ZoomInfo Adds AI Agent Teams to Run Multi-Step Sales Workflows

ZoomInfo launched Agent Teams on October 1, 2026, giving sales and marketing teams a product designed to carry work from a customer signal through research and outreach. It is available now in GTM Studio without a new contract or separate product SKU. The company’s promise: coordinated AI agents that handle multi-step sales work without someone supervising every handoff.

The launch draws directly on DoubleO.ai, whose orchestration architecture—the software that coordinates agents’ work—powers Agent Teams. ZoomInfo says it announced that company’s acquisition on September 30. Rather than presenting the acquired technology as a separate tool, ZoomInfo has put it inside its existing GTM Studio product. GTM means go-to-market: the work sales, marketing and customer-success teams do to find, engage and retain buyers.

ZoomInfo describes an architecture that evaluates the current account record as work progresses. Each specialized agent has its own role, instructions and tools, but receives context from earlier steps. The next action is chosen using the record’s current state, rather than simply following a fixed sequence. That design is meant to connect a company’s systems and incoming signals with ZoomInfo’s business intelligence, keeping research, decisions and follow-up inside one coordinated workflow.

Chief Product Officer Dominik Facher frames the problem as lost reasoning between agents: exchanging a finished output is not the same as preserving why an earlier decision was made. ZoomInfo says sharing that context lets agents build on one another’s work. CEO Henry Schuck links the approach to sales processes that change as people switch jobs or buyers return after a pause.

Revenue doesn’t follow a flowchart. Champions change jobs, deals stall, and buyers come back months later.

Henry Schuck, founder and CEO of ZoomInfo

ZoomInfo’s launch examples go beyond generating a reply or creating a task. They describe agents examining the history of a customer relationship, identifying new prospects and preparing or sending tailored messages. These are company-described workflows, not independently measured demonstrations.

• Follow a former customer advocate to a new job. If that person becomes chief revenue officer at a target account, Agent Teams checks whether the new company is already a customer and fits the seller’s ideal customer profile. It reconstructs the advocate’s role in the earlier deal, identifies the new buying committee and drafts outreach based on that relationship. A later visit to the pricing page can restart the workflow with its earlier context intact, rather than treating the visit as an isolated signal.
• Use a won deal to find the next prospects. Agent Teams reviews call transcripts, emails and customer relationship management records to understand why a customer bought. It then drafts a case study, identifies 10 competitors of that customer that are not already customers, and reaches out with messages backed by the case study. The example connects analysis of a completed sale with new outreach, rather than ending when the research or case-study draft is finished.

ZoomInfo says Agent Teams runs on its existing data and governance foundation, including privacy-rights management, SOC 2 controls, and systems certified to ISO 27001 and ISO 27701. Those are the security and privacy assurances attached to the launch; the company separately describes the agents’ ability to execute complex workflows reliably.

The release includes an endorsement from Caroline Burnett, sales operations leader at Silverfort. She says Agent Teams helped scale revenue operations and accelerate go-to-market work in both quality and outcomes. Her statement supplies a customer’s assessment, but gives no numerical result, comparison period or measured reduction in human supervision.

The launch release provides no benchmark or error rate for its reliability claims. It also does not specify whether outreach requires human approval before sending. That leaves an important operational boundary unresolved: the examples show work moving from account research toward customer contact, while the announcement does not explain where a person must authorize the next action.