Thyris Merchant Services as an AI Merchandising Platform
Thyris Merchant Services combines catalog operations, AI-assisted enrichment, provider controls, usage reporting, classification tags, and hyperpersonalization campaigns. This allows merchandising teams to improve product data and prepare targeted experiences without making every AI output immediately customer-visible.
When to use this scenario
Use this model when a merchant needs to:
- Improve incomplete product titles, descriptions, images, or tags.
- Test an AI provider or prompt on a controlled product set.
- Apply repeatable enrichment rules to a catalog segment.
- Track token, image, provider, status, and estimated cost history.
- Build product campaigns from reviewed classification tags.
- Keep draft, review, activation, pause, and expiry decisions explicit.
Platform composition
| Capability | Merchant Services role |
|---|---|
| Catalog | Provides source product data and stores approved results. |
| Instant Enrichment | Tests selected enrichment outputs on a controlled set. |
| Enrichment Rules | Defines reusable or scheduled processing policies. |
| AI Providers | Uses Thyris Managed AI or a configured compatible provider. |
| Usage | Reports execution, provider, token, image, cost, status, and error history. |
| Hyperpersonalization | Builds product campaigns from product and enrichment signals. |
Reference flow
Implementation journey
1. Define the quality problem
Identify the exact product fields and catalog segment that need improvement. Define acceptance criteria for accuracy, tone, completeness, brand constraints, prohibited claims, and image suitability.
2. Select and configure the provider
Use Thyris Managed AI or configure a supported provider. Set ownership for credentials, model choice, cost review, and provider changes.
3. Run a controlled pilot
Use Instant Enrichment on representative products. Include normal items, sparse records, edge cases, regulated categories, and products with unusual variants.
4. Review before scaling
Compare source and generated content. Check factual accuracy, unsupported claims, duplicated language, classification quality, image consistency, and expected usage cost.
5. Create a reusable rule
Once the pilot is accepted, create a rule for the approved catalog segment and outputs. Keep execution history and pause the rule when source data or provider behavior changes.
6. Prepare campaigns
Use reviewed product and enrichment tags to create a campaign. Validate product availability, discount values, message text, status, and activation window before publishing.
Review responsibilities
| Team | Review focus |
|---|---|
| Merchandising | Product accuracy, classification, and campaign relevance. |
| Brand and legal | Claims, tone, regulated content, and required disclosures. |
| Catalog operations | Source quality, identifiers, availability, and synchronization. |
| Platform engineering | Provider credentials, execution reliability, and cost controls. |
| Commerce operations | Campaign lifecycle, inventory changes, and rollback. |
Success measures
- Catalog completeness before and after enrichment
- Acceptance rate of generated content
- Manual correction rate
- Cost per accepted product update
- Enrichment failure and retry rate
- Campaign activation, pause, and expiry accuracy
- Product availability conflicts detected before activation
Production controls
- Preserve the source value or another rollback path.
- Do not publish unreviewed claims in regulated or high-risk categories.
- Separate AI generation from campaign activation.
- Pause campaigns when price, inventory, or policy changes invalidate them.
- Review provider usage and error history regularly.
- Test changed prompts or models on a pilot segment before broad execution.