Solo Agentic Devs Are Crushing It. So Why Is the Org Falling Apart?

Agentic devs don't want to work on projects with other people, not because they're antisocial, but because everybody they work with slows them down. And honestly, I get it. When you're moving at agentic speeds, coordinating five people, moving in the same direction on the same thing at the same time at that speed is extremely challenging. So is the team obsolete or are people just looking at the new world with old world glasses on the speed of agentic development has removed a constraint on generating output. But output volume and outcome success can often be totally unrelated. So now we have to look at the whole system and find where flow gets stuck. In this video, we're going to look at why teams can feel like overhead and where the bottleneck moved to. We're also going to explore why your org chart is built for problems that don't exist anymore and what teams and careers look like on the other side of agentic transformation. First, let's realize there's no universal detailed truth for what a team is. In some orgs, a team is a pool of resources working on tickets assigned to them. The business tells the PM what to turn into tickets. The PM gives the tickets to the EM, the EM gives tickets to the dev and the dev gives the tickets to QA and QA gives the tickets to operations. And each disconnected from the one before them and the one after. In other orgs, a team owns a flow of value from end to end. They hear what a customer needs. They decide what to build.

They ship it. They support it in production. They learn how it lands all inside one team. Both of those companies call those things teams, even though they're quite different, but on the macro scale, the idea seems to rhyme a little bit. You hand a unit of context to a person or a group with a shared mission and you trust them to deliver value inside of that mission. What's different is how they set up partitions around things like responsibility and accountability and cognitive load. Now think about what's changed for an agentically unlocked engineer. Five years ago, you needed a team of developers to ship a system that one agentic engineer can now ship in a few focus days. The places where you used to get leverage from coordinating with other humans doesn't pay off the same way anymore. For an agentic developer in the flow, every alignment conversation they would have had in a 2020 style setup looks like a global slowdown of the entire project. In 2020, it was necessary to pay the coordination tax to get more done with the help of others. But now it doesn't look like that from inside the flow and it can look that way from outside the flow too. When managers see what an agentically unlocked developer can push out the door. Well, we can see they're not necessarily wrong if we just think about speed of shipping code, but if that's all they think about, they're failing to spot and adapt to the change in front of us. And that change isn't that teams stop being useful. The issue is they're thinking about teamwork in an old paradigm of how teams once delivered value.

And they're operating with only one foot in this new agentic paradigm where the leverage equation is completely different. Now with agentic practices, those boundaries have dramatically expanded. Businesses have either used teams to push units of output out a door or to take a shared challenge and make them team solve it. And the purpose for a team still exists. You just can't see it. If you don't know what boundaries to look for, removing a non-bottleneck from a process does not improve system throughput. The bottleneck stays right where it was and still limits the entire system. And when you do remove a bottleneck, the real bottleneck, it just moves the bottleneck somewhere else in your system. It can only ever change places. So if teams are no longer constrained by output, where are outcomes constrained? Two places, coherence and sense-making. Coherence is what we need when many humans are individually staring at agentic speed and we want to keep them going in a common direction. Imagine an army with no coherence, just soldiers dropped into a battle with no plan, no ability to communicate. That's what an organization full of agentic engineers can look like right now. One person can cover a lot of ground, but you just can't ship anything that makes sense at the macro scale, just janky disconnected bits. Our second skill here is sense-making.

When the signals are messy and the answer isn't obvious, we need to make sense of the situation. There's no single framework for it either. That's what makes it challenging for humans and very hard to make an AI good at. Most of us are used to a categorization model, a two by two matrix where you fit data into a preexisting framework. In a sense-making model, the framework emerges from the data. Coherence is doing that work together. So with coherence and sense-making, a team of agentic engineers won't quickly go after five disjointed goals. And sense-making belongs everywhere, not just product strategy or high-level business goals, it's looking at logs and figuring out why your system is doing something weird under load. It's walking through a strange looking bug to see it's telling you about your architecture, it's reading what your customers are doing and asking why. Wherever the signals are messy, somebody has to make sense of them. AI can process the data. It can even automate some decisions, sometimes quite well, but there's a difference from sense-making or making good judgments about the situation we're working on. This brings me to the next issue that AI will not fix coherence issues. AI can facilitate coordination, it can summarize, it can sync, it can hand context around, but it doesn't do coordination. Coordination is humans agreeing on direction and then acting on that agreement, and that's not working the way it was before. Humans have stopped talking to each other in many of the ways that they used to.

Big tech is seemingly throwing gas on this fire too. They're cutting middle managers explicitly to speed up decision-making. Amazon flattened 30,000 corporate jobs last year. Microsoft and Meta announced their own waves in April of this year. The people holding coordination together are the ones being cut with no clear best practices to replace what they did. One of the major reasons we formed Teams was to keep alignment on a mission focused on a complex, ever evolving challenge. Sense-making together, coherence through shared context, judgment that draws on a group, not just one individual, that work didn't disappear. The need for it just got way bigger and the need is at a higher level than we were talking about in 2020. If you look at your company's cash flow statement and rotate it 90 degrees, it can often look very similar to your org chart. That's not an accident. Alfred Sloan invented this multi-divisional structure in the 1920s, functions as columns, divisions as rows, and it worked well enough that every large company for the next century adopted it. The orgs we have now are built around a world we do not have anymore. This isn't a new problem either. It's a century-long theme in management. The dominant business culture rewards local optimization and every generation, somebody tries to point out that you can't get a great hole by optimizing each part separately. Mary Follett told us that in 1920, Drucker told it to us again in 40s.

Deming spent the 50s saying it, Ackoff in the 70s, Goldratt in the 80s, Senge in the 90s, Team Topologies in 2019, even the DevOps has this trend and hundreds, if not thousands of other voices in between, all in different fields, all coming to the same point. Each generation, everyone nods, and then they get pulled back into this local optimization because frame questioning of the status quo threatens power who establish it. Career incentives reward silo loyalty. And we measure functional output because it's legible, but outcomes are harder to count. Business schools train analysts, not systems thinkers, and the consulting industry is often reinventing old wisdom instead of building on it. Together, all of these forces prevent new organizational ideas from becoming dominant. Every generation walks into the same trap. Each one notices it, names it, is celebrated for their insights and gets sidelined. And I see the Sloan shaped org as an evolutionary default destination, not a great idea. Agentic development changes the evolutionary equation. On a 10 year horizon, Sloan style, top down, command and controls, stay in your lane orgs probably won't survive. They were never optimal, but they were stable enough in a slow environment that it would remain the global default. Agentic engineering is interesting because it isn't a strategic choice. It's a change in the landscape strategies are forced to position themselves on. A new generation of businesses are running at agentic speeds and they're going to overwhelm these legacy corps, outmaneuvering them at every turn. They will constantly adapt in markets that legacy corps won't be able to comprehend in real time.

Agentic engineering behaves as a forcing function, not a strategy choice. It's evolutionary pressure on a landscape that just shifted. Sloan's old M form is simply incompatible with our future landscape. But the org chart isn't the only piece of this. The way we think about the work inside of it is also broken. In Dave Snowden's Cynefin terms, laying a brick wall is in the clear domain. It has codifiable rules, you know, poor mortar in the winter. You know, the thickness and height ratios cause and effect are obvious. Best practices apply. Building a disruptive fintech business on the macro level is nothing like that. Sure, inside that business, some bits are clear, like customer login flows. And many bits are complicated, but knowable, like fraud risk analysis, where experts can reason their way to good answers with good practices. But the higher level business is complex and the detail specific nuance that makes features market differentiating are often very complex. We can't perfectly predict what to build, how the market will respond, what bits compound, you only learn what works through exploring and taking actions. Seeing what happens when you make a change in complex work, the Cynefin prescription is probe, sense, respond, not analyze, decide, and execute. You don't plan your way through it.

You run a small safe to fail experiment and amplify what works, which is directly relevant for agentic engineers. Right now they're optimizing for execution speed in a domain where the bottleneck is their speed at navigating outcomes in complex domains. Agentic engineers have gotten really good at building brick walls really fast. The problem is they can't figure out how to get two brick layers to build one wall together, much less build something like a fintech. And even when they do coordinate, they're still building walls. Nobody's asking whether a wall is what this situation needs. Modern software organizations push back on the M form with end to end ownership teams that own a value stream from start to finish. Many orgs constrain their teams to live inside the product engineering silo only. The move now is to push streams up to enterprise scale across strategy, marketing, sales, engineering, and any function that is generating value along the value stream. A stream aligned team is partially a complex domain response. Keep the feedback loop tight enough that the team can sense and respond to what the system is telling them. But many orgs who have teams like this cut the team off at the edge of their product engineering silo. The people building the software are still disconnected from the people selling it, marketing it, setting strategy. Value still flows across two verticals of the cashflow statement. With cognitive load and toil dropping the way they have, the move now is enterprise scale value streams. The team is the unit that can hold a coherent mental model of the stream.

Orgs now need to furnish those teams with the skills they need to operate at that scale. So what does a team become now? A team in its purest form is a container for coherence under the cognitive load that comes from solving a mission. Like optimizing an entire value stream, reducing cognitive load and increasing output capacity on legacy team boundaries doesn't make that foundational need to set up reasonable boundaries go away. In the old version, the container was sized around how much one human could implement and how much coordination was needed to combine multiple humans into something bigger. But the work is different now. Humans steer systems they govern. Their outputs come from those systems and a team of humans gives those systems coherence and shared judgment. The only way to improve from here is to widen the breadth of ownership and raise the level of the team mandate to match their initiatives. It's a different kind of team than people are used to. People don't know what the new shape looks like. So they're operating in the old shape with new tools. Pick a ticket, fix a small bug, talk about it at stand up tomorrow and drown in pull request reviews. That mode is becoming background noise. None of that is where the alignment work should be happening anymore. It's not what the team should be talking about.

The thing that the team owns now is the cybernetic loop that improves a value stream. The thing that the team does together is keep that loop coherent. Does the work fit? Does it land? Does the system that produce it improve? And don't get tunnel vision on the product because things are moving so fast now, how you build it matters as much as the product itself, if not more. When your competitors are all moving at agentic speeds, the products you ship are just snapshots in time. You need a machine that moves with the market and that's a different cognitive load and a different set of skills than implementation coordination. You can still depend on other teams for shared services and platform work and you still need deep specialists for some things, but the default mode is stream-aligned teams that own the whole thing. Now the part you've been waiting for, how this shift impacts individual careers. In the last video, we talked about how toil shaped work is shrinking and the judgment layer is growing. This is the org version of that. Your career stops being defined by one narrow function front end developer, back end developer, product manager. What matters more is whether you can operate across the full value stream in partnership with the others from understanding what a customer needs, deciding what to build, shipping it, supporting it in production. Not every piece on your own, but co-ownership of the flow with your team. Not everyone dividing up all tasks equally, but everyone working together with their judgment and steering together.

That's the cognitive load container reshape at the individual level. Your container gets a different shape. You hold a wider scope, fewer specialized depths, more judgment calls about how it all fits together. Some careers will stay narrow and deep. That's fine. There are domains where deep specialization still wins, but let's take an honest assessment of the tech industry right now. A massive number of developers grind leaked code for interviews and never write a binary tree. And they're asked to write simple crud layers in a code base where the hard part is navigating the code base and never fixing the mess. Most careers won't sit at a narrow deep specialization and most never really did anyway, even if the work was complex and mentally taxing. I don't think any of this is going to be optional either. This will be a great filter event for many businesses. Orgs locked into this old M form just won't be fast enough. They can't function in this new economy. If your competitor has the raw power of multiple humans with agentic workflows, fully coordinated shipping end to end across the whole business, and you're still doing handoffs between five departments for every change, you're going to be left in the dust. Your competitors will be constantly sensing the market and what it wants, deciding what to build next, shipping it, learning from it in days. If you're scheduling a cross-functional alignment meeting to see if you can take it on in Q3, you're done.

The org structure is built on Sloan's M form will be eaten alive. They're inefficient, sure. But the bigger problem is that Sloan shaped orgs literally can't reshape cognitive load containers fast enough to keep up with the judgment calls. This new mode demands. To flow fast, you need a single threaded ownership of the full value lifecycle and not just inside of engineering at the level of business delivery, org charts built around mapping job functions to reporting lines onto a cashflow statement won't survive this. I'm not sure if I should even say this next part, but honestly, on a 10 year horizon, the layoffs we're seeing right now might be a good thing for those individuals. The orgs might not survive anyway, as they'll be too slow to change. So the people leaving early are leaving an obsolete economy with time to move on and adapt instead of when it falls apart on them. So even if it hurts the individuals a lot on a one year time horizon, it might be for the best on the 10 year horizon. So if this resonates, the career video goes into more depth about what that means for you personally, your skills and your job. If you want to go deeper into this and help your team adapt agentic practices, I teach hands on courses for developers and teams who want to level up. I also work with leadership groups to figure out how to adapt their organizations towards this new paradigm links in the description and subscribe so you don't miss the next one.