Weekly signal

This week (June 22–30, 2026) accelerated a practical shift: agentic AI is moving from pilot projects into contact-center operational fabric. Vendors shipped agent-first features that treat AI agents as first-class workers (voice agents, agent assist, workforce controls, and analytics), while cloud CCaaS and contact-center vendors added AI evaluation and richer integrations that make agents easier to govern and measure in production. Key developments point to three concurrent priorities for service teams: (1) integrating agents into agent workflows and identity/governance, (2) bringing AI into real-time voice and post-contact evaluation, and (3) operationalizing human+agent workforce planning.

What changed

• Google Cloud Contact Center (CCaaS) released updates on June 22 that expand admin control and CRM adapter behavior and continue productizing agent adapters and integrations for live agent workflows — an incremental but practical step for embedding agents into enterprise stacks.

• Microsoft published Dynamics 365 customer-service updates (June 22) and pushed workforce engagement management to GA on June 30 — explicitly building Service Agent / Copilot hooks so AI agents, live agents, and supervisors share a single operating plane and governance surface. The update highlights agent identities, auditability, and scheduling/coaching flows placed alongside agentic automation.

• Zendesk’s June 24 briefing emphasized traction for AI-native voice solutions (noting large-seat deployments) and framed voice + agentic agents as the next battleground for service automation and quality. This underlines that enterprises are deploying agentic voice at scale, not just text bots.

• Amazon Connect’s June 2026 release notes added generative-AI evaluation for self-service (AI-agent) interactions and expanded AI-powered post-contact summaries and language coverage — putting evaluation and QA of AI agents into standard supervisor tooling.

• Vonage Contact Center’s June 22 release introduced Agent Knowledge Assist and supervisor/agent workflow enhancements that reduce friction when agents and AI agents collaborate.

What to do with it

1. Treat AI agents as operational workers. Add them to workforce plans, capacity models, and escalation paths now — Microsoft’s GA WFM features show vendors expect agents to be scheduled and coached like humans.

2. Instrument and audit agent decisions. Use CCaaS and contact-center release tools (Google, Amazon, Vonage) to capture transcripts, post-contact summaries, and evaluation scores so you can measure resolution, compliance, and customer experience against AI-driven actions.

3. Prioritize voice in pilot-to-scale plans. Zendesk’s deployments show voice-agent rollouts are viable at large seat counts; test voice in the smallest risky domain first (billing, status updates) and bake in human takeover and QA.

4. Update governance and identity practices. Ensure every agent has identity, logging, and a clear audit trail (actions, connectors used, data accessed) before expanding them into transactional work. Microsoft and platform release notes indicate this is now table stakes.

5. Short-term tactical: add post-contact summarization and AI-evaluation to supervisor dashboards this quarter to speed quality feedback loops; these features are available in major CCaaS releases this month.