Pass Through or Recast: The Decision Every Software Layer Has Already Faced

TL;DR Every layer of software that accretes capability faces the same binary: keep the supplier’s unit and pass its cost through, or choose a unit of your own and recast the metric on what the layer now delivers. The industry has faced this decision under at least three names, per-call API pricing, OEM licensing, and now the AI harness, and keeps failing to recognize it because each generation renames the layer. The value metric decision, which unit travels with the grant, precedes packaging, which precedes pricing, whatever the layer is called this year.

Somewhere right now a product team is pricing a layer built on a large language model. The layer retrieves, plans, selects tools, verifies results, retries, and assembles raw model output into finished work. The team calls it a harness, because that is the current word, and because the word is new they believe the pricing question is new. It is the third time the software industry has priced this layer. It just was not called a harness the first two times.

One Decision, Two Doors

A layer that accretes capability on top of someone else’s unit eventually arrives at a binary. Door one: pass the underlying cost through. Bill in the unit you buy, calls, tokens, compute hours, marked up enough to carry your margin.

It is legible, easy to defend on day one, and it declares your product logistics for someone else’s meter. Door two: recast the metric. Define the unit on what the layer now does for the customer, the record enriched, the workflow completed, and accept that your input costs become your problem to manage rather than your customer’s line item to audit.

Neither door is right in the abstract. Pass-through fits while the layer is thin, when what the customer buys really is the underlying unit with plumbing around it.

It strains as the layer thickens, because the unit on the invoice describes what you consume, not what the customer receives, and a buyer will eventually hold your rate card next to your supplier’s. Recasting captures what the layer adds, but it moves consumption risk onto you, and it only holds if the new unit does what a good value metric must do: the buyer understands it, can forecast it before signing, and watches it track the value received. Where the line falls for a specific layer is diagnosis, not doctrine.

The Rename Is What Hides the Precedent

What makes the binary hard is not its difficulty. It is that the industry keeps renaming the thing that faces it. Every few years a new noun arrives at the application layer: SaaS, platform, copilot, agent, harness. Each noun brings a market map, a conference track, and a confident claim that pricing works differently here. The claim survives scrutiny because the noun severs the precedent. A team pricing a harness does not go looking for how connector vendors priced, because nothing in the word suggests the histories are the same history.

Bill Aulet, Managing Director of the Martin Trust Center for MIT Entrepreneurship and Senior Lecturer at MIT Sloan, made the underlying complaint at the Business of Software Conference USA in 2014: “Entrepreneurship needs a common language. Stop reinventing terms.” His specimen was “pivot,” a new name for changing your strategy, and his argument was about knowledge transfer: “If 20% of our population is going to be entrepreneurs we need to be able to share knowledge across it.” He was talking about entrepreneurship education, not software pricing.

The extension is ours, and we make it deliberately: same disease, different specimen. Rename the layer and the accumulated knowledge about pricing that layer stops transferring. Every rename resets the industry’s memory, and the decision gets rediscovered, at full tuition, by people who believe they are the first to face it.

They are not. We have watched this decision get faced at least three times.

Sighting One: Pennies per Call, Until the Call Stopped Being CRUD

The API era opened with a unit that fit. Early integration layers billed pennies per call while a call was CRUD: create, read, update, delete.

Then business logic moved into the layer, connectors began finishing work no call described, and the layers that had become applications met the two doors for the first time at scale. The per-call rate card outlived the transition only where a call remained one uniform act. We walk the whole arc, and the test that fell out of it, in the API pricing era; what matters here is that the decision predates every noun that followed.

Sighting Two: OEM Has Asked the Question for Decades

The decision is older than the API era. When one vendor embeds another’s product, the license grant has to answer a question we examined in channel, OEM, and white-label licensing: whose units is the grant denominated in?

Denominate the OEM fee in the supplier’s units and the host vendor is passing through: relaying a supplier’s meter to its own customers with a royalty schedule attached. Denominate the fee in the host product’s units and the component has been recast: the supplier’s meter stops at the host’s cost line, and the host prices what the composite offering does for the end customer.

Every OEM negotiation that stalled on denomination was arguing pass through or recast, in contract language predating this year’s noun. The paper is older; the decision is identical.

Sighting Three: The Harness Buys in One Unit and Sells in Another

Which brings us to the newest noun. The vendors shipping agent harnesses as products buy capacity in a supplier’s metered unit and sell capability in units of their own. That is the OEM denomination question under this year’s vocabulary, and the default answer repeats the API era’s opening position: the supplier’s token, passed through with margin, describing the vendor’s consumption while the customer bought a finished task.

Credits soften the invoice without changing the decision. We take the harness apart on its own page; the point this piece needs is narrower. The harness is not facing a new decision. It is the connector of the API era standing in front of the same two doors.

The Invariant: Metric, Then Packaging, Then Price

Whatever the layer is called, the sequence does not move. The value metric decision, which unit travels with the grant (the industry also calls it the licensing metric), comes first, because everything downstream operates on the unit it defines. Packaging comes second: editions, add-ons, and allowances only mean something once there is a unit to allocate across them.

Pricing comes last: a rate on a unit nobody chose deliberately is a rate defending nothing. The industry runs the sequence backwards at every rename: pick the pricing shape the new category has converged on, inherit the unit embedded inside it, discover the mismatch at renewal. We made the sequence argument for the SaaS vocabulary in SaaS pricing models; it holds for every vocabulary since.

The moment of recognition repeats across the patterns in our corpus, layer after layer: a vendor describes its product in the current year’s noun, and its pricing problem turns out to be a decision an earlier layer already faced under an older one.

The diagnosis conversation is shorter than most teams expect. Name the unit on your invoices out loud, say whose meter prints it, and much of the architecture work reveals itself. When the answer is uncomfortable, talk to a pricing expert before the renewal cycle makes the decision for you.

Which layer owns which decision?

The sequence assigns custody: the value metric to licensing, allowances and editions to packaging, the rate card to pricing. A dispute that will not settle is usually being argued in the wrong layer’s vocabulary.

The Test to Run on the Next New Name

The vocabulary will churn again. Whatever noun replaces harness, the test is short, and you can run it the day the noun arrives.

Name the unit the customer is billed on. Then ask whether that unit existed under the previous name. If it did, the pricing question is not new, whatever the category branding claims. You are looking at pass through or recast, a decision some earlier layer already faced under an older vocabulary, and its history is available the moment you stop calling your situation unprecedented.

One further question: does the unit describe what the customer receives, or what you consume? A consumed unit passes a cost through, however many abstractions sit between the meter and the invoice; a received unit is a recast metric, with everything recasting demands of it. Neither answer is a verdict on its own. But a company that cannot say which door it stands before has usually walked through one by default, and the default is pass-through, because the supplier’s meter was already running.

If the unit on your invoices belongs to a supplier, and your product long ago started doing more than that unit describes, you have already faced this decision. You just have not answered it yet. The answer, and the packaging and pricing architecture that follow from it, is work we do with software companies every week. Bring us the layer, in whatever vocabulary this year gave it, and we will show you which era already priced it.

FAQs