The unit economics arrived quietly, then all at once. A human agent costs about $7 to $12 per call. A voice AI can resolve the same interaction for roughly $0.40. That gap, approaching 95% on a per-call basis, is no longer a projection. It is the operating reality reshaping every large contact center this year.

Microsoft put a dollar figure on it that other CFOs will find hard to ignore. Microsoft’s chief commercial officer, Judson Althoff, said internally that AI delivered more than $500 million in savings over the prior year in Microsoft’s call center. The comment circulated publicly in July 2025, days after Microsoft announced cuts affecting about 9,000 roles. That kind of documented outcome changes every renewal conversation a business process outsourcer will have through 2027.

The companies supplying the infrastructure are raising capital at a pace that reflects the demand. Berlin-based Parloa raised $350 million in Series D funding at a $3 billion valuation, just eight months after closing a $120 million round at a $1 billion valuation, with General Catalyst leading and customers including Allianz, Booking.com, and SAP. PolyAI raised $86 million in December 2025 at a $750 million valuation, making it one of the best-funded voice AI startups in the contact center space. The investor thesis is straightforward: contact center outsourcing alone is commonly estimated at more than $100 billion a year globally, with labor as the dominant cost line.

What has changed in 2026 is not the promise. It is the fluency. Modern voice AI can pause more naturally, handle interruptions, and modulate style cues such as politeness and speaking pace. The rigid menu trees that defined a decade of customer frustration are functionally obsolete. Enterprise teams report Tier-1 automation and containment in the 40% to 60% range for the call types they target first, and those rates translate directly into avoided labor spend.

The workforce consequences are real but more nuanced than the headlines suggest. What is actually happening is a restructuring of the contact center workforce: Tier-1 headcount on routine inquiries is shrinking through attrition in many organizations, while Tier-2 and Tier-3 staffing for complex escalations becomes more valuable. A separate Gartner finding adds a meaningful caveat: by 2027, 50% of companies that attributed customer service headcount reductions to AI will rehire staff, because enterprises are discovering that removing humans entirely degrades customer experience for the interactions that require judgment, empathy, and creative problem-solving.

The smarter frame is cost per resolution, not cost per call. A Forrester Total Economic Impact study found that PolyAI customers achieved 391% ROI over three years, with about $10.3 million in agent labor cost savings for a composite enterprise model. That figure reflects a specific profile, large enterprises running millions of calls annually, but the structural logic applies at smaller scale too. Any company still staffing a full Tier-1 team for order-status checks and balance inquiries is effectively paying a premium to do work that voice AI already handles reliably.

The question CFOs are now asking is not whether voice AI saves money. For many routine call types, the economics are now broadly understood. The question is sequencing: which call types, which volume thresholds, and which escalation protocols preserve the customer relationships that justify the brand in the first place.