AI analyst Nate B Jones argues that AI eliminates coordination overhead, not core human value
A solo analysis by Nate B Jones of AI News & Strategy Daily on how AI-driven agentic systems restructure organizations by eliminating coordination work rather than simply automating tasks.
Summary
Nate B Jones argues that the standard framing of AI's impact on jobs — mapping AI capabilities onto existing task lists within fixed org charts — is fundamentally wrong and understates the real transformation by a factor of two or three. His central claim is that most of what knowledge workers do is not value creation but coordination overhead: meetings, PRDs, status updates, handoff documents, and sprint planning that exist solely because the execution layer is made of humans. When agentic AI systems can move directly from insight to code, the coordination layer doesn't get automated — it becomes structurally unnecessary. Jones uses the recent Block layoffs, where Jack Dorsey attributed job cuts to AI, as an example of a leader using the AI-automation narrative to obscure what was actually a correction for overhiring. Jones concludes that removing coordination work concentrates human effort on the highest-value activities — product vision, brand thinking, genuine customer relationships, systems architecture — work that coordination overhead has been crowding out for decades.
Key Takeaways
FULL TRANSCRIPT
The Wrong Question Everyone Is Asking About AI and Jobs
Nate B Jones: One of the lessons that we are learning — one of the bitter pills that we need to swallow — is that AI is telling us our jobs were never the real job. And that's actually good news. I'm not just making that up. I'm going to get into why it's good news.
Right now, every serious conversation about AI starts with the question: what tasks can AI do? Consultants will decompose roles into task lists for CEOs. Researchers will survey the frontier. Somebody publishes some kind of percentage estimating it, and now it's a headline, and everyone maps the existing organization and says, "Oh my gosh, how am I affected?" And it's this cycle of panic.
The implicit assumption is that the organization is a fixed structure — which is wrong, by the way — and that it's a set of roles performing a set of tasks — also wrong, by the way — and that AI is a force that acts on that structure, automating some cells and leaving others — also wrong. These are all the wrong questions. And getting it wrong doesn't just produce a slightly off answer. It produces a catastrophically off answer, one that understates the impact of AI on organizations by a factor of two or three.
I am here to tell you that we would rather be in the future where AI is more impactful than the future where it's less, because we humans benefit more at work specifically. I'm going to get into it. I doubt you believe me, but I'm still going to get into it and explain why.
The Coordination Tax: What Knowledge Workers Actually Do All Day
Here's what's going on. The tasks that exist in a 200-person tech company are not a natural set of things that need to be done. People are going to chuckle at this point, and it's fair to chuckle. They're artifacts of human coordination. Most of what knowledge workers do all day is not value creation — whatever they tell their boss, and whatever the CEO thinks. It's just coordination overhead.
Writing specs so that someone who wasn't in the room can act on a decision. Sitting in meetings so that eight people who can't share a brain can synchronize their state. Preparing decks so that an executive who doesn't have time to read the primary source can make a decision. Filing tickets so that work can be tracked across people who can't see each other's progress directly. PRDs, sprint planning, status updates, cross-functional syncs. Are you tired yet? Onboarding documents, design handoffs. No wonder we want to Netflix and chill. Backlog grooming, postmortems, retros.
None of these are the value. The code is the value. If you are in any kind of software company, the code is the value. If you are in any kind of product company, the shipped product is the value. Everything else is the cost of producing that value with an execution layer made out of humans.
When AI agent harnesses can go directly from insight to code in one big loop — which they can now, I did a whole video on this — you don't just automate tasks within your existing organization. You delete the need for the org to be structured this way at all.
Take a deep breath. We're all going to be fine.
The coordination layer evaporates. The roles that existed to manage human-to-human handoffs disappear. Not because they're automated — because the handoffs they managed no longer happen. And that changes the math on everything.
The Calendar Test
I want you to try something. Pull up your calendar from last week. Not the idealized version — the real version. Count up the hours.
How many hours did you spend in meetings whose primary purpose was transferring information from someone who had it to someone who needed it? Status updates, sprint planning, design reviews, standups, one-on-ones where you spent 20 minutes giving your manager context they need for the meeting they have after yours.
Now count up the hours you spent creating documents — creating artifacts whose primary purpose was translating your knowledge into a form someone else could do something about. PRDs are this. Specs are this. Decks are this. Tickets are this. Handoff documents are this. That email you drafted for 45 minutes because three different stakeholders needed to understand the same decision from three different angles is also that.
Now count the hours doing the thing you were actually hired to do: writing the code, designing the interface, closing deals, building a strategy, creating something that did not exist before.
If your ratio looks anything like the research suggests, you spend roughly 60% of your time on those first two categories and 40% on the third, if you're lucky. So 60% of your time goes into meetings and creating documents to coordinate with people, and only 40% goes into doing the real work.
Microsoft's 2025 Work Trend Index found that the average employee spends 57% of their time communicating and 43% creating. Sana's Anatomy of Work study put it at 60% "work about work" — very similar numbers. The average knowledge worker now sits through 11.3 hours of meetings a week. I think that's low. I think that's really low. And that number has tripled since 2020.
Those are not productivity failures. Those are organizational mechanics that are required to coordinate in the world we find ourselves in. I call it the coordination tax.
A PM's job is writing PRDs. An EM's job is sprint planning. We say those things. We don't experience these activities as what they really are — as overhead. We experience them as the role. But the value is working software that solves a customer problem. That's it. That's the value.
If you work in software, everything between "we understand what to build" and "the thing exists and it works" — also true of physical product, by the way — is process. And process exists because the execution layer is made of people.
This is not an insult to the people performing coordination work. That work is real. I've done it. It's often difficult, and the people doing it are frequently very good at it. But that work exists because of a constraint: the constraint of human bandwidth, the constraint of human context limits, the constraint of human communication. It doesn't exist because it's a direct path to value.
If you could snap your fingers and have everyone share perfect context with zero latency and zero translation loss, you would delete most of our roles overnight. Not because the people are unnecessary, but because the problem they solve would not exist anymore.
Why Agents Don't Pay the Onboarding Tax
Think about what happens when a new engineer joins the team. For weeks, they produce almost nothing. Not because they're incompetent, but because they're absorbing context. Every decision the team made in the past year lives in someone's head and probably in some scattered documents. The new hire has to reconstruct this through conversations, through old PRDs, through commit history, through Slack threads where three people answer the same question all differently. Which one's right?
An agent harness starting a new task has a much simpler, cleaner problem set and goes faster. The agent reads the codebase, reads the progress file and the git history — full context in seconds. The onboarding tax, the coordination tax — neither of those exist, because the problem that onboarding solves, transferring institutional context from human brains to a new human brain, just doesn't exist for the agent.
In a video a few days ago, I talked about the fact that multi-agent coordination harnesses all converge on the same structural pattern: decompose the work, parallelize the execution, verify the outputs, iterate toward completion. And I showed that the key variable determining what's in range is whether or not you can iterate against something that passes a sniff check — whether the work can be turned into little problems where we can recognize meaningful progress.
This video is really about what happens when you apply that insight to an organizational structure.
The Org Is Moving to Code
The key thing to think about is this: if we take seriously the idea that agentic coding harnesses can generalize — that they can solve a lot of knowledge work — what we are really saying is that the org is moving to code.
When I say the org is moving to code, I do not mean everyone then becomes a software engineer. That is the 2015 version of this conversation and I'm not interested in it. What's changed is that code is now readable and writable in natural language. You describe what you want in English. The agent writes the implementation. The interface between human intent and machine execution is just a conversation. It's not even a programming language.
"The org is moving to code" means the artifacts of work are all becoming machine-inspectable and testable. But there's a deeper point here that's very easy to miss. When the translation layers disappear, everybody gets closer to the product. Not closer to a description of the product — closer to the product itself.
The PM isn't writing a document that an engineer will interpret. They're shaping the actual artifact. The designer is not producing a mockup that approximates the final thing. They're working on the final thing. The marketing leader isn't writing a brief for someone else to execute. They're actually just building the landing page, seeing what it would look like, testing the conversion — all in the same session.
That's not a productivity gain. That's the wrong word for it. It's just fundamentally different, and it's a much better relationship with work for us, for humans.
Here's what this looks like in practice. A person can sit at the command line and iterate directly on the design, the specification, and the code in the same session. The spec and the implementation converge into one artifact. There is no PRD — and by the way, I'm not making that up. Anthropic has talked about the fact that they don't use PRDs anymore. There is no sprint planning meeting. There is no status update. The state of the work is the code. It's inspectable all the time. There is no design review as a separate ceremony. You iterate on the design and the implementation at the same time.
This is not theoretical. For some AI-native organizations, this is just an ordinary Tuesday. And for a growing number of people, this is already how work happens. It's not the future — it's the present.
And here's what I want you to notice. It's not that the coding bit got automated. It's that the translation layers between people got deleted. So there was no PRD needed, and the person with customer insight could work directly with the agent to get that vision to come to life. There was no sprint planning, and that means there aren't eight engineers who need to coordinate, because they're all working with agents and they can get the context they need. There was no status meeting — the status is the commit history. No design-to-engineering handoff — you just iterate on the artifacts. No QA as a separate function — verification is built into the agent loop.
Every one of those is a coordination function at a big company. Every single one evaporates when the execution substrate shifts from a team of humans to a harness of agents directed by a human with good context.
This is not something you can do overnight. I know that for many organizations this is a massive leap, and I know that there are many interstitial states — many steps along the way to move in that direction. You may be at a point where you are using Claude Code to generate smarter PRDs, and those PRDs are informing your engineers and saving you time, and you celebrate that, and you should. But I want to be honest with you that there is a world here where all of those coordination artifacts go away. Because if we can't see this world — the world some of us already live in — we are going to get surprised when it comes to us. And I don't want you to be surprised.
Whether we like it or not, every step along the way to the world I'm describing is going to be temporary. We are going to move to this world. And why? Because it's so much more efficient at generating customer value. And frankly, because for a lot of us, all of that coordination work was miserable. Do you love the 11.6 hours you spend in meetings? Really — I don't think you do.
The Flywheel: Why the Standard Forecast Gets the Math Wrong
This is where the analysis goes to really interesting places. It's one of the most important structural arguments that I think I'm going to make this year.
When you remove humans from the execution layer because you have agents, you don't just eliminate the coordination roles that existed to manage those humans. You actually make the remaining work more verifiable than it was before. In other words, as you simplify the organization to work with agents, the remaining work becomes more easily verifiable by agents. You are beginning to get into a loop. We call this a flywheel. And you are beginning to turn a flywheel that is pushing the work of the organization relentlessly toward humans managing teams of agents that produce customer value. And that's the whole org.
A marketing campaign becomes a deployed landing page with conversion metrics — an agent can test whether the page converts. A sales proposal becomes a configured offering with win-loss data — an agent can optimize the configuration against historical outcomes. Once the artifact is code, you can test it. You can iterate on it, and you just need less coordination.
So we are getting into a compounding loop: less coordination needs fewer coordination roles, makes remaining work more verifiable, means agents do more, means the whole loop turns around again, and we need less coordination. Each turn accelerates the next.
The implication is that the standard forecast miscounts what's really going on here. When AI changes the structure of work, the entire distribution of work changes with it. Most tasks that exist today are coordination tasks, and coordination tasks exist because the execution layer is human. If you change the execution layer, you don't just automate tasks inside that org chart — you are going to delete the need for a lot of the org chart to look this way at all, to look like the current org at all.
What Survives: Genuine Judgment vs. Coordination That Feels Like Judgment
Now, there are many natural responses to this, and I'm going to answer all of them because I can hear you thinking: Nate, there is no way this turns into good news for people.
The natural response is to say, well, the routine stuff gets automated, but the judgment-heavy work stays human. And that is partly correct. But the category of work that we call high-judgment is smaller than we would like to think. We have been conflating two very different things: genuine judgment under uncertainty, and coordination overhead that feels like judgment because it's complicated.
Think about strategic vision. Most strategic decisions decompose into probabilistic bets with researchable evidence. Market size is a calculable metric. Competitors are researchable. Capabilities are auditable. Financial models are computable. It feels non-verifiable because the synthesis traditionally happens inside a single person's brain in this opaque process that we call judgment. An agent harness processing all those inputs simultaneously could probably do a similar synthesis, and the reasoning would be inspectable.
Once you internalize that pattern, it repeats everywhere you look. What appears to be a single judgment-heavy role is almost always a composite of verifiable prep work and a layer of genuine non-verifiable judgment over the top. But we've bundled this all together because having a single person do both was super efficient organizationally. Agent harnesses suggest that we unbundle them. The preparation may become automated and the judgment layer remains human. And while it's thin, that does not mean it's not important.
Here is where I'm going to tell you what survives. This is the part where I'm going to say: no, we are not done being people at work. There's a lot of exciting work ahead. In fact, there's more exciting work ahead, I would argue, than there was before.
Product vision — not the PRD, but the upstream conviction about what your product is and who it's for. That's a human thing. Brand as deep thought work — not the guidelines document, but what your company means. Genuine care and customer relationships — reading between the lines of what a customer says to understand what they deeply need. Engineering architecture as systems thinking — designing under deep uncertainty at the frontier to bring a product to life. And ultimately, designing agentic systems themselves: building, tuning, and evolving agentic harnesses, figuring out how to teach this new skill to others in the org, learning, prompting, figuring out how to build internal tools.
Someone has to define all of this, and someone has to figure out across all of these different fields: what are the task decomposition strategies? What are the design verification criteria? What are the ways we debug failure modes? This is all human work. This is an entirely new set — or family — of disciplines. And within a few years, it will be the most important operational competency in technology.
Notice what all of these have in common, though. They're the hard skills — the ones requiring the deepest expertise, the most accumulated judgment, the highest-stakes decision-making. And they're the skills where coordination overhead has been starving us of time to do these things for decades.
The typical product leader spends at best 5 to 10% of the week on product vision, and that's really optimistic. A senior engineer might spend 10% of their time on genuine architecture thinking — again, at best. So much of the rest is meetings, alignment, sprint ceremonies, cross-team negotiation, and just junk.
The coordination task did not just waste our time. It suppressed the highest-value work that we people are doing. And so removing that work and giving it to agents does not diminish the role of humans in organizations. It concentrates human effort on the work that was always the most important — the stuff we said we'd get to, and that we were always too busy coordinating to do properly. I would like that work back. Thank you very much.
The Block Layoffs: What Dorsey Said vs. What Actually Happened
I'll be honest — the math for a typical 200-person tech company is pretty brutal. Because if you were spending most of your people on coordination roles, those people may think that their work is the stuff that they complain about. Their work is the meetings. Their work is the artifacts that just go to make more meetings. And it's not. That's not the reason the company is there. The company is there to produce value.
I recognize that this is a massive change. And as much as I am excited — because it asks us to have a deeper vision of work and the value that we can bring to the table — it is a massive, massive shift.
I know that the standard narrative frames this as an automation story with a floor: you automate the automatable tasks and then you stop. This is the narrative we saw most recently when Jack Dorsey from Block told the stock market that he was automating with AI and cut a bunch of the people at his company — when in fact what he was doing was correcting for overhiring. But he used the same narrative. He said he was going to be able to do more with less.
The thing I want you to recognize is that that perspective is limited. It is absolutely true that you can delete an enormous number of coordination functions inside a business, and that those coordination functions are sometimes held dear by some people as if they are the heart of the role. But the heart of the company and the heart of the role has always been about things that are more interesting — about things that allow us to get closer to the work and the customer.
100% of Your People Touching the Product
The secret of agents starting to move code into more of the workspace — starting to convert more of our work into something that is closer to the final product — is that we get to touch that product more. We get to dig into the real work ourselves more. We get to get fingertip-close with the product.
Imagine instead of a world where only 20% of your people — the engineers — touch your actual product, 100% of your people touch your product. What would that look like? How much more could you accomplish if that were true? That's the challenge that we have in front of us. That's the opportunity for human-friendly work that we have in front of us. That is why AI is showing us the job was never the job. And it's actually good news.
The Two Qualities That Determine Who Thrives
The qualities that stand out for people who make this shift — there are two that just keep popping up over and over again in people from a really wide range of backgrounds. I've seen this in people who were educated in music. I've seen this in people who are developers. I've seen this in people who are deep into product, deep into customer success.
The two things that seem to persist in people who make this adjustment — who see this world coming and say, "That's a wave I'm going to catch, not a wave I'm going to get drowned under" — are agency and ramp.
Agency is: I'm going to sort this out. This is a skill issue. I can learn this. I can figure this out. That attitude in spades. And ramp is the ability to learn quickly, be extremely curiosity-driven, and just go dive into it. Those two qualities together occur in every single person I know who is jumping into this head-first and thriving.
So if that's you — you're looking at your calendar next week and thinking, "Wow, I have a lot of meetings I hate" — this is one situation where AI is actually coming to help. The two qualities you need are high agency and high ramp. Lots of curiosity to learn.