For customer experience leaders staring down another budget cycle, the instinct to reduce headcount is understandable. But simply making fewer people do the same amount of work is not the answer. Granted, agent salaries account for close to two-thirds of a typical U.S. contact center's operating expenditure, with a further 11 percent spent on management, IT, and ancillary staff, and that concentration makes labor the obvious lever to pull.
But the operational reality tells a different story: Average speed to answer and call abandonment rates are climbing, not falling, and the industry has seen agent salaries rise by an average of 5 percent annually between 2019 and 2025, compared to just 0.5 percent a year in the preceding five years. Despite that investment, attrition has not improved; it averaged 31 percent a year between 2019 and 2025, up from 29 percent in 2014-2018.
~2/3 of contact center opex is spent on agent salaries. | $12.3 billion is the annual U.S. industry cost of manual customer authentication. | $14,000 is the estimated cost to replace one experienced agent. |
Cutting staff without changing how the operation runs would only worsen queue times and abandonment, both of which already damage revenue and experience. The more productive question for senior leaders isn't how many agents can we remove but how much more each interaction can accomplish.
Start With First-Contact Resolution
Customer satisfaction is the single most influential metric contact centers track, and first-contact resolution (FCR) consistently ranks among the top three drivers of it. Yet FCR has quietly declined across the industry, even as calls grow more complex. Every unresolved call becomes a second call, and a second cost.
AI-enabled interaction analytics can reverse this by mining historical data for the issues and solutions that have worked before, surfacing the most likely fix to an agent during the very first call. Root-cause analysis goes a layer deeper, identifying the recurring problems that generate repeat contacts in the first place, so the underlying business process gets fixed rather than just the symptom. Intelligent routing complements this by matching callers to the right agents based on skill, intent, and regulatory certification, cutting the transfers and callbacks that come from mismatched routing.
Proactive contact extends the same logic outward, using purchase history and interaction data to reach customers by email or SMS before a problem becomes a phone call, and using sentiment analysis to intercept frustration before it escalates into an inbound complaint.
Where the Call Time Actually Goes
To find savings within the call itself, it helps to know how a typical call's time is spent: roughly 10 percent on identity verification, 72 percent on talk time, and 18 percent on post-call wrap-up.
Authentication is a clear target. A mean 73 percent of inbound U.S. contact center calls require identity verification, consuming an average 42 seconds—10.2 percent of call length at a cost of 73.4 cents per call. Currently, 84 percent of that verification is done manually by agents. The industry-wide bill comes to an estimated $12.3 billion a year, for a step that adds nothing to the customer experience. Replacing agent-led verification with an AI-enabled voicebot preserves the same security process while freeing agent time and can flag unusual stress or anxiety for further review, providing a fraud-detection layer that manual authentication doesn't have.
Talk time itself holds further opportunity. The average agent works across 3.5 applications during a call and another 2.3 afterward; only 3 percent of U.S. contact centers give agents a single unified desktop, while 40 percent require juggling four or more systems. AI-enabled desktop automation can surface the right information and populate the right fields without agents hunting across screens, cutting both handle time and the error rates that themselves generate the extra putting-things-right calls that contact centers are trying to avoid.
A subtler cost hides in simple repetition. In a survey of 1,000 U.S. customers, 53 percent said they very often or fairly often had trouble hearing an agent or were asked to repeat themselves, rising to 56 percent among the youngest respondents. ContactBabel estimates the industry-wide cost of this repetition at more than $2.5 billion annually for U.S. contact centers, translating to more than $250,000 a year for a typical 250-seat operation. AI-driven voice isolation, stripping background noise from both sides of the conversation in real time, addresses this line item directly.
Post-call wrap-up—the 18 percent of call time spent on notes, disposition coding, and triggering back-office processes—is arguably the easiest win of all. It's internally facing, so the risk profile is low, and generative-AI call summaries, checked and adjusted by the agent, apply to nearly every interaction handled. That combination of low risk and universal applicability is exactly what senior leaders should look for when prioritizing an AI business case.
Digital and Self-Service: The Next Frontier
Digital channels like email, web chat, social, and messaging now account for roughly 28 percent of inbound interactions, at a meaningfully lower cost than phone: a mean $7.20 per phone interaction vs. $4.91 for email, $5.84 for web chat, and $5.50 for social media. Yet 45 percent of emails are still answered from a blank page, and only 3 percent use AI assistance. Shifting the email channel toward AI-drafted, agent-checked responses (and eventually toward AI answering without agent involvement) is one of the more immediately actionable cost levers in the report.
Self-service offers the deepest cost reduction of all. A web interaction runs to a few cents, vs. 25 cents to 75 cents for IVR, but it's underperforming its potential. Around 25 percent of inbound calls come from customers who tried and failed at web self-service first, and 30 percent of voice self-service attempts are abandoned in favor of a live agent. The fix is to make chatbots and voicebots sophisticated enough to handle genuine intent rather than rigid keyword matching, with a seamless, visible escalation path to a human when confidence is low.
The Hidden Line Item: Attrition
Replacing an experienced agent costs an estimated $14,000. For a 500-seat operation running a 41 percent attrition rate, that adds up to nearly $3 million a year, with agent absence costing roughly the same again. AI can't fix engagement on its own, but predictive analytics that track performance shifts, absenteeism, schedule mismatches, and disengagement signals give management the early warning needed to intervene before an agent walks.
None of these use cases are directly focused on replacing agents with AI. They require redirecting AI at the moments, like authentication, wrap-up, repetition, routine email, and failed self-service, that add cost without adding value, allowing agents to focus on customer loyalty, empathy, and solving difficult issues without being pressured to close calls early.
Steve Morrell is managing director of ContactBabel.