No Jitter Brief:

A coordinating layer above the AI agents is required to manage the sprawl, Opus wrote, naming its proposed framework the “AI agent control plane.” This five-layer plane is comprised of journey and intent state, identity and consent, policy and guardrails, knowledge governance and evaluation, audit and continuous testing. Each layer handles a different aspect of managing agent sprawl and coordinating autonomous AI agent action across systems.

Under the Opus framework, the journey and intent layer would provide AI agents with the context and progress of where they’re at across all interactions and channels. The policy and guardrails layer would enforce business policies, compliance and guardrails, while the evaluation and testing layer would validate behavior and detect issues. “The AI agent control plane…assumes interconnection across all five layers…so they can inform one another rather than operate as isolated tools. That whole system doesn’t yet exist as an actual thing brands can simply go out and buy,” Opus said.

Related:Who owns the customer truth in an AI-driven contact center?

No Jitter Insight:

Organizations must make the individual pieces that are available work together — and vendors are ready to meet that need. For example, CCaaS vendors like NiCE and Genesys have incorporated agentic AI capabilities into their platforms while providing a framework for controlling those agents. System of record vendors like Salesforce and Zendesk have moved into the CCaaS space while also introducing AI agent control capabilities. ServiceNow approaches the same issue from the workflow automation space with its AI Control Tower.

Many other vendors provide other pieces of the agentic puzzle, from orchestration and runtime platforms to context and memory engines to purpose-built evaluation and testing.

The strength of these and other vendors “is depth inside their own ecosystem,” Opus said. “Inside their walls, they can deliver something close to a control plane experience.”

Brands evaluating agentic AI vendors “should look closely at how openly each one supports cross-vendor agent coordination, how willingly it exposes journey state and policy to systems outside its own, and how seriously it engages with emerging interoperability standards rather than treating them as a threat,” Opus said.

Related:How AI is reshaping UCaaS and CCaaS pricing