream. And when a bottleneck moves to the people who make product decisions, three things happen: they cut corners on requirements (shipping half-baked ideas because the team can turn them around fast), they feed developers busy work just to keep them occupied, and — worst of all — they lose the time needed to push through complexity to find elegance. Clarke references Steve Jobs's insight: Apple kept working past "peak complexity" until they reached "peak simplicity." That's where great products come from. But a product manager juggling work for 100 developers has no time for that journey. Elegance goes out the window.
Why Giving AI to Product People Almost Makes Things Worse
"If you want to wear your dog out so she sleeps, don't take her for long walks. Make the dog think. Brain games exhaust the dog faster than running."
Â
The obvious fix — give product people AI tools too — sounds right but misses the point. AI can handle the easy parts of product work: drafting user stories, generating specs, compiling research. That's the equivalent of taking the dog for a run. But the hard parts — the deep thinking about what to build, why it matters, how features interact — that's brain work. And brain work is exhausting in a way that volume work is not. Clarke works with senior leaders whose biggest challenge is pacing themselves. Heavy cognitive lifting burns through energy fast — your brain consumes 30-40% of your body's glucose when you're thinking hard. When AI handles the easy work, the proportion of your day spent on exhausting brain work jumps from maybe 15-20% to 50% or more. It's like lifting weights for six hours straight. You don't get stronger — you break down. On top of that, product people go from coordinating one stream of work to juggling many simultaneous initiatives. Clarke calls these "idea grenades" — and when you're juggling chainsaws with grenades attached, you start dropping things.
The Real Danger: Going in the Wrong Direction, 100x Faster
"If you change the relative capacities and make some of them much, much faster, the bottleneck's gonna move. My next book, jokingly, is gonna be called 'Who Moved My Bottleneck?'"
Â
There's an amplification effect that makes this worse than a simple throughput problem. An error in a line of code affects one line. An error in a design document ripples into hundreds of lines. An error at the strategic level — building the wrong features entirely — can be a disaster for the company. Now add AI speed to that equation. Overwhelmed product people making rushed decisions don't just slow things down — they point the entire organization in the wrong direction, and AI-powered developers execute that wrong direction at 20x speed. As Clarke puts it: you crash into the mountain, faster. The fundamental Theory of Constraints insight applies: if you speed up a non-bottleneck resource, you don't speed up the system. You just create more work-in-progress, more chaos, and more cognitive load for whoever the real bottleneck is.
Four Experiments to Try Before Cognitive Crush Hits Your Team
"Quality will come from actually slowing down. Money, profits will come from slowing down, building very good products, focusing on why we're building these products, not just how do we keep the AIs working."
Â
Clarke offers four practical experiments for teams navigating this shift:
Â
Get product people working with AI — as a thought partner, not a turbo boost. Teach them to delegate the routine work to AI so they can protect their cognitive energy for the decisions that actually matter. Think of AI as a delegation tool, not a productivity multiplier.
Help product people find their sustainable pace. Like Clarke's gym trainer who said "don't come five days a week or you'll never come back" — the people doing heavy cognitive lifting need to pace themselves. Old-school agile called this sustainable pace. It's never been more relevant.
Don't try to keep developers (or AI) busy all the time. The instinct to maximize utilization is the instinct that creates the problem. With AI, you're renting capacity by the minute, not paying salaries. Use it at the pace of good product thinking, not at maximum throughput. Turn the tap on and off as needed.
Measure what matters: value delivered, not stories completed. If 60-70% of features rarely get used today, imagine what happens when you 20x the feature output without improving the decision quality upstream. More features, more waste — at scale.
About Clarke Ching
Clarke Ching is "The Bottleneck Guy" — a Theory of Constraints and lean expert who wrote Rolling Rocks Downhill, the agile+lean business novel that never mentions agile, and The Bottleneck Rules. Born in New Zealand, he spent 20 years abroad (15 of them in Scotland) before returning home. He's spent decades helping teams find and manage the one constraint that controls everything else. LinkedIn
Â
You can link with Clarke Ching on LinkedIn.
New Zealand