AI Voice Agents, WhatsApp in the Agent Desktop, and AI Triage at the Front Door

Home Depot Brings AI Voice Agents to the Front Door of Support

The Home Depot this week announced AI-powered voice agents with Google Cloud and Gemini Enterprise, replacing IVR menus with conversational intent detection and moving call navigation, routing, and first-response resolution from agents to AI. That matters because voice is the hardest service channel to automate, and this puts AI at the front door of support, not just behind the scenes.

The pattern is moving in the same direction, but the evidence is still case-based. Klarna said genAI handled up to 75% of support interactions before bringing more humans back into complex and premium cases; NAGA Group said AI fully resolved about 66% of chat support and plans to extend automation to email. Other firms have cut human oversight after deployment: Uber reduced about 10% of its Community Operations workforce, Microsoft and Commonwealth Bank of Australia reportedly trimmed Tier-1 support roles, and Commonwealth Bank later reversed course after its AI voice bot struggled with call volumes. ServiceNow has taken a different approach, keeping staff in the loop through dashboards and escalation review.

For service leaders, this is the next step beyond supervised copilots: governing end-to-end flows across voice, chat, routing, and QA. The career edge now belongs to people who can tune escalation rules, read containment and resolution metrics, and define where human judgment stays mandatory.

United Telecoms Brings WhatsApp Into the Same Service Desktop

United Telecoms this week added direct WhatsApp integration to its contact-centre platform through Meta’s WhatsApp Business Platform, letting agents handle WhatsApp in the same desktop as voice, live chat, and SMS with one customer record behind each conversation. That is the clearest sign yet that WhatsApp is being treated not as a separate inbox, but as a native support channel inside the case-handling environment where history, follow-up, and handoffs already live.

The shift extends the unified-context model from a shared customer view into channel absorption. The point is not just that agents can see more of the customer in one place; it is that a high-volume messaging channel is being pulled into the standard workflow so continuity survives channel changes. United Telecoms did not publish WhatsApp-specific response-time or productivity data, so the evidence is architectural, not performance-based. But the direction is clear: support stacks are being built so WhatsApp, like voice or chat, feeds the same service record and can connect onward to CRM and helpdesk systems such as Salesforce, HubSpot, Zendesk, and Freshdesk.

For support professionals, the progression is from managing channels to managing cases across channels without losing context. Teams will need agents who can read cross-channel history, maintain asynchronous conversations, and execute follow-up inside one integrated workflow instead of bouncing between tools.

OKI Puts AI Triage at the Contact-Center Front Door

OKI’s CTstage AI Concierge shows the next step in the sequence: conversational intent detection, initial inquiry triage, guided self-resolution, live-agent fallback, and response support that summarizes calls and logs interaction history. That is a narrower, more defensible use case than full autonomous answering, and the source does not claim end-to-end replies from scratch or quantify deflection or average handling time gains.

Read after Peresoft’s governed retrieval into self-service actions, the pattern is tightening around the front door of service operations. AI is now being inserted to sort, route, and summarize high-volume work before a human takes over, rather than only surfacing knowledge or executing contained workflows. OKI’s target sectors—finance, retail, government, transportation, manufacturing, logistics, healthcare, and defense—are exactly where labor pressure and traceability requirements make this easier to deploy. For support leaders, the progression is clear: the near-term job is not replacing agents, but redesigning intake so AI handles the repetitive first pass, preserves auditability, and frees staff for the cases that actually need judgment.