Most org charts make sense when you understand how they got there.
Expertise had to live somewhere. Information was expensive to move. Software was specialized. Work crossed functions through meetings, reports, tickets and managers because that was often the only practical way to coordinate it.
Those constraints shaped the company.
What interests me now is that some of them are changing faster than the org chart.
The old boundaries were not arbitrary. They were expensive.
Departments solved real problems. They concentrated expertise, created accountability and gave leaders a manageable way to divide work.
But every boundary also introduced a handoff. Marketing handed something to sales. Operations waited on technology. An analyst produced a report for someone who then translated it for someone else. A manager existed partly because information had to move through a person.
That arrangement can still be exactly right.
It just should not be treated as permanent because it is familiar.
AI changes the cost of coordination.
A lot of work that once required a chain of specialists can now happen much closer to the person who owns the problem. Research, analysis, prototyping, workflow automation and even lightweight software development are obvious examples.
That does not eliminate expertise. In some cases it makes expertise more important because the routine production work falls away and judgment becomes the scarce part.
What it does change is the number of handoffs we should accept without asking questions.
If a marketing leader can build a working prototype without entering a traditional development queue, does every customer experience need to follow the old path? If an operating team can query structured data directly, should recurring reporting still be a production process? If AI can translate information between systems and functions, which coordination roles are still necessary and which ones are artifacts of how hard information used to be to move?
I would not start this exercise by asking which department to cut. That gets political fast and usually produces the wrong conversation.
Start with an outcome instead.
Acquire a customer. Launch a product. Resolve a service issue. Produce a forecast. Then map the actual work required to get there.
Which steps require judgment? Which require deep expertise? Which are rules? Which exist because one system cannot talk to another? Which are approvals that protect the business, and which are approvals that simply accumulated over time?
Only after that would I draw the organization around the work.
The test is simple: would you build the boundary again?
That is the question I keep coming back to.
If we started today—with modern data systems, automation, AI and none of the inherited reporting lines—would we put the same work in the same department? Would we create the same management layer? Would we require the same handoff?
Sometimes the answer will be yes. Regulation, risk, scale and specialist depth are real reasons to preserve structure.
Sometimes the answer will be no.
I am less interested in predicting the death of departments than in separating what is structurally necessary from what is merely inherited.
Technology has changed the economics of coordination. Eventually, organizational design has to notice.