What if every new box on the org chart is a loan against company speed that nobody plans to repay? Founders draw the chart when the week feels heavy. The interest accrues quietly as lag between decision and action, as translation layers between the people doing the work, and as meetings whose only job is keeping the boxes synchronized.
There is no single correct shape. Flat structures get credibility because early tech and AI companies live and die on speed, and hierarchy is the natural enemy of speed. Traditional charts get credibility because new hires can answer two simple questions: who do I report to, and where is my support. Both can be right on Tuesday and wrong on Thursday. Structure should follow the work, not the latest mythology about how legendary companies run.
If the company still has heavy research and product discovery left, departmental fiefdoms are a tax. Territorial fights over ownership slow the only thing that matters. In that phase the founder’s real job is making sure people talk to each other. Flat is a coordination choice. If the business is a multi-unit operation with repeatable local systems — think restaurants that franchise cleanly — the chart mostly draws itself. Each unit needs a manager; managers need shift leads; pretending that is Jensen’s sixty-direct-report model is cosplay.
Even famously flat tops hide nested structure. Leaders with dozens of direct reports usually have mini-CEOs underneath, each running some internal shape of their own. Flat at the apex does not mean formless everywhere. It means someone chose where the bottlenecks should live on purpose.
Staffing timing matters as much as shape. Overstaffing feels responsible and often is just anxiety with a budget. Seats added before the work is clear create managers of work that does not exist yet, then invent process to justify the headcount. Hiring behind growth forces prioritization and keeps people close to customers and product. The especially expensive hire is the first manager brought in because the founder is tired of management tasks. Management is psychological safety, accountability without humiliation, and an environment where people feel seen and still responsible. There will never be an AI manager for that. AI can process time-off requests. It cannot build culture.
AI does change the economics of work, just not the way org-chart cosplay claims. Modern companies run on knowledge work: creativity, judgment, experience, intuition. Treating AI as a headcount deletion machine confuses mechanical chores with expertise. Building agents to “replace” skilled roles often devalues the expertise those roles required. AI becomes powerful when an experienced person stays in the loop. It can also destroy productivity. Founders burn hundreds of hours on the wrong AI task because it felt like progress. The real opportunity is the twenty to thirty percent of most jobs that is mechanical or administrative: research that should land as a dashboard, outreach that should not mean portal slavery, bureaucracy that makes skilled people miserable.
Use case zero for company-wide AI is training and adoption for everyone at a keyboard — real tools, real practice, systems built around practitioners. At scale the scarce profile is the person who understands the business strategically and knows how to automate functions inside it. Job shapes change. They do not disappear into a prompt.
When a company hits the bloated hundred-person stretch and margins tighten, tactical downsizing arrives. Annual bottom-ten rituals and “trim the fat” language can buy a short-term bump founders mistake for a long-term trajectory. Permanent job fear is not a healthy operating system. Careful hiring standards matter. Confusing fear with excellence does not. The honest audit asks whether the organization drifted from purpose, whether boxes were loans against speed that never got repaid, and whether managers were hired to avoid management rather than multiply it.
Org design is a series of loans. Take them deliberately. Match structure to the actual work. Keep managers human. Use AI to strip bureaucracy from skilled people, not to pretend expertise is optional. And treat every new box as interest-bearing debt against the one resource early companies cannot buy back cheaply: speed.
Watch the Full Episode on Lean Org Structures below: