The company of 2030

Organisations take the shape of the information they process. AI is changing how information moves inside companies more than anything since the personal computer, so it is worth asking what a mid-sized company will look like by 2030.

Shape follows information

Twentieth-century companies grew middle management because someone had to collect information, summarise it, pass it up and pass decisions down. Personal computers removed the typing pool, but the coordinating layer survived, because summarising, chasing and reporting still needed people. Much middle management today is exactly that: finding out how things are going, and telling someone else.

What changes by 2030

Agents are good at precisely that work. A project that reports on itself needs no one to ask five people for updates. We expect three consequences.

Teams grow more slowly than the work. Not through mass dismissals, but because companies reach their next size without another coordinating layer.

Roles get sharper. The salesperson sells, the engineer builds, the service lead solves the hard cases.

Decision rights become explicit. When agents act, the business must write down what they may do and who approves the rest.

Approval rules as an organisation chart

Today's chart shows who reports to whom and says little about who may decide what. Approval rules are a precise map of authority: this agent may draft and schedule; this person approves anything sent to a customer; discounts above a set level go to that person. The map can be read, tested, changed and audited, which makes delegation explicit and reversible, and the quality of decision rules a competitive factor.

The old chart shows who reports to whom. The new one shows who may decide what.

Managers as editors

An editor does not write every word. An editor decides what matters, reviews the work, sends it back when it is wrong and takes responsibility for what is published. That is the manager of agent-assisted work: a more demanding job than coordination, not a lesser one. Companies will need to promote people for the quality of their decisions, not for gathering and relaying information.

What stays the same

Companies will still win on trust, a few key relationships and people who care. These will matter more: processes that live in software anyone can buy are easy to copy, while a culture of good decisions and people who know customers by name are not. With fewer routine moments of contact, leaders will also have to create on purpose the shared experiences old ways of working provided by accident.

The case against

Predictions about the organisation have a poor record; the paperless office and the death of hierarchy were both announced early. And AI projects often fail. In 2025, 42 per cent of companies abandoned most of their AI initiatives.¹ MIT's NANDA initiative found that buying from specialist vendors succeeded about 67 per cent of the time, against about 33 per cent for internal builds.² Gartner put transformational generative AI deployments at $5 million to $20 million.³

We read these as a warning about how change happens, not whether. Large transformation programmes cost more than a mid-sized company can risk. The companies that reach 2030 in a new shape will get there one process at a time.

What this means for a mid-market leader

Question the next coordinating hire. If the role is mainly finding out and telling, ask whether agents can do it.

Write down your decision rights. Who may approve what, and up to what limit. It will show where authority is unclear today.

Develop managers as editors. Train them to review, challenge and decide.

Buy before you build, and start with one process. The evidence suggests that is where success lies.

The company of 2030 will arrive through sensible changes, each making the organisation a little lighter in coordination, clearer in authority, and more human in how its people spend their days.