The Specific Answer

Vol 1 argued that an authorization decision has to run at a specific place. Vol 2 is the era where specific answers get built, examined, and bought. This is the first.

Why this essay names a product

Vol 1 of this publication held a strict line. Nine essays, no vendors named except in editorial reporting on their public actions. The argument came first because the argument had to settle before any specific product could be discussed without the discussion collapsing into positioning.

The argument is settled. Vol 2 flexes that line by design. The first essay of Vol 2 names the control plane the author of this publication is building, because to do otherwise would hide the specific thing the editorial arc was pointing at the whole time. Readers deserve to see what the arc applies to. That is a different piece of writing than a sales page. A sales page lists features. An editorial examination names what the thing ships today, what it does not ship, who the honest buyer is, and where adjacent vendors sit against it.

The scene the product has to answer

Two procurement agents. Shared $10,000 budget. Each submits a legitimate $6,000 order through a different provider at the same moment. Both authentic. Both approved by the vendor's own controls. Neither looks like fraud. Treasury owes $12,000 against a $10,000 ceiling, and nothing in the current enterprise stack is built to stop it.

FederatedIQ, or FIQ, is an authorization layer built to answer that question. Every irreversible commitment an agent is about to make passes through one decision point. The decision reads the current aggregate exposure across every agent acting under the same authority, applies the policy that authority has set, and returns approve or decline before the commitment lands. Each decision produces a signed record the controller, the auditor, or the regulator can verify without trusting the agent or the provider that acted.

That is the architecture Vol 1 argued for, implemented as a specific product.

Where other vendors sit

Vol 2 names adjacent vendors by name, because any reader building an enterprise agent program needs to tell what each actually covers.

OpenAI's ChatGPT Wallet, shipped November 13, governs the per-transaction consent and biometric confirmation on the user side. It authorizes one purchase at a time. It does not compose across transactions or across agents sharing authority. That is correct for what it is. It is wrong to read as governance of the aggregate.

Sardine's Know Your Human, now integrated with ChatGPT Wallet, provides device and intent attestation so merchants and issuers can approve agentic transactions with confidence. It is fraud detection and acceptance-side risk scoring. It sits on the opposite side of the transaction from the enterprise authorizing its own agents. Different buyer, different side of the fence.

Agen.co, built by the team behind Frontegg, is an identity-native agentic governance platform doing per-action policy checks at agent runtime, with customer logos including Cisco, Palo Alto Networks, Siemens, and Rapyd. Its anchor is identity: once an agent is authenticated under a human identity, Agen.co enforces policy on what that identity's agent is allowed to do. Strong at per-agent governance. Less strong at the cross-agent aggregate that a shared budget creates, because the hinge is identity rather than federation.

FIQ sits in the slice none of those three natively covers: cross-agent, cross-provider aggregate authority against shared constraints, computed at the moment of the commitment, verifiable after. If an enterprise has only one of those three problems, the right vendor is probably one of the three above. If the enterprise has the aggregate-across-agents problem on top, that is the problem FIQ is for.

What FIQ ships today

The arc's editorial discipline has been to flag what is proven versus what is claimed. The product deserves the same treatment.

Shipping today: the per-action decision point, the signed decision record with external verifiability, the policy service that evaluates against live state rather than cached entitlement, and the out-of-process placement that keeps the agent from reaching the mechanism that governs it. These are the properties Vol 1 argued for, and they are in the product because the argument shaped the product.

Shipping with specific integrations rather than universally: cross-agent aggregate exposure read in real time. This works today where the agents and the systems of record speak protocols FIQ has adapters for. It does not work universally yet because universal composition across heterogeneous enterprise substrates is the engineering problem the Parakhin paper named in N° 032 as the open edge of the field. FIQ's bet is that the integrations harden one substrate at a time, starting with the ones where the concurrent-commitment pain is most expensive.

Not shipping, honestly: a universal cross-vendor federation that works out of the box against every enterprise's existing stack. No vendor can honestly claim that, and anyone who does is selling a demo. What ships is integration work against specific systems with specific contracts, growing the covered surface.

The one test a buyer can run

The specific test a treasurer can apply before signing a pilot is this. Two agents, shared ceiling, concurrent commitments. Does the system stop the second one before the first settles, and does it produce a signed record anyone with the public key can verify afterward.

If a vendor's demo shows that, the vendor is operating in the authorization-decision slice. If the demo shows per-transaction approvals, per-agent policy checks, device attestation, or post-hoc fraud scoring, the vendor is in one of the other slices named above. Each of those is a real product. None of them is the thing the concurrent commitment demo tests.

The one-page pitch for FIQ, with the pilot shape and the buyer question, is at fixgovernance.ai/fiq. The product itself is at federatediq.ai. Both exist so the reader who wants to pressure-test the specific answer has a place to start.

What Vol 2 does with this essay

This is the first essay of Vol 2 and the first time the publication has named the author's product in editorial voice. The voice-product boundary that Vol 1 held is flexed here deliberately. It is not dissolved. Future Vol 2 essays will continue to name adjacent vendors by name, examine what they ship, and apply the same verification discipline to every claim, including the ones made about FIQ.

Readers who read this as a sales page will not be wrong about the author's interest. They will be wrong about the editorial standard. The standard is the same as every essay in Vol 1: name the specific claim, show the evidence, flag what is not yet proven. If that discipline survives into Vol 2, the voice-product boundary flexes without breaking. If it does not, the publication is no longer worth reading. Readers will know by the end of any Vol 2 essay which way the discipline went.

The argument is general. The test is specific.

Cross-link: first full Vol 2 essay, following the capstone/opener N° 039 (After the Argument). Named vendors: OpenAI's ChatGPT Wallet (launched November 13, 2026, verified in N° 038); Sardine's Know Your Human product per their public partnership announcement with OpenAI for the ChatGPT Wallet rollout; Agen.co by Frontegg per their public site and customer logos (Cisco, Palo Alto Networks, Siemens, Rapyd, Tipalti, Flowcode, Cyera). The concurrent-commitment wedge was defined in the post-ship analysis of the Sardine and Agen.co competitive landscape and codified in the FIQ one-pager at fixgovernance.ai/fiq. The open-engineering-problem framing of cross-vendor federation at machine speed is from N° 032 (The Authorization Gap Has a Literature Now), specifically the Parakhin paper on cache-coherence mapping to multi-agent authorization revocation (arXiv:2603.09875). The verification discipline applied to FIQ's shipping claims in this essay (what ships today, what ships with specific integrations, what does not ship) was confirmed with the product author rather than asserted from marketing materials. Reciprocal: the product lives at federatediq.ai.

End N° 040 · Vol 2 · 01