A New Kind of Workforce
In this guide on AI workforce management, workforce management in contact centers has always been a complex optimization problem: forecast demand, schedule enough agents to meet service levels, manage adherence to schedules, handle the inevitable absences and fluctuations, and do it all while controlling labor costs. WFM tools from vendors like NICE, Verint, and Aspect have refined this discipline over decades, using sophisticated forecasting algorithms, scheduling optimization engines, and real-time adherence monitoring to squeeze maximum efficiency from human agent pools. But the introduction of AI agents into this equation changes the optimization problem fundamentally, because AI agents do not behave like human agents in any of the ways that traditional WFM was designed to manage. They do not need schedules, do not take breaks, do not call in sick, and do not have skills that degrade over time. They also do not handle every interaction type equally well, do not provide the empathy that certain situations require, and cannot make the judgment calls that complex cases demand.
The result is a new workforce management challenge: optimizing a hybrid team of human and AI agents, each with different capabilities, cost structures, and operational characteristics. This is not simply a matter of adding AI capacity to the existing WFM framework – it requires rethinking fundamental assumptions about how work is distributed, how capacity is planned, and how performance is measured. The AI handles a variable share of total interactions depending on the mix of call types, time of day, and AI confidence levels. The remaining interactions flow to human agents, but the volume and complexity of those remaining interactions change as the AI’s capabilities evolve. Human agents who previously handled a mix of routine and complex calls now handle primarily complex calls, which changes their skill requirements, training needs, and emotional labor. The entire workforce dynamic shifts in ways that traditional WFM tools were not designed to anticipate.
Planning for a Hybrid Operation
Capacity planning in a hybrid human-AI operation requires modeling two interdependent systems rather than one. The AI’s capacity is theoretically unlimited – it can handle any number of concurrent calls – but its effective capacity is bounded by containment rate: the percentage of calls it can actually resolve. If the AI’s containment rate is 75%, then 25% of all calls will still require human agents. But this 25% is not a random sample of all calls – it is disproportionately complex, emotionally charged, or unusual, which means human agents handling escalated calls need higher skill levels and more time per interaction than they did when handling a mix of routine and complex calls. Planning human agent capacity based on the overall call volume reduction without accounting for the increased complexity per call is a common mistake that leads to understaffing of the human tier.
Assembled, a workforce management platform based in San Francisco, has built its product specifically for this hybrid model, managing the scheduling and performance of human agents, AI agents, and BPO (Business Process Outsourcing) providers within a unified framework. Its approach recognizes that AI agents, human agents, and outsourced agents are not interchangeable resources but complementary capabilities that need to be orchestrated together. The platform uses ML-powered forecasting to predict not just total call volume but the likely distribution between AI-handled and human-handled interactions, and it generates schedules that ensure adequate human coverage for the expected escalation volume. For enterprises already using NICE or Five9 for workforce management, these platforms are adding hybrid workforce capabilities as well, though the integration of AI agent capacity into their forecasting models is still evolving.
Quality Management Across Humans and AI
Quality management in a hybrid operation requires different approaches for human and AI interactions, but a unified standard for the customer experience that both must deliver. For AI interactions, quality management is primarily about monitoring containment rates, accuracy of information provided, appropriateness of escalation decisions, and customer satisfaction scores. The advantage of AI quality management is that every interaction can be evaluated automatically, compared to the 1-5% sample rate that is practical for human agent call reviews. For human interactions, traditional quality management approaches – call monitoring, scoring rubrics, coaching sessions – remain essential, but the nature of the calls being scored changes. When human agents handle only escalated and complex calls, the scoring criteria should reflect this: evaluating de-escalation skills, problem-solving creativity, exception handling, and emotional intelligence rather than the routine call-handling skills that dominated quality scorecards when agents handled a mix of everything.
The organizational culture challenge of hybrid workforce management should not be underestimated. Human agents who see AI handling an increasing share of calls may feel threatened, devalued, or uncertain about their future. Clear communication about the role of human agents in the hybrid model – emphasizing that they are handling the most important, complex, and valuable interactions rather than the routine work that AI has taken over – is essential for maintaining morale and performance. Some organizations have reframed the role entirely, positioning human agents as “AI supervisors” or “escalation specialists” with higher status and compensation reflecting their advanced responsibilities. Others have created new roles focused on AI optimization – reviewing AI performance, improving knowledge bases, and designing conversation flows – that leverage agents’ deep customer service expertise in a new context. The organizations that manage this cultural transition thoughtfully will retain their best people and build a hybrid operation that genuinely delivers better customer experiences than either humans or AI could achieve alone.
