A LEED plaque compresses years of invisible work into one glance.
Energy models. Commissioning files. Materials decisions. Systems reviews. Most people entering the building will inspect none of them. They see a plaque, recognize the signal, and understand that a defined body of work sat behind it.
The plaque is not the work. It is what makes the work legible.
AI assurance has the same visibility problem. A professional-services firm may register its AI uses, restrict client data, test outputs, preserve evidence, assign accountable humans, and monitor changes. The buyer cannot see any of that from a proposal. The provider cannot turn it into an intelligible commercial distinction by pointing at a policy PDF.
That is why AI assurance needs two products.
→ Buyer permission: a bounded, defensible basis for saying yes to a defined use of AI.
→ Provider status: buyer-recognized standing earned by doing the difficult work required to support that permission.
The first creates utility. The second creates the incentive to keep investing in the first.
Status is a market-readable result
Eugene Wei's 2019 essay “Status as a Service” separates utility from social capital. His subject was the social network. His sharper point travels.
Status does not appear from an icon. It comes from a proof of work that other people recognize and cannot cheaply copy.
On Instagram, the proof may be the photograph. At Harvard, the proof is admission. In LEED, the proof sits across prerequisites, credits, documentation, review, and a completed project. Different games. Same mechanic.
side note: importing Wei's frame does not turn assurance into a popularity contest. It tells us why a signal with no demanding work beneath it becomes decorative almost instantly.
For AI assurance, the proof of work is operational:
- Name the exact service boundary.
- Register the material AI uses inside it.
- State what data, tools, people, and jurisdictions are covered.
- Connect each consequential claim to current evidence.
- Preserve client restrictions and human decision rights.
- Test what the system does in practice.
- Report material changes and lose positive status when the claim stops being true.
That work is expensive in attention, authority, and operating discipline. It should be. If a firm can earn the same signal by filling out a form and paying a fee, the signal tells the buyer very little.
The buyer has to receive something useful
Provider status is not the primary job.
The buyer is deciding whether a firm may use AI in important client work. That decision needs an object: one legal entity, one service boundary, named uses, current evidence, explicit exclusions, and a status that can change.
The answer is never “our AI is responsible.” It is closer to this:
This named service uses AI for these registered purposes, under these client restrictions, with these tools and data classes, during this period, subject to these exclusions and this current review status.
That sentence is less glamorous than a seal.
Good. It can support a decision.
A buyer may still say no. A client may add a restriction. Counsel may require more proof. Certification could never replace those decisions. Its job would be to reduce the amount of basic truth each buyer has to reconstruct from scratch.
Status gives providers a reason to do the hard part
Risk reduction is necessary. It is rarely an attractive internal rallying cry.
Ask a provider CEO to fund months of evidence work, workflow mapping, testing, training, and recurring review. “This may reduce an undefined future risk” is a thin commercial story. The cost is immediate. The reward is abstract.
Visible, buyer-recognized standing changes the equation. The same work can become a market signal, a procurement asset, a client-confidence object, and a talent signal.
LEED shows the shape of that system. USGBC's history records a 19-project pilot in 1998, a public launch in 2000, later versions, professional credentials, and an independent certification function. Its current rating-system page combines prerequisites, optional credits, third-party review, certification levels, education, credentials, a public directory, and promotion tools.
Research has found evidence of strategic differentiation inside those visible levels. A 2020 Management Science study found that builders chose LEED levels partly in response to the levels achieved by earlier nearby projects. That supports a status mechanism. It does not prove status caused LEED's full adoption.
The distinction matters.
Two loops, one integrity rule
The proposed AI Service Assurance system contains two reinforcing loops.
The utility loop
Clear boundary → current evidence → better buyer comprehension → more usable permission decision → clearer demand for proof
The status loop
Difficult operating work → visible bounded standing → provider recognition → more providers pursue the work → the signal becomes easier to recognize
The utility loop has to govern the status loop.
Reverse that order and the system decays into badge theater. Providers chase the symbol. Buyers infer more than the claim supports. Weak operators copy the image. The market learns to ignore the whole thing.
No proof. No permission.
Current readiness work is not certification
Petrichor's current AI Assurance method builds claim-to-evidence controls, service boundaries, current operating records, and client-facing permission material.
Petrichor's current AI Assurance work is readiness and implementation, not certification. The certification, registry, and mark described here are a proposed market architecture that still requires independent governance, validation, and legal review.
A legitimate certification institution would need an independent standards owner, assessor rules, decision independence, surveillance, complaints, appeals, a live registry, lawful mark ownership, and tested suspension behavior. None can be conjured by designing a handsome badge.
The status object comes last.
Proof. Permission. Recognition.
In that order.
Credible status must be able to turn off
A mark has value when the market knows what was earned, what is covered, and whether the claim remains current.
That produces the next question:
What happens when the facts change?
Read why a credible assurance badge must be able to turn off.
Frequently asked
Is provider status the same as reputation?
No. Reputation is diffuse. The status proposed here is a bounded signal attached to a defined service, standard version, evidence period, and current registry state. It matters only when buyers recognize the underlying proof.
Does buyer permission mean automatic approval?
No. The proposed system supplies evidence for a permission decision. The buyer keeps authority, client-specific restrictions, and the right to refuse the use.
Why would a provider pursue this work?
The provider may gain a reusable client-confidence object and market distinction from work it already needs to perform. That is the hypothesis. It still needs buyer and provider testing.
Is Petrichor issuing certifications now?
No. Petrichor currently performs readiness and implementation work. Independent certification, a public registry, and a mark remain proposed and gated.