Growing without hiring new people for every step in revenue

Why headcount growth eats into margin

Ask a traditional consultant how to get from ten to twenty million in revenue. The answer is a spreadsheet: new sales people, new account managers, someone in HR to administer the new hires. Costs grow roughly in step with revenue. Margin stays flat at best.

For a long time that was the only way, because thinking work could only be done by people. Anyone who wanted to process more orders needed more clerks.

For a growing share of that work this no longer holds. Reading orders out of emails, preparing quotes, asking follow-up questions, reconciling data between systems: programs can do this reliably enough today to take load off a person instead of just offering a form. What this work costs has been falling fast for years. We described how fast in Cognitive Deflation.

What Coase says about the size of firms

The economist Ronald Coase asked in 1937 why firms exist at all. Why not contract every task out to someone on the open market?

His answer: because of the costs every single engagement brings. Finding someone, briefing them, checking their work, paying them. For every small task that is too much effort. So we hire permanently and trade flexibility for less coordination.

Exactly these costs shrink when a task is taken over by a system that is already set up. It does not need onboarding, because it accesses the company’s instructions and data. It does not wait for the next job, it only costs when it runs. For part of the work that removes the reason to create a position for it.

The firm of the future is therefore more of a small core of people who decide, know the customers and handle exceptions. The recurring work in between runs in systems that these people steer.

Two weaknesses of the linear model

The model “more revenue, more people” has two weaknesses that weigh heavier with every growth stage.

Coordination grows faster than the team. Fred Brooks described this for software projects: with every additional person the number of connections in the team grows faster than the number of people. Beyond a certain size you hire managers to coordinate other managers. The share of administration grows and productivity per head falls.

Staff cannot be adjusted quickly. In a slow period you cannot have fewer people at short notice. In a busy period it takes months until new people are found and trained. You are structurally always late, in both directions.

A system does not have these two problems in this form. It costs per case, not per month. In a quiet month it costs little, in a strong month it simply runs more often. It is not entirely free of fixed costs, because servers and upkeep remain. But the jump from a hundred to a thousand cases does not require a job posting.

A typical case from practice

A pattern we keep meeting in projects looks like this. Customers send orders as PDFs, by email and sometimes via messenger. Several people in order entry transfer these details by hand into the inventory system. The work is monotonous, error-prone and grows with every new customer.

The setup that replaces it has five parts.

1. One central point receives all incoming messages, whatever the channel.
2. A model that can read images and PDFs pulls out article numbers, quantities and delivery addresses.
3. A second model checks the details against stock levels and the customer master in the inventory system.
4. When something is unclear, the system itself sends a follow-up question to the customer, naming the details that are missing.
5. Only when everything is checked is the order created. Whatever cannot be checked goes to a person.

What changes is less the number of people than their job. Order entry becomes exception handling. How much this saves depends on volume, on the quality of incoming orders and on the interface to the inventory system. We deliberately give no figure, because without those three inputs it says nothing.

Both models side by side

	Growth through staff	Growth through systems
Costs at higher volume	rise roughly with revenue	rise per case, noticeably flatter
Knowledge	sits in heads and leaves with them	sits in instructions and data that stay
Response time	months for search and onboarding	days to weeks for adjustment
Exceptions	handled well by people	must be handed back to people
Risk in a slow period	fixed costs remain	costs fall with volume

Why hardly anyone sells you this

Agencies, consultancies and IT service providers earn on hours and day rates. A setup that abolishes work also abolishes the billable hours. That makes nobody an adversary. It does explain why recommendations often end up as more staff or more licences. More on this in The Great Filter and in Legacy Tax.

What this means for the balance sheet

A salary is a running expense. When the person leaves, what they knew leaves too. A system you run yourself behaves differently. The instructions, rules and data stay in the company even when people change.

In accounting terms it is not as clear-cut as it is often presented. Use of language models is billed per call and is therefore also a running expense. Whether you may capitalise your own development depends on your accounting rules and is a question for your accountant. The real difference is not in how it is booked. It lies in the fact that the system can be improved with every case it handles, because every case is documented. How company knowledge can be captured this way is covered in Why Your Most Valuable Employee Is a Vector.

How to tell whether it is working for you

The simplest metric for this is revenue per employee. It is rough, but over several years it shows whether your growth comes from structure or from headcount. If revenue rises and the metric stays flat, you are growing linearly. If it rises along with revenue, you have moved tasks out of heads and into systems.

Three questions you can ask before the next job posting.

1. Can the task be written down? If you can describe the work in a set of instructions with clear rules, it is a candidate for a system. If the instructions would have to defer to experience in every second case, it is not.

2. How often does it occur? A setup only pays off above a certain volume. When that point is reached depends on the effort of building it and on the cost of the current way. That is a calculation, not a guess.

3. Who handles the exceptions? A system needs someone who sees errors and adjusts the rules. That person is not a saving, they are part of the setup.

The technology is here. Whether it works for you depends on your starting point. If you would like an honest assessment, you can reach us via the contact page.