# Buying Instruments.

Type: Note
Date: 2026-06-01
Tags: SaaS, enterprise software, tools for thought, JA Westenberg
Location: Sydney, Australia
Canonical: https://www.aaronroot.net/journal/pseudo-markets

Offer someone a house, a car, or the weekly shop for half the usual price and they pay attention. Offer an organisation the same software for half the price and often nothing happens. [JA Westenberg](https://www.linkedin.com/pulse/pseudo-markets-ja-westenberg-d66vf/) calls these pseudo-markets. They have prices, procurement teams, RFPs, and vendor scorecards, so they look like markets. But price often does not decide the deal. Status, regulation, switching costs, and whether the choice can be defended later matter more.

This is familiar in B2B SaaS. The person who buys the software is often not the person who uses it. A CIO buys the ERP. A procurement team approves the platform. Then thousands of people work inside that decision for years. The buyer is insulated from daily use, so the purchase favours defensibility. The safe vendor. The approved category. The decision no one gets blamed for making.

For workflow software, that often makes sense. Organisations need convergence. Shared records, common processes, permissions, reporting, audit trails. Known work benefits from fewer versions of the truth.

But AI is moving software back towards senior buyers. After years of bottom-up adoption and product-led growth, many tools are being bought top-down again. Security, data access, legal risk, model governance, cost control, and executive anxiety all push AI tools towards procurement.

That matters because many AI tools are not just workflow systems. At their best, they are instruments for thought. They help people model, write, design, research, explore, and notice things the official process has not yet named. Their value comes from useful divergence.

Procurement is good at seeing the cost of variety. More licences, more vendors, more duplicated data, more security risk. It is worse at seeing the cost of sameness. The idea that never forms. The weak signal that gets missed. The person who stops experimenting because the approved tool cannot support the way they think.

This connects to the factory and lab distinction in [the conversation between Max Schoening and Geoffrey Litt](/journal/cognitive-debt). A factory is for work that is already understood. A lab is for work where understanding is made by doing the work.

Procurement can buy for both, but each needs a different question. Factory software should reduce variation. Lab instruments should create useful variation. Treat one like the other and the tool will be judged by the wrong standard.