Zoho announced DataPrep 3.0 on October 28, a release that lets a user state a data goal in plain language and have an agent plan, build and run the pipeline.
What happened
In the Zoho announcement, the agentic version of Ask Zia plans the pipeline, and the user can review and edit each step. Nothing reaches the destination until the user approves it, and Ask Zia remembers preferences across conversations. Import, Transform and Export agents handle the stages, and a Data Quality agent suggests fixes. Zoho says a rebuilt engine called dForge gives 100x faster previews, 70% faster transforms and 50% less storage. Code Studio lets users write custom Python transforms and test them on sample data before a full run, next to more than 400 built-in transforms. A headless DataPrep exposes an MCP server so other AI agents can create and run pipelines. New export connectors include HubSpot, Klaviyo, Brevo and Mailchimp. A 15-day trial needs no credit card, and a launch event was set for October 29.
Why it matters
Zoho’s own example for the sales side has Ask Zia sending enriched leads to the Zoho CRM Leads module after the user confirms. That is the step between a list import and the CRM where duplicates, bad fields and routing mistakes get in. The approval gate before the destination is the control to look at first. See our recent reports on SAP’s lead routing agent and Zoho’s WhatsApp billing change.
One original insight
In our reading, a RevOps team trialling the Data Quality agent should keep a fixed sample of 200 lead rows, run the agent over them, and compare the suggested fixes with what an analyst would have changed by hand. The 15-day trial is long enough for that test, and the result says more about fit than the speed figures Zoho reports for its engine.
Source: Zoho Blog, Zoho DataPrep 3.0
