Why the best AI strategy in customer support is a human elevation strategy

New research from MIT and Oak Ridge National Laboratory quantifies something CX leaders have felt intuitively for years. AI's real value in customer support isn't replacing agents. It's clearing the path for your best people to do their best work.

The Iceberg Index study maps which skills are most exposed to AI automation and reveals a counterintuitive truth. The tasks AI absorbs most easily are the ones that were slowing down your highest performers all along.

What the research actually found

The research introduces a "skills exposure" lens that changes how leaders should think about AI investment. Instead of measuring impact by headcount or deflection rates, the study identifies specific skill clusters and how exposed they are to automation.

High-exposure skills cluster around repetitive cognitive work. Classification. Summarization. Workflow preparation. Knowledge retrieval. Record updates. These are the tasks that happen before and after nearly every customer interaction, and they collectively represent the largest source of operational drag inside support teams.

Low-exposure skills are different. They cluster around empathy, creativity, synthesis, judgment, and complex problem solving. These are the moments where AI consistently struggles and where the best support agents thrive.

The research also highlights an important framework called EPOCH from MIT Sloan's Workforce Intelligence research. Empathy. Presence. Opinion and judgment. Creativity. Hope and leadership. These capabilities directly correlate with employment growth as AI scales. They represent the human differentiators that remain central to long-term customer trust.

The implication is clear. AI should absorb the high-exposure cognitive load while human expertise is concentrated where it drives the greatest impact.

Two paths forward for support teams

The research points to two distinct opportunities for how support organizations can evolve alongside AI.

Path one. Elevate your top performers

Your most effective agents are the ones who excel at creative problem solving, emotional intelligence, and turning difficult moments into loyalty. They are also the ones most likely to be dragged down by routine cognitive tasks that fragment their focus.

When AI handles classification, tagging, knowledge surfacing, workflow preparation, and conversation summaries, these high performers get something valuable back. Time and attention. They can focus entirely on the complex, emotionally nuanced issues that actually require their unique capabilities.

This is not about giving agents fewer responsibilities. It's about giving them the right responsibilities. The research validates what the best CX leaders already practice. Freeing up your top talent from operational friction is not a luxury. It's a productivity strategy.

Path two. Transition team members into AI development roles

The second opportunity involves the portion of your team currently handling more routine work. As AI absorbs high-exposure tasks, these team members can transition into roles that shape, coach, and improve AI quality over time.

Think of it like an innovations team. These individuals review AI outputs, identify errors, flag edge cases, write new knowledge content, and help the system get smarter. They bring frontline context that engineers don't have. They know how customers actually phrase things. They know what a good answer looks like in practice.

This is a redeployment strategy. And it mirrors exactly what the MIT research recommends. Human involvement in AI design, oversight, and continuous improvement.

What this means for how we measure success

Traditional KPIs still matter. That includes handle time, first contact resolution, and customer satisfaction. But they don’t capture the deeper shift happening inside support teams when AI is deployed well.

Deflection rate, in particular, misses the point entirely. It measures volume avoided, not value created, and it reinforces an outdated either/or mindset that pits AI against human service instead of treating them as complementary forces.

The research points to a better approach, one rooted in accelerating customer loyalty through a layered, skills-based lens. Measure which skill clusters are moving from humans to AI. Track whether high-exposure work is being absorbed. And monitor whether your top performers are spending more of their time on the low-exposure, high-judgment conversations that actually deepen customer relationships.

This is the skills-exposure framework the study introduces, and it predicts where work is shifting far more accurately than staffing ratios or volume metrics ever could.

Organizations that understand this shift will gain a real advantage. They won’t just automate faster, they’ll redesign their workforce more intentionally and use AI in service of their customers, not in spite of them.

The operational model that emerges

When you combine these insights, a clearer picture of modern support operations emerges.

AI handles high-volume cognitive load at scale. It resolves routine issues, surfaces relevant knowledge, prepares workflows, and eliminates the operational friction that fragments agent attention.

A dedicated team coaches and develops the AI. They read one-star reviews. They identify gaps. They write new content. They ensure the system reflects the brand voice and handles edge cases correctly. They treat AI like a new team member that requires ongoing investment.

Specialist agents focus on the relational moments that build trust. They own the complex issues. They repair emotional breaks in the customer experience. They turn frustration into loyalty. These are the interactions customers actually remember.

This is not automation replacing humans. It's a redesign of how human potential gets deployed.

The strategic takeaway

The MIT and Oak Ridge research gives CX leaders evidence for something many have believed intuitively. AI's greatest impact in support comes from removing cognitive load so humans can lean deeper into creativity, empathy, and complex problem-solving.

The organizations that treat AI as a human elevation strategy will outperform those that treat it as a cost reduction tactic. They will retain their best talent. They will build stronger customer relationships. And they will create a support model that adapts as AI capabilities continue to evolve.

Research now backs what the best CX leaders already knew. The future of support is AI and humans working as a system, not AI working instead of humans.

See how Gladly brings AI and human agents together here.