Is Your Platform Agent-Ready? A Practical Audit for the Machine Customer

Part three of the series. By Saurabh Sachdeva, Founder & CEO, Assurex.

In Part one I argued that AI visibility gets you found, not chosen. In Part two I argued that “chosen” is no longer the end of the story, because the AI now completes the purchase.

Since then, the most common question I’ve been asked is a practical one: how do we know if we’re ready?

This piece is the answer. It’s an audit any digital or commerce team can run against its own estate in an afternoon, built on the three layers from Part two: readable, transactable, chosen. Each layer has a handful of checks. Score each one honestly, and you’ll know where you stand before your competitors do.

What changed since Part two

The infrastructure stopped being a preview. In the last few months:

• The protocols became a stack, not a standards war. Industry coverage now describes merchants facing several protocols at once, each doing a different job: discovery, checkout, payments, agent identity. Waiting for one to “win” is the mistake.
• Google’s commerce protocol went from launch to rollout. It added multi-item carts, live catalogue access and identity linking, and its governing council expanded to include Amazon, Meta, Microsoft, Salesforce and Stripe.
• The card networks built agent identity into payments. Visa launched a single integration that accepts agent-initiated payments across four protocols. Mastercard has started scoring how likely a transaction is to have come from an agent.
• Retailers split into two camps. Shopify opened checkout to authorised browser agents this month, while Amazon has been blocking some third-party shopping agents.

For UK brands, the timing matters. Native AI checkout launched in the US first, and the UK is on the roadmap rather than fully live. That’s not a reason to wait. It’s the window to get ready while competitors are still reading about it.

Layer 1: Readable

Can a machine understand exactly what you sell, without guessing?

An agent doesn’t see your hero image or feel your brand. It reads data. If that data is incomplete, inconsistent or locked inside page layouts, you don’t exist to it.

1. Structured product data is complete. Every product carries price, availability, specifications, delivery terms and returns policy as structured fields, not just prose on a page.
2. Your feeds match your site. The price and stock in your merchant feed agree with what a customer sees at checkout. Agents notice mismatches, and they treat them as risk.
3. Content lives in your platform, not your templates. If key facts only exist inside a page design, an agent can’t reliably extract them. A headless or API-first content model fixes this at the source.
4. Terms are explicit. Shipping costs, delivery windows and return rules are stated as facts, not hidden behind “see T&Cs”.

Most enterprise teams pass check one on paper and fail it in practice. The data exists, but in three systems that disagree.

Layer 2: Transactable

Once an agent has decided, can it actually complete the purchase?

This is where most brands discover their checkout was built for a person with a mouse.

1. Checkout is exposed through APIs. Cart, pricing, shipping and order creation can be called programmatically, not only clicked through a browser flow.
2. You can accept agent-initiated payments. Ask your payment provider which agentic protocols and network programmes it supports today, and what’s on its roadmap.
3. Your fraud rules won’t block legitimate agents. Agents don’t behave like humans: no mouse movement, no browsing, instant checkout. Tuned for bots, your fraud stack may decline your best new channel.
4. Customer identity carries through. A customer’s loyalty status, saved preferences and negotiated pricing should survive an agent placing the order on their behalf.
5. You stay merchant of record. Know who owns the customer, the order data and the refund when the transaction starts somewhere else.

The good news: this layer is engineering, and engineering has a plan. It’s the most fixable layer of the three.

Layer 3: Chosen

When an agent compares you with a competitor, does the evidence favour you?

This is the layer from Part one, and the one no integration can buy. Readable and transactable get you into the shortlist. Chosen decides whether you win it.

1. You know which customer you’re the best choice for. Not “everyone”. A defined customer with a defined need, where your offer is genuinely strongest.
2. Independent evidence backs that up. Reviews, ratings, comparisons and third-party coverage say the same thing about you that your website does.
3. You deliver what you promise, and it shows. Delivery times, returns and service outcomes are where agents look for proof. Operations is now marketing.
4. You test how AI describes you. Ask the major assistants to recommend a product in your category for three different customer types. Note who they pick, and why.

If you fail here, no amount of work on the first two layers will save you. You’ll simply be a very easy brand for an agent to rule out.

Score yourself

Score each check 0 (not in place), 1 (partly) or 2 (fully). Thirteen checks, 26 points.

20 or more: you’re ahead of most of your market. 10 to 19: the gaps are fixable, but the clock is running. Under 10: start this quarter, beginning with Layer 1.

One pattern to watch for: a high score on the first two layers and a low score on the third. That brand will be easy for agents to buy from, and just as easy for them to pass over.

Where to start: the first 90 days

1. Weeks 1–2: run the audit. Get commerce, engineering, payments and marketing in one room. Score honestly. Disagreement between teams is itself a finding.
2. Weeks 3–6: fix your data. Close the gaps in Layer 1 first. Everything else depends on it, and it improves human search and AI visibility along the way.
3. Weeks 7–10: map your checkout. Document which parts of the purchase flow can be called by API today, and talk to your payment provider about agentic support.
4. Weeks 11–13: test the third layer. Run the AI recommendation test for your core customer types, and compare what the assistants say with what you believe about your own positioning.

Agentic commerce rewards the brands that are easy to understand, easy to buy from and genuinely the right choice. The first two are an engineering project. The third has always been the job.

If you’d like a second pair of eyes on your score, Assurex runs this audit with commerce and digital teams. Get in touch.