Your own people write the specifications. Your orchestrators govern the agent fleet and the data it reaches. The record is institutional memory rather than an audit file. We move onto the gate and out of the work, which is the only ending we are interested in.
When machines absorb execution, the roles that matter are the ones that decide what good looks like and answer for the result. A principal company is organised around that: a core of people who specify and answer, orchestrators who govern the agents and the data, and agents underneath doing the volume. We run this way ourselves, which is the only reason we are willing to sell it.
They decide what good looks like, keep the gates that carry consequence, and are named against outcomes. Fewer of them than a traditional structure, and each one has more reach.
They own the specifications, the permissioned paths and the agent roster. A role most companies do not have yet, and the one that decides whether any of this scales.
Working from specifications, stopping at gates, leaving a record. Capacity you can add without a hiring round.
Multi-year, and the whole design points at us being unnecessary in the work.
Your people learn to say what they mean precisely enough to survive literal execution. This is a skill, it takes months, and without it every agent stays fragile.
Defined, and filled from your own organisation wherever possible. We train into it rather than occupy it.
Every application, and eventually every supplier's agents, running on the same specifications, gates and record.
An accumulating account of how your work goes wrong and what you did about it. It answers auditors, it survives staff turnover, and no competitor can rebuild it.
Agents absorb work that people are doing today. We do not start this without a named plan for those people and what they move on to, because a transformation that only produces a headcount saving does not survive contact with the organisation it is done to.
The people who know the process best are the ones best placed to specify it and to keep its gates. That is a promotion in substance, not a redeployment on paper.
Most companies do not want fewer people. They want more output from the ones they have, and a way to grow that is not linear in headcount.
Who is affected, what they move to, and when. We will not start a programme without it, and you can hold us to that.
The economics push everyone the same direction, so the destination is not the differentiator. Being three years into it with the practice in your own people, the ground already laid and a record nobody else has is a position competitors cannot buy their way into quickly.
Our working paper sets out why the firm is changing shape and what replaces it. The operating model is that argument, installed.
Multi-year, and built so your own people run it. Start with one application and a number.