Only 4% of Service Leaders Hit Their AI ROI, and Some Are Rehiring Agents

Customer service was supposed to be the place where AI paid for itself first. The interactions are high in volume, the work is repetitive, and the cost per contact is easy to measure. Two years into heavy investment, the function has a scorecard, and it is sobering. New Gartner research presented at its Customer Service & Support Conference in Denver last week finds that almost none of the organizations spending on AI have seen the return they planned for, and the workforce savings that were meant to fund the bet have mostly not arrived.

A 4 Percent Success Rate

The headline figure is stark. According to Gartner's November 4 release, just 4 percent of customer service organizations have achieved the ROI they expected from their AI initiatives. The finding comes from a survey of 275 service and support leaders conducted in July through August 2026, and it describes a function that is buying enthusiastically but struggling to turn that spending into value.

"While customer service leaders are eager to accelerate their use of AI, many are failing to unlock its value," said Christopher Sladdin, Senior Director Analyst in the Gartner Customer Service & Support Practice, during the conference's opening keynote. His prescription was not more AI but different AI: apply it more strategically, rethink how it is funded and deployed, and redesign teams and processes so they can actually capture what it delivers.

The gap is not a lack of activity. Practitioners are moving quickly into more complex automation. A NiCE survey of more than 118 customer service practitioners at over 70 organizations, conducted from September 22 to October 10, found that approximately six in ten respondents are already running a multi-agent AI ecosystem. More than half cited oversight, auditability and human-in-the-loop controls as what lets them expand automation with confidence. Deployment is accelerating. Returns are not.

The Headcount Savings Never Came

Much of the original business case for AI in service rested on smaller teams. That is not what the data shows. A companion Gartner release on November 5, drawn from the same survey of 275 leaders, found that only 21 percent said their organizations had reduced human agent headcount because of AI. Among those that did, the average reduction was 16.7 percent of the workforce over the previous 18 months.

Most leaders are reshaping the workforce indirectly instead. Forty-five percent report a hiring freeze and 39 percent are shifting agents into other business units, while only 20 percent are using forced reductions. At the same time, 71 percent are moving agents into different roles within service and support, and more than half have hired specialized staff to support AI initiatives. "Service leaders are largely using AI to reshape work and increase capacity, not replace their workforce at scale," Gartner concluded.

Some of the organizations that did cut are already reversing course. Seven percent of leaders said they had rehired human workers after previous reductions, and 11 percent plan to rehire within the next 18 months. Gartner VP Analyst Kathy Ross called it the "cut fast, rehire later" trap: reducing headcount based on expected AI value rather than demonstrated results. When the technology cannot handle the anticipated volume or complexity, service quality suffers and rebuilding capacity means recruiting, onboarding and training all over again, while knowledge lost in the cuts has to be restored.

What the Top 4 Percent Do Differently

Gartner's research points to three habits that separate AI-forward service organizations from the rest. The first is aiming AI at customer value, not just cost reduction. These organizations treat interactions as strategic assets and use AI to analyze conversations and feedback at scale, finding recurring friction and the drivers of successful outcomes. "AI's superpower is that it enables leadership to learn from customer interactions at scale," Sladdin said.

The second is building portfolios rather than collecting pilots. Gartner predicts that by 2028, organizations that build AI portfolios will achieve twice the business value of those pursuing standalone use cases. Sladdin urged leaders to deploy "fewer, higher-value and interconnected use cases" and to strengthen the data and knowledge management foundations that let them work together. AI-forward organizations invest nearly 1.8 times more in those foundations than in the use cases themselves.

The third is redesigning teams around what humans do best. "Humans and AI are not fit for the same job," Sladdin said, arguing that leaders should stop asking how to remove people from interactions and ask instead where their involvement creates disproportionate value. Gartner's Day 2 conference highlights push the point further: Director Analyst Andy West said that by 2029 AI will automate 70 percent of current customer service tasks, but total task volumes will increase as AI creates new opportunities for value-added services. Human work, in his words, is evolving rather than disappearing.

Demand is changing shape too. Senior Director Analyst Daniel O'Sullivan noted that customers are already three times more likely to consult a third-party generative AI tool than a company-provided chatbot when they need help. As customers' own AI lowers the effort of chasing a refund or a resolution, requests that once died as "not worth the hassle" will reach the service desk.

A practical playbook for operations leaders:

• Define the objective before the use case: Decide whether AI is meant to cut cost, retain revenue, strengthen loyalty or support growth, because each goal calls for different workforce and investment decisions.
• Fund the foundations first: Follow the AI-forward pattern of investing more in data integration and knowledge management than in individual AI tools, so each new use case builds on the last.
• Prune the pilot list: Retire standalone experiments that do not connect to a strategic objective and concentrate budget on a smaller, linked portfolio of use cases.
• Make headcount an outcome, not an input: Reduce staff only after AI has proven it can deliver the required outcome at an acceptable cost under real operating conditions, not on the strength of a vendor forecast.
• Mine interactions for root causes: Use AI to analyze conversations at scale and eliminate the recurring issues that generate contacts in the first place.
• Redesign roles around judgment: Move agents toward complex, emotionally charged work and hire for resilience, empathy and adaptability, the traits Gartner says are hardest to teach.

The 4 percent figure is less an indictment of AI than of how it has been bought: one use case at a time, justified by headcount that was never going to disappear on schedule. The service organizations pulling ahead are spending on the plumbing, connecting their use cases and redesigning work around people. For everyone else, the next budget cycle is a chance to stop adding pilots and start building a system.