You Are the Middle Manager

The middle manager Jack Dorsey wants to eliminate is you.

Block's CEO posted in April that there's "no need for a permanent middle management layer," and said he was compressing his org from five reporting tiers down to two or three. Meta's Reality Labs swapped traditional supervisor titles for "AI builder," "pod lead," and "org lead." Atlassian announced 1,600 layoffs in January with 900 cuts coming from research and development, framed as an AI pivot, despite Mike Cannon-Brookes telling 20VC five months earlier that the company would employ more engineers in five years. Salesforce cut 4,000 customer support roles and replaced them with AI agents, dropping that group's headcount from 9,000 to 5,000. Workday gutted Global Customer Operations.

The story has a name. Press is calling it the Great Flattening. Analysts at Gartner project that one in five firms will use AI to compress their middle management ranks by the end of 2026.

You are a founder running 22 people at $3M ARR. You read all of this. You start to think about flattening too.

Here's where the trap opens. The founders “cargo-culting” the Block playbook into a $3M ARR company are about to make their bottleneck worse than it has ever been.

Stay with me on the layer math for a second, because the geometry is doing more work in those headlines than anyone is saying out loud.

Block has roughly 6,000 employees. Meta has 70,000-plus. Atlassian sits around 12,000. Salesforce has more than 70,000 even after the cuts. Each company sits on top of an executive bench: a CEO, a COO, a CFO, a CTO, heads of functions, regional leads, lifers in seats with budget authority.

Below the bench, the middle layer. People owning teams of 8 to 15. Strategy comes from above. Execution lives at their feet.

A middle manager at that size splits the week into two buckets. Translation work, where they convert the executive's call into something a team can ship this quarter. And admin work: standups, ticket routing, performance reviews, status compilation, the weekly "tell me what your team did."

The agentic AI argument lives in the second bucket. Modern tools absorb scheduling, status digests, ticket routing, first-pass perf summaries, the quiet logistics of running a team of 1,000 indirect reports. If you have 5,000 employees and 400 middle managers spending 60% of their time on admin, you can compress that layer with software and save real money.

Now run the same math at your company.

Last month, 48% of visitors to documentation sites across Mintlify were AI agents, not humans.

Claude Code, Cursor, and other coding agents are becoming the actual customers reading your docs. And they read everything.

This changes what good documentation means. Humans skim and forgive gaps. Agents methodically check every endpoint, read every guide, and compare you against alternatives with zero fatigue.

Your docs aren't just helping users anymore. They're your product's first interview with the machines deciding whether to recommend you.

That means: clear schema markup so agents can parse your content, real benchmarks instead of marketing fluff, open endpoints agents can actually test, and honest comparisons that emphasize strengths without hype.

Mintlify powers documentation for over 20,000 companies, reaching 100M+ people every year. We just raised a $45M Series B led by @a16z and @SalesforceVC to build the knowledge layer for the agent era.

Make Your Docs Agent-Ready

You have 22 people. Two team leads. Each of them runs a group of seven or eight. There is no executive bench. There is no COO. There is no CTO with a director layer between her and the engineers.

It's you, your two team leads, and 18 people doing work.

Look at what those leads spend their time on. I'd bet most of it is translation. They take whatever you said in Monday's all-hands and turn it into the three things their team will ship this week. They explain why a priority shifted. They mediate between the engineer who wants to refactor the auth module and the customer who wants the integration finished by Friday. They sit in the customer call your sales hire just botched and reconstruct what the customer meant. They tell you, on Wednesday, that one of the engineers is one bad week away from quitting, while there's still time to do something about it.

A small fraction of their week is the admin job from the press releases. They run standups. They write a sprint update. They sit through your Tuesday leadership sync. Those are the parts that look like the AI-replaceable middle-manager work the headlines describe.

If you flatten the leads out of the structure, you delete the translation. The admin was 15% of the role. The other 85% was the work buffering you from doing it yourself.

The week the flattening lands on your org chart, your direct-report count goes from two to 20. The standup runs through your calendar. The customer escalation routes to your inbox. The new engineer who needed two weeks of careful onboarding now needs you to do it. The Tuesday Slack from your team lead saying "Mike is checking out, I think we're going to lose him" never arrives, because Mike now reports to you, and Mike is not going to tell you he's checking out.

You absorbed every job the layer was holding. The only thing that flattened was your week.

He's a founder I've been checking in with for a few months. The day after Dorsey's memo went around X, he pulled the trigger.

He has 22 people. Two leads. About $4M ARR. Reasonable margins. Nothing on fire. He cut both leads back to senior IC roles and announced it to the team as a flattening.

The first couple of weeks he loved it. He was in every standup. He saw every pull request. He sat in on every sales call. He told me he hadn't felt this connected to the product since they were five people in a co-working space. He used the term "founder mode" without irony.

Six weeks in, our check-ins have a different texture.

His sales team is waiting on him to weigh in on every deal of meaningful size, because the leads who used to make those calls are now ICs without the authority. He's been behind on deal reviews for two weeks. A customer QBR got rescheduled twice and the customer is annoyed. The product hire who joined right before the flattening was supposed to be onboarded by one of those leads; that hire now reports to him directly, and he's been too underwater to give her more than a quick check-in since the move.

Every one of those traces back to the same root cause. He absorbed the layer he flattened, and he's now doing its work himself.

Under that is something worse. The leads sitting in IC seats still get pulled into the decisions they used to own, because the team's muscle memory takes months to forget. They answer differently now. They hedge. They send the question up to him and let him make the call. A fresh demotion makes people small for a while, and these two are exactly that fresh. He kept them on the payroll and lost the thing he was paying them to do.

He hasn't sent me the "I think I broke our company" message yet. In my experience watching this play out, that message lands somewhere between week six and week 10. He's at the front edge of that window right now.

I'm writing this newsletter to keep the next 50 founders from sending it.

Here is the part that's hard to hear when the news cycle is full of CEOs being celebrated for doing the same thing.

The playbook the headlines describe solves a problem you do not have, and creates a problem you cannot afford.

The problem big-company flattening solves is that a 6,000-person org accumulates managers as a function of headcount. When the technology that handles the routine parts of management improves by a generation, you can pay for fewer of those managers. The operating-expense delta is hundreds of millions. The press loves it. The stock loves it. The analysts model it. The CFO has a number to point at.

Your flattening creates a different problem. You, personally, were already the binding constraint on every part of the company before you started cutting layers. Sales escalations. Hiring panels. Pricing exceptions. Custom contract terms. The architecture decision your CTO is half-sure about. The Friday-night fire when production goes down and somebody has to call the customer at home.

The middle layer in your company is the buffer absorbing the work that would otherwise hit you. When you flatten that buffer, the work doesn't disappear. AI does not absorb a customer escalation about a contract clause your sales team wrote six months ago. Claude does not pick up the call from your largest account when the integration breaks at 2 PM Eastern. An agent does not have a one-on-one with the senior engineer who is about to take a Stripe offer.

The CEOs being quoted in the headlines all sit on top of executive teams that absorb the work the middle layer used to handle. You sit on top of nobody. You are the executive team. You are also, now, the middle layer. You will burn through your hours faster than any agentic tool can give them back.

A startup making the rounds in the press has one junior employee and lets an AI handle performance feedback, task allocation, and skill development. Read it out loud. The founder is the executive layer, the senior IC, the customer-facing layer, and the manager of the one employee. That story is a gif of a man on a unicycle juggling chainsaws.

Run that math against your own company. You have customers, you have a team, you have seven directions to push in this quarter, and one finite human nervous system to do it with.

Ready to transform artificial intelligence from a buzzword into your personal revenue generator

HubSpot’s groundbreaking guide "200+ AI-Powered Income Ideas" is your gateway to financial innovation in the digital age.

Inside you'll discover:

Download your guide today and unlock a future where artificial intelligence powers your success. Your next income stream is waiting.

Get Your Guide

Here is what flattening should do at your scale. It should pull you out of the execution work and protect the layer that sits between you and it.

Here are five moves I'd take this week with a founder in your shoes.

Audit what your leads do. Sit with each of them for an hour. Pick a recent day that wasn't a fire drill, and walk it through in 15-minute blocks. Put each block in two columns: translation, or admin. If your leads are 80% translation, the AI flattening argument doesn't apply to them. If they're 80% admin, you have a different problem, which is that you hired the wrong leads or you scoped the role badly. Fix the role before you cut the layer.

Audit what you do. Same exercise on yourself. Pull a recent normal day into 15-minute blocks. Look for the work that should not be on the founder's calendar at $3M ARR. First-pass code reviews. Standup attendance. Status check-ins. Customer support tickets you handled because the response time was bad. Sales discovery calls below your ACV threshold. Onboarding the new engineer. Any of these on your week is a sign that you are the layer doing layer-work.

Use AI where the big companies are using it, but inside your own job. Status compilation, meeting summaries, customer-call transcripts with action items, weekly priority drafts, first-pass writeups on inbound. Any agentic tool that gives you back two hours a day on admin. The flattening argument the press is selling applies cleanly to one role at your company, and that role is yours. Take the time back from your own job and reinvest it in the executive work nobody else can do.

Give your leads a raise. I know this sounds backward when the news is about cutting them. The leads doing translation work at a growth-stage company are some of the highest-impact people you've ever employed. They are why you can think for more than 11 minutes at a stretch without an interruption. If they leave because the comp band is wrong or because the Block headline made them nervous about their own stability, you will pay six figures and four months to replace each one. Run a comp check, give them a meaningful raise, and tell them out loud what they are doing for the company.

Resist the urge to broadcast a flattening as a strategy. I have watched several founders in the past month frame "we are getting flatter" in town halls as if it were a vision statement. The team hears: cuts are coming, my role is on the list, my manager doesn't have a future here, I should update my LinkedIn tonight. Even when you are not cutting anyone, the language costs you trust. Save the strategic framing for things that are strategy. Layer changes are operational; talk about them operationally, when they happen, with the people they affect.

A February paper from the National Bureau of Economic Research found that 80% of companies actively using AI reported no measurable productivity impact. METR's developer-productivity study, revised in February, showed that experienced engineers using AI coding tools believed they were 20% faster, while objective testing measured them as slower. McKinsey's survey of 4,500 developers caught a 46% reduction in time on routine coding tasks, paired with 91% longer pull-request reviews. Three credible studies, three different metrics, one uncomfortable conclusion: the productivity claim the press is selling lives mostly inside its own coverage.

The flattening story has the same shape. Block, Meta, Atlassian, and Salesforce are running a real experiment, and we will know in 18 months whether the layer they cut was admin overhead or institutional knowledge. That is their experiment to run. They have the cash, the executive bench, and the headcount to absorb a wrong answer.

You don't. The Block experiment, replayed at 22 people, looks like a founder back in every meeting at 11 PM Tuesday wondering what just happened.

The founder I described above hasn't reversed it yet. He still has time. In my experience watching this play out, he has about two weeks before the leads sitting in IC seats start interviewing in earnest. Six weeks before the engineers stop bringing him problems they could have brought a manager. Somewhere between three and six months before a board update lands with a number he wishes he didn't have to write down.

Some version of him is going to message me in the next month or two. The note will say he thinks he broke his company. He will have signed for it himself the morning after the Block memo went around, and it will take some explaining before he sees that the cargo cult is what cost him.

I'm writing this so other founders don't have to be the ones who text.

The Block playbook is for Jack Dorsey. The Meta playbook is for Mark Zuckerberg. The Atlassian playbook is for Mike Cannon-Brookes. Your playbook is quieter, less photogenic, less likely to make TechCrunch, and a hell of a lot more relevant to the company sitting in front of you.

The middle manager Jack Dorsey wants to eliminate is the middle manager Jack Dorsey runs. The middle manager you keep meaning to deal with is the one buying you back 10 hours a week.

If a piece of news doesn't fit the company in front of you, all it does is make the bottleneck worse. The bottleneck has been you the whole time. Use the AI tools Block is using on your own admin work, not on the layer that's protecting you from it.