In a soft market, the first line items cut are the ones nobody can explain to a CFO. That is the entire argument for running engagements on receipts.
The cut order is not random
When a budget gets compressed, finance does not cut by category — they cut by defensibility. A line item with a named outcome, a runbook, and a monthly cost estimate is defensible in a five-minute conversation. A line item that reads "AI transformation retainer" is not. The retainer goes first; the receipts line survives. We treat this as a structural property of how budget reviews work, not a prediction (Class C — the configuration of the engagement determines which side of the cut line it sits on).
This is the part of the partnership-vs-vendor argument that gets framed wrong in good markets. In a good market, the receipts discipline looks like a cost — more documentation, more runbook writing, more falsifiers in the contract. In a down market, the same discipline is what keeps the engagement off the chopping block, because the work has a name a non-technical executive can defend.
What a vendor relationship looks like under pressure
A vendor relationship in a soft market degrades in a predictable sequence. The monthly check-ins get shorter. The deliverables get vaguer — "ongoing optimization," "platform tuning." The renewal conversation surfaces a clause nobody remembered signing that makes leaving expensive. The client either pays the exit cost or pays another year of the retainer, and either way the vendor wins.
We have watched this happen in engagements where we replaced the prior vendor. The prior vendor had delivered something real in month one, and then the relationship calcified into a flat-fee maintenance contract whose actual scope had drifted to almost nothing. The client could not cut it because they could not tell what would break if they did. That is the trap a vendor relationship is structurally built to produce.
What a partnership looks like under the same pressure
A partnership engagement under pressure does something different: it shrinks honestly. Because the ledger lists every artifact, every dependency, and every cost line with a substitution note, the client can walk into the budget review and say "we are keeping these three artifacts, retiring this one, and the monthly run cost drops from X to Y." The engagement gets smaller without getting opaque. The relationship survives because it never depended on the client not understanding what they were paying for.
This is the counter-cyclical part. The receipts discipline is more work to set up, and that work pays back exactly when a vendor relationship would have been cut.
What the hiring side already knows
Jay Kumar Chimata, who runs a 22k-user AI talent platform, did an interview with Themesis covering what employers in 2026 are actually paying for versus what candidates are studying: Meet Jay Kumar Chimata: JobFirst.ai and the Real AI Job Market (Class E — primary-source interview with the operator of a hiring marketplace).
Our one-line read, in our own words: when hiring tightens, the candidate who can hand over a runnable artifact and explain its failure modes gets the offer; the candidate who can only describe a vendor's UI does not. The agency-side version of that asymmetry is the same. In a down market, the engagement that ships an exit-ready artifact survives the cut, and the engagement that ships a login does not.
The questions to put in the budget review
If you are inside a company that is about to compress its AI services spend, these are the four questions worth asking each line item before deciding which to keep:
- What artifact do we own from this engagement? Can someone outside the team re-run it from a runbook?
- What is the monthly run cost if we keep only the artifact and end the engagement?
- What is the falsifier that ends this engagement, and has it fired?
- Where is last quarter's ledger?
A line item that cannot answer those four is, by our working definition, a vendor relationship — and it should go first. A line item that can answer them is a partnership, and cutting it is throwing away the receipts you already paid to build.
What this is not
We are not arguing that every engagement should survive every budget cut. Some of ours should not. The point of the receipts discipline is not that the work is sacred; it is that the keep-or-cut decision is made on the merits, with the evidence on the table, by the people whose money it is.
That is the whole moat. It is not glamorous, and it is exactly what holds up when the market stops being kind.