Documentation

Overview

Improvado’s AI Agent can now assist you directly inside the Transformation Layer — across Wizard Flows, Blend Data steps, filters, and formulas.

It sees what you see:

• Current steps
• Mappings between sources
• Applied filters and formulas
• Output schema and preview state

This means you can ask the Agent to help with almost anything:

• Suggest or create custom formulas
• Modify existing transformation logic
• Apply or adjust filters
• Identify missing steps
• Explain model logic in plain language

Use Cases

Blend Data

Blend Data can involve dozens of fields across multiple platforms — and the AI Agent is here to help.

• Autocomplete output schema based on inputs
• Match fields from each source to the schema
• Spot missing or incorrectly mapped fields
• Suggest standardizations across channel names, currencies, or KPIs

Examples:

• “Map all campaign_id fields across Google and Facebook to the output schema.”
• “Add missing creative fields from Facebook.”

{%docs-informer info%}

AI Agent doesn’t change anything without your permission. It is up to you to accept changes or not.

{%docs-informer-end%}

Formula Helper

The Agent is especially powerful in Custom Formula, where it can help non-technical users write complex calculations using plain English.

• Ask for a new column based on logic
• Request nested conditions or multi-source calculations
• Automatically build CASE statements or IF chains
• Preview the result before applying it

Examples:

• “Add a column for ROAS = revenue / cost.”
• “What does this formula do?”
• “Convert CPM to CPC using impressions and clicks.”

Formula Explainer

Not sure what a formula does? Just ask the Agent.

• It can break down any expression into plain English.
• Explains each function, operator, and field used — so even non-technical users understand the logic.
• Perfect for debugging legacy models or onboarding new teammates.

Example:

• “What does CASE WHEN cost > 0 THEN revenue / cost ELSE NULL END mean?”

Filter Logic

AI Agent enhances your filtering experience by letting you apply and debug filters faster and more intuitively — across both dataset and table levels.

• Ask the Agent to apply simple or compound filters using natural language
• Get help filtering by campaign name, ad type, spend, or date ranges
• Let the Agent debug why a filter is returning empty results
• Ask for a breakdown of which rows were excluded

Examples:

• “Filter out campaigns where spend = 0.”
• “Only include countries US and UK.”
• “Why is this model returning nulls after filter step?”

Join Data

The AI Agent is fully integrated with the Join Data operator, providing intelligent assistance as you merge datasets.

• Autofill Join Keys: The Agent can suggest join keys based on data types, column names, and historical patterns.
• Explain Join Logic: Unsure which type of join to use? Ask the Agent to explain Left vs Inner Join in context.
• Detect Conflicts: It can identify mismatches between field types and warn you if a join may produce duplicates or nulls.
• Field Selection Support: Get help deciding which fields to keep from each table to avoid bloated outputs.

Examples:

• “Help me join Facebook Ads and GA4 data using campaign name and date.”
• “Why is this join returning too many rows?”
• “Which metrics should I keep from the secondary table?”

The Agent acts like a smart assistant embedded in your modeling flow — reducing trial-and-error and helping you ship clean joins faster.

Group Data

Group Data can be complex — especially when dealing with multiple levels of hierarchy, duplicate logic, or advanced aggregations. That’s where the AI Agent steps in.

• Suggest Grouping Keys: Based on your goal, the Agent can recommend which fields to group by.
• Write Aggregation Logic: Whether it’s SUM, COUNT DISTINCT, or custom ratios, the Agent can generate the formulas for you.
• Debug Unexpected Results: Ask why a grouped output is missing values or producing incorrect totals.
• Explain Hierarchical Structures: If you’re working with multi-level data (e.g., brand > product > SKU), the Agent can help you structure the grouping correctly.

Examples:

• “Group by campaign_id and get total spend.”
• “Roll up ROAS by platform.”
• “Why are my totals higher after grouping?”

Conclusion

The AI Agent isn’t just a support tool — it’s a partner in building and maintaining your most complex models. From schema mapping to formula creation, it lets you move faster, stay consistent, and reduce dependency on manual QA.

Whether you’re building from scratch or improving an existing flow, the Agent brings structure, speed, and insight — right inside your transformation environment.