Prepare once. Keep testing everywhere AI commerce moves.
Herm connects your existing product information, tests how it performs across important AI commerce environments, tracks changing requirements and shows what needs attention when a channel or standard changes.
You keep your feed, PIM and merchant tooling. Herm measures whether the information they already publish works when AI systems use it: read the methodology.
Your catalogue is interpreted consistently inside Product Readiness. External channel requirements are tracked around it, not treated as the source of truth for your brand.
Four statuses, and they are never merged into a single logo row.
Live, connected, tracked and roadmap mean four different things on this page. Collapsing them into one row of logos is exactly the overclaim the vocabulary exists to prevent.
Herm runs Product Readiness tests here today.
Herm reads your product data from here. Your system stays in place.
Herm follows the published requirements. Not certification or validation.
Planned. Not currently available.
Protocol readiness is one input to sellability, not the definition of it.
A technically conformant feed can pass every channel requirement and still fail the shopper it was meant to serve. Passing a channel requirement does not prove an AI agent can satisfy a customer's need.
That is why Commerce Channels reports where you are being evaluated, and Shopping Tests report what actually happens there.
“Find a waterproof trail shoe under £150 for wide feet.”
Where does this catalogue behave differently?
Stage-level results for the same catalogue and the same test set, per live environment.
Where evaluation conditions differ enough that a single number would be misleading, Herm shows the difference rather than averaging it away.
| Product Readiness stage | Herm · Claude | Herm · OpenAI | Herm · Gemini |
|---|---|---|---|
| Find | 88% | 84% | 81% |
| Answer | 61% | 58% | 55% |
| Trust | 82% | methodology confirmation required | 80% |
| Close | 93% | 91% | 90% |
| Sellable Intent Rate | 68% | methodology confirmation required | methodology confirmation required |
Kessock Outdoor sample workspace · test set v7 · 148 shopping tests per environment · GB market. Rows marked “methodology confirmation required” are not yet published as directly comparable across environments.
What is live, what is connected, what is tracked, what is planned.
Coverage changes as the ecosystem does. Each item carries its real status, and nothing on the roadmap is presented as available.
- Product feed URL
- Google Merchant Center
- PIM export
- Commerce API
Herm reads product data from these. Your systems stay in place.
- Agentic Commerce Protocol (ACP)
- Universal Commerce Protocol (UCP)
- Google Merchant Center requirements
- OpenAI commerce and merchant requirements
Tracking means Herm follows changes and requirements. It is not certification, validation or compliance.
- Shopify-specific agent environments
- Custom or merchant-owned agent profiles
- Additional market-specific commerce standards
Planned. Not currently available.
A real commerce-agent architecture, extended for independent evaluation.
Herm's primary shopping simulation environment uses and extends Anthropic's open-source Commerce Agents reference implementation. Herm adds its independent Product Readiness measurement layer.
The evaluator matters more than the agent vendor. Herm’s measurement layer is identical across every test profile, so results stay comparable as architectures change.
Anthropic is not affiliated with or endorsing Herm. Results are produced by Herm’s evaluation layer and do not represent any provider’s live consumer product. No partnership, endorsement or privileged access is claimed. Provider names are used descriptively.
Standards change. Product intent shouldn’t have to.
ACP, UCP and similar specifications are external commerce standards Herm tracks as the ecosystem evolves. Herm does not use any single external protocol as the internal source of truth for your brand.
Your product information is interpreted consistently inside Product Readiness, while relevant external channel and standard requirements are tracked around it.
| Standard | What it covers | Status | |
|---|---|---|---|
| ACP | Agentic Commerce Protocol | Agent-to-merchant commerce interactions. | Tracked |
| UCP | Universal Commerce Protocol | Cross-platform commerce interoperability. | Tracked |
| GMC | Google Merchant Center | Product feed and eligibility requirements. | Tracked |
| OpenAI | Commerce and merchant requirements | Product surfacing and merchant expectations. | Tracked |
Precise language. Herm does not certify, validate or confirm compliance with ACP, UCP or any provider specification. Where a status says tracked, it means tracked.
Know when the rules change before failures become invisible.
Environments and requirements move. When they do, Herm re-runs the affected Product Readiness tests and reports what changed for your catalogue, not just that something changed somewhere.
Tests are versioned, and every result carries the environment profile it ran against.
- 04 Aug Feed requirement updated new eligibility attribute · 4,182 SKUs re-checked 12 tests
- 11 Aug Destination policy changed market routing rules tightened for GB 14 tests
- 19 Aug Standard revision published ACP draft revision · no test impact none
- 30 Aug Re-run after update affected tests re-run · result restored restored
What changed, which tests are affected and what to review are shown. Detection and requirement-to-test mapping are proprietary.
Don’t just tell me the standard changed. Tell me what it means for my catalogue.
A changelog reports the industry. An impact chain reports your products, your affected stage and the action your team can take.
The same catalogue fails for different reasons in different places.
Herm reports the root-cause category behind each failure, per environment. Where an environment does not expose the information needed to attribute a cause, Herm says so instead of guessing.
| Root-cause category | Herm · Claude | Herm · OpenAI | Herm · Gemini |
|---|---|---|---|
| Product information incomplete | ● | ● | ● |
| Variant not distinguishable | ● | ● | — |
| Claim unsupported | ● | — | ● |
| Destination invalid | ● | ● | ● |
| Retrieval behaviour | ◌ | ◌ | ◌ |
A correct recommendation still fails if the shopper has nowhere valid to go.
Product Readiness evaluates commercial facts as supplied by the connected source at test time, and whether the resulting recommendation is valid for the shopper's market and context.
- Supplied product URL resolves
- Retailer destination is valid
- Availability as supplied at test time
- Market eligibility for the shopper
- Create a cart
- Execute checkout
- Process payment
- Verify real-time global stock
The product was selected correctly. The destination supplied for a GB shopper resolves to a US retailer.
Herm evaluates supplied destination, availability and market eligibility. No checkout, cart or payment is executed.
Sellability is market-specific.
One product, three markets, the same shopper intent. Price, availability, retailer and eligibility are read from the connected source at test time.
| Supplied by source | GB | US | EU |
|---|---|---|---|
| Price | £185.00 | $219.00 | €209.00 |
| Availability | in stock | in stock | out of stock |
| Retailer | kessock.com | kessock.com/us | partner retailer |
| Eligibility | eligible | eligible | not eligible |
Price and availability are evaluated as supplied by the connected commerce source at test time. Herm does not independently verify real-time global stock.
Product Readiness tells teams when changes materially affect AI sellability.
Alerts are tied to evidence: the environment or catalogue change, the tests it affected and the products behind them.
Herm evaluates whether your existing commerce information works when AI systems actually use it.
Herm may identify channel-specific information requirements. You keep your existing commerce infrastructure and the systems your team already runs.
Nothing here asks you to migrate your catalogue or route your commerce data through a new system of record.
- Reads the product information your systems already publish
- Tests it in live AI commerce environments
- Tracks changing external requirements around it
- Reports what needs attention, with the products behind it
- A PIM or a feed platform
- A certification or compliance authority
- A new system of record for your catalogue
- A checkout, cart or payment processor
Common questions
Does Herm certify ACP or UCP compliance?
No. Those are external specifications Herm tracks as the ecosystem evolves. Tracking a standard is not certification, validation or a compliance confirmation on anyone’s behalf. Where a status says tracked, it means tracked.
Do we have to move our catalogue to Herm?
No. You keep your feed, PIM and merchant tooling. Herm reads the product information those systems already publish and measures whether it works when AI systems use it. Nothing asks you to change your system of record.
Can results be compared across environments?
Only where the evaluation conditions justify it. Where they differ enough that a single number would mislead, Herm shows the difference and marks the row rather than averaging it away.
What does Herm do when a standard changes?
It re-runs the affected Product Readiness tests and reports what changed for your catalogue: which tests were affected, which products sit behind them and what to review. Tests are versioned and every result carries the environment profile it ran against.
Does Close mean Herm completes a purchase?
No. Close evaluates the supplied destination, availability and market eligibility. Herm does not create carts, execute checkout or process payment.
Keep your catalogue ready for wherever AI commerce goes next.
Connect a feed, run one live environment and read the result against real shopper intents.
Sample figures throughout.