Services became the economy without building an operating system

Services became the economy without ever building an operating system for itself.

Manufacturing spent the twentieth century learning how to make repeatable things repeatably. Standard work, quality gates, unit economics, flow. Agriculture mechanised and then industrialised its supply chain. Services absorbed the workforce of both and imported almost none of the discipline.

The gap does not show up as bad work. Professional services firms are full of capable people doing good work. It shows up as rework — the same job described three times, the same fact re-entered four times, the same month-end reconstructed twelve times a year.

The size of the shift

The numbers are not marginal. On the World Bank’s ILO-modelled series, services accounted for 35% of global employment in 1991 and just over 50% by 2024. In high-income countries the share is about 75%, while agriculture has fallen to roughly 3%.

Half the world and three-quarters of the rich world now works in services. The operating practices most of those firms use were not designed for that scale. They were inherited from a period when a professional firm was a partnership of a dozen people who could hold the commercial position of every engagement in their heads.

That is not a criticism of the people who built them. It is an observation about what happens when a way of working outlives the size it was designed for.

Why services did not inherit the discipline

Manufacturing had an unforgiving teacher: physical inventory. A factory that loses track of its work in progress can walk onto the floor and count it. The discrepancy is visible, embarrassing and immediate.

A services firm’s work in progress is a claim about time somebody spent, under terms somebody agreed, against a scope somebody interpreted. You cannot walk onto the floor and count it. You can only ask people, and people answer honestly and differently.

So the feedback loop that forced manufacturing to get rigorous never closed in services. A firm can run for years with an unreliable WIP number, because nothing physically piles up in a corner to prove it wrong. The correction arrives eventually, in the form of a margin that was lower than anyone thought, but it arrives too late and too diffusely to teach anybody anything specific.

The six drivers

In a services firm, productivity is not about optimising physical inputs. It is about the quality of the operating layer that turns expertise into outcomes, consistently, with minimal rework. Six things move it.

Standardise delivery. Templates, playbooks and quality gates, so that less of each engagement is invented from scratch.

Structure the data. One source of truth across sales, delivery, time and finance, rather than four functions each maintaining a defensible version.

Cut hand-offs and rework. Clear ownership, fewer queues, a shorter path from sold to done to billed.

Track the right unit economics. Project margin, utilisation, write-offs, ageing WIP, cash conversion. Revenue and headcount are the metrics that hide the problem.

Price to value. Package outcomes, tighten scope, control change deliberately rather than absorb it politely.

Invest in capability. Training, coaching, and tools that remove cognitive load rather than add a system to check.

None of those is novel. Every one of them is a direct import from operational practice that manufacturing settled decades ago. The novelty is only that services firms have not generally been asked to run this way.

Where the analogy breaks

This is the part that most versions of this argument leave out, and without it the argument is glib.

In manufacturing, variability is the defect. In professional services, variability is frequently the product. A client is paying for judgement applied to a situation nobody has seen in exactly that form before. A firm that standardises its way to a factory has usually standardised away the thing it was hired for.

So the six drivers apply to the operating layer, not the expert layer. Standardise how an engagement is scoped, approved, tracked and billed. Do not standardise what the expert concludes. The distinction is not always obvious in practice, and firms get it wrong in both directions — some by refusing to standardise anything, some by writing playbooks so prescriptive that senior people stop thinking.

There is a second limit. Operational rigour has a cost and a payback period, and below a certain size the payback does not arrive. A twelve-person practice where the person who sold the work also delivers it does not have a continuity problem, and installing governance to solve one is pure overhead. The argument here is about firms large enough that commercial truth has to travel between people who were not in the same room.

Treat operations as a product

The firms that will do well as services keeps absorbing the workforce are the ones that treat their own operating model as something designed, owned and improved — not as administration that happens around the real work.

That means somebody senior owns the flow from prospect to payment, the way a manufacturer owns a line. It means the unit economics are read rather than assembled. And it means the operating layer is deliberate enough that the expert layer can be genuinely bespoke without the firm losing track of what it agreed to.

Our own view of what that layer needs to hold is set out on our management consulting page and, for technology firms, on the IT services and consulting page. The argument came before the software, and it stands whether or not you ever buy any.