Decision throughput: what it means for senior product and engineering leaders in the AI transition

AI made shipping cheap. Bad decisions got expensive.

That gap is the one most senior leaders feel right now and can't quite name. Their teams ship more code, more features, more experiments per week than they did 18 months ago. The Director of Product who used to push two big releases a quarter is now pushing six. The VP of Engineering who used to approve one architecture change every six weeks is now approving four. None of that velocity helps if the calls underneath it are the wrong calls. And the calls are getting harder. They're harder because the cost of a wrong call compounds faster when execution is fast. They're harder because AI tools can build the wrong thing very, very quickly. And they're harder because senior leaders are still making them with the same brain, the same hours, and the same calendar they had before any of this started.

I've coached over 100 senior leaders through this gap. I sat on the senior leadership team at NHS England Bookings & Referrals and the Department for Education's Teacher Services. I built AI products at HomeX through a $90M Series A. The pattern is the same in every team I've worked with. Teams ship more. Leaders decide less. The thing that needs to scale isn't engineering capacity. It's the rate at which a senior leader can make high-quality calls under pressure. I call that decision throughput.

What I mean by decision throughput

Decision throughput is the rate at which a leader or a team can make high-quality decisions under pressure.

Three things matter in that sentence. Rate, quality, pressure. Take any one out and the term collapses into something else.

Rate alone is just decision velocity. Make a thousand calls a week and most of them will be wrong. Quality alone is just thoroughness. Spend two months on a single call and you'll be thorough and irrelevant by the time you ship. Pressure alone is just stress. Lots of senior leaders are stressed. Most of them aren't actually deciding faster.

The combination is what counts. High decision throughput means you make more good calls per week than the average leader in your seat would. Low decision throughput means you either ship lots of bad calls or you don't ship calls at all.

The term has a cousin in the work of Martin Eriksson, the product leader behind The Decision Stack. His framework asks teams to align on mission, strategy, objectives, opportunities, and solutions in that order. The Decision Stack is about decision shape and order. Decision throughput is about decision rate at quality. They sit next to each other. One is the architecture of how a team decides. The other is the metric you watch when the architecture is working.

I've used the term decision throughput in coaching since 2024. I haven't seen anyone else write it down with this definition. So this article is partly a definition and partly a flag in the ground. If you're a senior leader and you've felt the gap I described, this is the term for it.

The reason it matters more in 2026 than it did in 2022 is simple. AI tools cut the cost of execution. They didn't cut the cost of judgment. A team that could ship one experiment a week now ships four. The leader sitting on top of that team has to decide which four matter, which one is the right bet, which architecture choice locks in compounding value, and which one is a sunk cost two quarters from now. The decisions used to be slower because the work was slower. The work is no longer slower. The decisions still are.

The 3 places it drops

Decision throughput tends to fail in three places. They're different problems and they need different fixes.

Personal. The calls a senior leader carries alone. Most senior leaders have three to five calls live in their head at any moment that they haven't told anyone about. The £4M vendor switch that's still on the fence. The performance call on a Director who's brilliant but doesn't scale. The architecture choice that's going to define the next two years and that no one else has the context to weigh. These are the calls that wake leaders at 3am. These are also the calls where outside structure makes the biggest difference, because the leader has been chewing the same loop for weeks and can't see past it.

A Director I worked with last quarter was carrying a £40M AI integration call at a regulated UK firm. She had the data. She had the team's input. She'd run three internal reviews. She'd been carrying it for seven weeks. We structured it in 90 minutes. Trade-offs, second-order effects, reversibility. She made the call that afternoon. The decision wasn't the bottleneck. The decision space was. She needed someone to hold the trade-offs flat so she could weigh them.

Team. Alignment that doesn't hold. The senior leadership team agrees in the room on Monday and diverges by Wednesday without anyone saying so. Two weeks later you discover Engineering and Product have been building two different things, both consistent with what was said, neither one what was meant. This is the failure mode CEOs notice first because the cost shows up in shipped work, not in calendars. It's also the one most coaching engagements completely miss, because most coaching is one-to-one and team dynamics need protocol-based work, not reflective conversation.

System. AI accelerating execution faster than judgment. This is the new one. Most senior leaders have not adjusted their decision rhythm to match what their teams can now ship. A weekly review of priorities was fine when the team shipped one thing per week. With four shipping, by Friday three of them are headed in a direction the leader hasn't actually approved. The system is running open-loop. Junior people are making calls that should be senior calls because senior leaders haven't built the throughput to absorb the volume. This is where decision quality drops first, because mid-level people make safer calls, not bolder ones. Innovation slows even though shipping speeds up.

The fix is different in each place. Personal needs structure. Team needs protocol. System needs cadence redesign. A blanket “leadership coaching” engagement won't move any of them, because none of them are mood problems. They're operating problems.

The 4 archetypes that carry this differently

In coaching senior leaders through decision throughput problems, four patterns recur. I think of them as archetypes. Most leaders are blends of two. Knowing which two helps a lot.

The Mover. Decides fast, ships fast, sometimes regrets it. Mover leaders have high throughput on rate but volatile quality. The fix is rarely “slow down.” It's usually a structuring frame applied right before commit, so the speed survives but the quality climbs. Movers thrive with tight, decision-by-decision pressure-testing.

The Builder. Reasons in systems, decides through models. Builder leaders make excellent calls in the categories they've already modelled and freeze on novel calls because the model isn't built yet. Their fix is reusable decision frames that compress the model-building step. They tend to dominate technical leadership roles in scaled teams.

The Reader. Listens to the team, weighs context, defers to people who know more. Reader leaders make great calls in steady states and underperform in crises because they're waiting for context that won't arrive in time. Their fix is a pre-agreed default rule for under-information conditions, so they ship at quality even without complete data.

The Anchor. Holds the line on quality, slows the team to do it, takes the hit on velocity. Anchor leaders are often the strongest in regulated or high-stakes environments. Their fix is rarely about quality. It's about identifying which decisions actually need anchoring versus which ones are throughput drag dressed up as care.

The diagnostic I run starts here. Four archetypes, two dominant, one or two specific decisions to work through live. From that I can tell which of the three drop-points is dominant, and the engagement designs itself. Take the 5-min version here.

Why most leadership coaching misses this

Most senior leadership coaching is reflective. The model is borrowed from therapy. A trained coach asks questions. The leader thinks aloud. Insights surface. Patterns repeat. Over time the leader integrates the work and acts differently.

Reflective coaching is real and useful. It is also not the right tool for decision throughput. Decision throughput is a live operating problem. The leader is sitting at their desk on a Tuesday afternoon with a £40M call in front of them, three open browser tabs, two Slack threads they're avoiding, and 90 minutes before the next meeting. What they need is not a question that helps them sit with the discomfort. What they need is structure that lets them weigh the trade-offs in the time they have, see the second-order effects clearly, and commit.

That's live decision work. It's a different practice from reflective coaching. It looks more like what a Chief of Staff does when they're at their best. Or what a senior advisor does in a board context. Or what a strong product partner does on a call you're carrying alone.

Most senior buyers know the difference once they feel it. They've usually had a coach before. The coach was useful for personal patterns and unhelpful when an actual call landed on the desk on a Wednesday morning. That's why they're looking again. Not because reflective coaching failed them. Because they need a second tool, and the market has been pretending one tool is two.

This matters commercially. Senior buyers in product and engineering roles aren't shopping for emotional regulation. They're shopping for partnership on the calls they can't take to anyone inside the company. The vocabulary the market uses (“executive coaching”) is the vocabulary of the wrong tool.

What changes when judgment catches up to speed

When decision throughput climbs, three things change in a senior leader's week.

First, decisions move faster with less re-litigation. The same call that used to surface in three different forms over three weeks gets made once and stays made. Not because the leader becomes more decisive in a personality sense. Because the structure they apply to each call is consistent enough that the team trusts the output and stops bringing it back. Re-litigation is the silent tax on senior calendars. Cutting it back creates more time than any calendar audit ever did.

Second, trade-offs get clearer. A Director who used to feel torn between three options now articulates which option dominates on which dimension and what they're choosing to give up. That clarity travels. Their team makes better aligned calls because they understand the dimensions the senior leader is actually weighing. Communication overhead drops without anyone running a workshop on communication.

Third, alignment holds outside the room. Decisions made with structure are easier to communicate, because the structure carries forward. The Engineering Director and the Product Director walk out of the room with the same understanding of what was decided and why, because the same trade-off analysis is sitting in front of both of them. The Wednesday divergence stops happening, not because everyone's better at meetings, but because the artefact from the meeting is sturdier.

This is where six months of work starts to compound. Most leaders who go through it don't notice the change in week one. They notice it when their team's velocity holds steady through a hard quarter and the team's quality holds steady alongside it. Will Larson's writing on engineering strategy points at the same idea from the engineering side. The leaders who maintain alignment through scale are the ones who can decide cleanly under pressure. The ones who can't end up with sprawling architectures and political teams. Decision throughput is the underlying capacity that lets the strategy hold.

What to do this week

Three actions. None of them require buying anything.

Take the diagnostic at /diagnostic. Five questions, three minutes. Outputs your dominant archetype and the most likely drop-point in your decision throughput right now. Same one I use in the first 15 minutes of any new engagement.

Bring one decision you've been carrying to your next 1:1 with whoever you talk shop with. Not to ask them to make the call. To structure it out loud. Name the three options, what each one trades off, what makes it reversible or not. The act of structuring surfaces the answer in 80 percent of cases.

Name what you're really waiting for. Most decisions live more than three weeks aren't waiting for information. They're waiting for permission, or for someone else to move first, or for the leader to feel ready. Naming the actual blocker is usually enough to release it.

The call you can't make alone

If you've recognised yourself in this article, the next step is probably not buying coaching. It's getting clear on which of the three drop-points is yours.

If structure is what you need, I run a 30-minute discovery call. You bring a real call you've been carrying. We work it live. By the end you'll either have made it or you'll know what's missing. No pitch. Book at /work-with-me.

For the AI strategy side of the same problem, my colleagues run a deeper diagnostic at aiimpactsystem.com. The two pair well. Most of my clients run them in parallel.

Bring a decision. Work it live.