Fixed-price sprint · Four weeks · One brand, one market, end to end
AI surfaces are already recommending brands. The next step is transacting with them — and that runs on infrastructure most merchants have never been asked for: verifiable product data, callable commerce capabilities, recognizable agent traffic, and evidence that a human authorized the spend. The uncomfortable part is that failure here is silent. You are not outranked. You are excluded, with no impression data to explain it.
The five layers we assess
Can an agent verify that your product satisfies a buyer's stated constraint?
Can it resolve, quote and commit against you — outside your storefront?
Can you tell a buying agent from a scraper — and what are you blocking today?
Can you accept an agent-initiated order and defend it in a dispute?
Who owns this across merchandising, digital, security, payments and brand?
agentic commerce readiness = minimum(product knowledge, commerce capability, identity, authorization, operating model)
— the weakest layer is the ceiling, and it is almost never the one you expect.
The distinction that decides everything
Most brands are working on the first and have done nothing about the second. That is survivable right now, because agents still hand the buyer back to a browser. It stops being survivable the moment they don't.
The model knows your brand and says nice things about it. When the buyer says "order it," the agent reaches your site, cannot establish a price for this customer in this market, cannot confirm the delivery date, cannot create an order without a session — and hands the intent back to a human. The demand was real. The conversion happened somewhere else.
The agent's request is recognized at your edge. It verifies your product against the buyer's constraints from data you publish. It gets a real total with tax and a real delivery date. It creates the order with authorization evidence you retain. You remain merchant of record, you keep the customer, and you can prove afterwards what happened.
Scope · One brand, one market, five layers
One brand in one market, end to end. You pick the scope — a category, a region, a business unit. We assess it as it actually behaves toward machine traffic today, not as the roadmap describes it. Readiness varies enormously by market and category, and a group-level average hides exactly the gap you need to find.
We take the constraints buyers actually state in your category — fits model X, waterproof, under 12 kg, delivered before Friday, REACH compliant — and test whether each one is machine-checkable in your data. Not overall: by category, weighted by revenue. We assess identifier discipline, variant modelling, classification and units, claim provenance, compliance linkage, and how stale your price and availability get before anyone notices.
Output → constraint-verifiability scorecard + prioritized fix list by revenue
Every agentic commerce protocol reduces to three demands: return the authoritative record for an item, return the real total for this basket in this context, and create the order. We map which of those you can serve as callable services today, independent of the storefront checkout — and which are trapped inside a session-bound funnel. Identity linking for contract or loyalty pricing gets particular attention, because it is where B2B either works or doesn't.
Output → resolve / quote / commit gap map against UCP and ACP
The layer everyone skips and the reason agent traffic quietly fails before it ever reaches commerce. We pull the actual picture from your edge: how much agent traffic arrives, what is challenged or mitigated, whether anything is cryptographically signed, and whether your policy distinguishes a buying agent from a catalogue scraper. Most teams have never looked. Most are surprised.
Output → agent traffic & access report + per-path policy recommendation
"The buyer says they never authorized this" is a conversation you will have, and the first time will be at a bad moment. We define your posture: what evidence you require before accepting an agent-initiated order, what you retain and for how long, what merchant-side ceilings apply by value and category, and what the dispute path looks like. Reviewed with the people who will actually own it when it happens.
Output → authorization & evidence posture, agreed with risk and legal
Product data sits with merchandising. Commerce capability sits with digital. Identity sits with security. Authorization sits with payments and legal. Presentation sits with brand. In most organizations these five have no shared meeting, no shared vocabulary and no shared metric — which is why this work stalls even when everyone agrees it matters. We name the owners and the decision rights.
Output → named accountabilities + the twelve questions for your platform vendor
Deliverables
All five layers scored for your brand and market, with the ceiling identified — the one layer capping everything above it, and the shortest path to raising it.
Category by category, weighted by revenue: which buyer constraints an agent can check against your data and which it cannot, with a fix list tied to revenue rather than to effort.
Your resolve, quote and commit surface as it exists and as it needs to be — the callable core that makes any protocol an adapter instead of a project.
What is reaching you, what you are blocking, and the per-path policy — plus the authorization evidence and merchant-side limits you should require before accepting an agent order.
An evidence-backed protocol recommendation with a scoped path. "Wait" is a legitimate answer — it means not funding an integration whose constraints are already visible.
Sequenced so each quarter ships something valuable on its own, with named owners across the five functions and the vendor questions folded into your next renewal.
Delivered as a working session with the five owners in one room — merchandising, digital, security, payments and brand. For most clients that meeting is the first time those five have discussed this together, and it is often worth more than the document.
Honest scope
Data fixes are prioritized against the constraints buyers actually state, by revenue. The smallest work that makes your product verifiable — not a multi-year master data programme.
We do not build your UCP or ACP endpoint. We tell you whether you are ready for one, which to adopt first, and what it will cost you to support the second.
Whether models recommend your brand is a different question with a different answer. That is the Agentic Shelf Audit. This one starts where that one ends.
We assess what your current stack can serve and where it caps you. Sometimes the answer is a broker or an adapter, not a migration.
Scope, price, timeline
If your commercial surface turns out to be agent-ready, you will know exactly which layer is carrying it and what to protect. If it is not, you will have the shortest path to making it so — with each gap tied to a specific buyer constraint, a specific capability, and a named owner. Either way you leave with a canonical view of what you sell and what an outsider's software can do with it, which is worth having whichever protocol prevails.
| Week | What happens |
|---|---|
| Week 1 | Scope agreed. Edge and traffic data pulled. Catalog extract and constraint list assembled. Interviews with the five owners. |
| Week 2 | Constraint-verifiability testing by category. Commerce capability probing — what can actually be called, and what breaks under machine rate. |
| Week 3 | Authorization and evidence posture worked through with risk, legal and payments. Protocol options assessed against your platform's real roadmap. |
| Week 4 | Readiness map, capability design and four-quarter roadmap finalized. Working-session readout with all five owners. |
€14,500
Fixed · one brand, one market · four weeks
The methodology, in the open
The Agentic Commerce Stack — free, no email wall — maps all eight layers of the emerging machine-readable commerce layer: product feeds, UCP, ACP, MCP, A2A, AP2 and Verifiable Intent, the Trusted Agent Protocol, Web Bot Auth, x402, and the Digital Product Passport. It includes the canonical-layer architecture, the four-quarter agenda, the twelve vendor questions and a scored 48-item self-assessment.
We give the methodology away deliberately. The value was never in knowing the steps exist. It is in running them across five functions that do not currently talk to each other, and arriving at a decision someone will fund.
Free field guide · auxfirst
The Agentic Commerce Stack
From the auxfirst canon
Each one answers a single question with evidence. This one asks whether your commercial surface can be transacted with. Its siblings ask about your workflow, your agent, and your brand.
Ready when you are
One category, one region, one business unit — or the market where you already suspect AI-referred demand is arriving and not converting. Book the scoping call and bring whoever owns product data if you can; that conversation gets useful fastest.
In the message, mention "Agentic Commerce Readiness Sprint" plus the brand and market you have in mind — that's all we need to scope it.