Vendors Now Hire Staff to Define What Success Means for You

← Back · ACT II · Chapter 14: Who Controls Value Engagement?

Boris Cherny, who leads Claude Code at Anthropic, published a short observation about how AI is reorganizing engineering work. He argued that titles like engineer, designer, product manager, and analyst are becoming lagging indicators, and that what actually persists is a set of five archetypes: Prototyper, Builder, Sweeper, Grower, Maintainer. The mix shifts as a product matures. The archetype, not the title, describes what a person does.

It is a good observation about engineering. It traveled quickly, and it did not stay in engineering.

A value-management software vendor took the archetypes to its own function and extended them. Its additions are worth reading closely, because one of them is not a description of work at all. It is a claim of custody.

Three Additions, One Consequential

The extensions were a Scout, who continuously discovers customer patterns, benchmarks, and emerging value preferences. An Orchestrator, who coordinates people, agents, customers, partners, and systems. And an Editor, whose stated remit is to govern outcome frameworks, ontology, methodology, and the quality of AI-generated recommendations.

Scout and Orchestrator describe activity, and the Editor describes authority, which is not the same kind of thing at all.

Ontology is a different kind of object from a workflow. It is the set of decisions about what things are, what counts as an outcome, which outcomes are commensurable, and what evidence is admissible against them. Whoever governs the ontology has already determined what any subsequent measurement can conclude. Everything downstream, every dashboard, every quarterly review, every renewal case, inherits those decisions and cannot argue with them.

The vendor is not concealing this. It published it as a role design.

A Role Is More Durable Than a Deliverable

Sellers have always supplied buyers with value models. That is not new and it is not the story here.

What is new is the institutional form. A deliverable is an event. It arrives, it is reviewed, it can be rejected, and it expires. A staffed seat is continuous. It has an occupant, a mandate, a budget line, a successor when the occupant leaves, and a standing claim to the question it was created to answer.

When the seller staffs the ontology seat and the buyer does not, the question of what counts as value has been permanently assigned. Not negotiated. Assigned. The buyer will still approve things. It will approve inside a structure it did not build and cannot see the edges of.

This is the difference between buying a service and inheriting a standard. You can replace a service. A standard you have absorbed keeps governing you long after the vendor is gone.

The staffed seat is also not the only form this move takes. In agent-to-agent commerce the same custody is arriving as a machine-readable schema. A seat and a schema are the same claim in two forms: the definitional layer, held on the seller’s side of the table.

Automation Makes The Asymmetry Permanent

Cherny’s original point was that AI changes how work is organized, not merely how it is performed. That is correct, and it is what makes the extension consequential rather than merely interesting.

An ontology held in a slide deck degrades, because people argue with it, forget it, and reinterpret it in the next planning cycle. Encode that same ontology into the system that generates the recommendations and it stops being arguable and simply runs. The Editor seat exists to govern the quality of AI-generated recommendations, which means the ontology is being written into the generator, not into a document.

At that point the definition stops being a position anyone holds and becomes the substrate everyone reasons on top of. It is no longer visible as a choice.

The Missing Seat Is On The Buyer Side

Ask a straightforward question of any organization. Who here governs the definition of what our investments are supposed to produce, and where is that definition written down in a form that a vendor could be measured against?

In most organizations, no one holds that. Finance owns the budget. Procurement owns the contract. The sponsoring function owns the deployment. The definition of the outcome itself is unowned, which is precisely why it is available to be supplied. This site has documented that vacancy before: the standard has no custodian.

It will be supplied. It is being staffed for right now, on the other side of the table, by people who are good at it and who are open about doing it.

What The Buyer Has To Hold

The answer is not to reject the vendor’s model. Vendor models are often better built than anything the buyer would produce alone, and blanket refusal leaves the buyer with nothing to test against.

The answer is to hold the ontology as a buyer-side asset before any vendor model is admitted. What counts as an outcome here. What evidence is admissible. What baseline stands. Which definitions are ours and are not open to substitution. Declared first, in writing, by the organization that will live with the consequences, because the buyer sets the definition.

That does not require the buyer to mirror the hire. Value does not need a new seat; it needs a standard, and a declared standard is what makes any seat, on either side of the table, accountable to something it did not write.

A buyer that has declared its own definitions can accept a vendor’s model as an input and test it. A buyer that has not declared them cannot test anything, because the model and the yardstick arrived in the same box.

The seat is being filled. The only open question is whether anything on the buyer’s side of the table is written down for it to answer to.

Sources: Boris Cherny (Anthropic), public post on engineering archetypes, 29 June 2026, describing the Prototyper, Builder, Sweeper, Grower and Maintainer archetypes. The extension of those archetypes to customer value management, including the Editor role described as governing outcome frameworks, ontology, methodology and the quality of AI-generated recommendations, was published by the chief executive of a value-management software vendor and is described here structurally.

Keep reading: ACT II · Chapter 14: Who Controls Value Engagement?