Recommendation Agent: Item filters for AI Recommendations, in plain language

Subtitles available in English, Spanish, Brazilian Portuguese

TL;DR

Recommendation Agent turns natural-language descriptions into IQL item filters for AI Recommendations and explains existing ones. It can build filters that incorporate customer context like favorite brands, viewed categories, or preferred sizes, adapting recommendations to who is browsing. It works inside Synerise AI Hub, validates referenced parameters against the live item catalog, suggests corrections for typos or missing fields, and ships a plain-language summary with every filter. Available now in private preview, enabled per workspace via your CSI manager or a Service Desk request.

How close to plain language can filter logic for recommendations get?

Close enough that describing it is the work. We've introduced Recommendation Agent in Synerise AI Hub, where it runs on the same behavioral AI infrastructure as Shopping Assistant and Automation Agent. The agent reads your live item catalog and customer context, generating filters that reference real parameters and adapt to customer preferences like favorite brands or preferred sizes.

What this shifts in practice

The team closest to the campaign owns the filter logic. A marketer types "When a customer is in their favorite category, promote their favorite brand. When they're somewhere else, stick to the category they're currently viewing". The filter is built, no handoff needed.

Filter iteration speed matches campaign tempo. A failed A/B test no longer means rebuilding an IQL query. Describe the change in language and apply it. The bottleneck moves from query syntax to merchandising judgment.

Filter logic stays auditable as the team changes. Every filter the agent generates ships with a plain-language summary. Existing IQL filters can be explained on demand. Campaign handovers stop depending on whoever wrote the original query.

What is Recommendation Agent

Recommendation Agent is an AI agent in Synerise AI Hub that operates across the AI Recommendations module. Today, the agent generates and explains IQL item filters: users describe filter conditions in plain language, and the agent produces the matching expression. IQL is Synerise's query language for item-level filters on the product catalog, referencing attributes like color, category, price, or availability. The agent reads the live item parameter schema, so it validates references against real catalog fields and flags typos with suggested corrections. The output is the IQL filter, plus a plain-language summary.

The agent's planned scope across AI Recommendations also includes creating and editing recommendation campaigns, configuring A/B tests, querying performance statistics, and answering questions about recommendations for specific customers or products.

Key capabilities

Natural language filter creation

Type the condition the way you would say it out loud. "Color should be red or black." "Products from shoes and accessories, brand X or Y, price between $100 and $200, in stock." The agent produces the matching IQL filter expression. No need to know IQL syntax or look up parameter names. Input works in many supported languages.

Parameter-aware validation and suggestions

The agent reads the live item parameter schema for the workspace it is running in. If a user references a parameter that does not exist, or misspells one that does, the agent flags it and suggests alternatives from the catalog. This prevents invalid filters and reduces debugging time.

Filter summary generation

Every filter the agent produces comes with a plain-language summary describing what the filter does, which parameters it references, and what conditions it applies. The IQL expression is still there for anyone who wants to read it. The summary is what makes the filter configuration transparent and easier to review.

Existing filter explanation

Recommendation Agent can explain filters that already exist in a recommendation campaign. When a user opens an existing filter, Recommendation Agent can describe how the filter was constructed, what logic it follows, and what item attributes it takes into account. Useful when someone else built the filter, or when the original author is not around to ask.

Authors

Release note contributors: Małgorzata Wojtowicz (Product Marketing Manager), Aleksandra Wyszkowska (Product Evangelist), Michał Pastuszak (Head of Product), Kamil Gaczoł (Senior Product Manager)

Find out more

See how Recommendation Agent fits into the Synerise Agent family. Each agent specializes in one domain of the platform, with several already live and more on the way.

Synerise Agent →

FAQ

How does Recommendation Agent fit into the broader Synerise Agent family?

Recommendation Agent is one of the live agents in the Synerise Agent family, alongside Shopping Assistant, Synonym Agent and Automation Agent. The pattern across the family is consistent: the user describes intent in plain language, and the agent handles the syntax. Today, Recommendation Agent works inside AI Recommendations to build and explain IQL item filters.

Will Recommendation Agent replace IQL or the Visual Builder?

No. Both stay. Recommendation Agent is an additional way in. It suits users who prefer natural language, and filters complex enough that hand-writing IQL slows the team down. The Visual Builder remains available for simpler conditions, and IQL queries can still be written directly.

Where in Synerise is Recommendation Agent activated?

In AI Hub > AI Recommendations > Models. Open a campaign, navigate to Items Filters, and click the Synerise Agent button in the filter dialog. No separate installation, no configuration step.

Does using Recommendation Agent change how filters are stored or reused?

No. The filter the agent generates is regular IQL, stored alongside any other filter in the campaign. It can be edited in the Visual Builder, modified directly in IQL, or re-fed back to the agent for a new revision. Existing filters made the old way work exactly the same as before.

Is filter creation Recommendation Agent's full scope?

No. Filter creation and explanation are what's live today in private preview. Recommendation Agent is a domain agent for AI Recommendations, and its planned capabilities span campaign creation and editing, A/B test configuration, performance statistics queries, and answering recommendation queries for specific customers or products.

Key facts