Just 4% of customer service leaders have gotten the payoff they expected from AI. Here’s how to beat the odds

On Wednesday, Gartner, the business and technology research firm, opened its customer service conference in Denver with a sobering number: just 4% of customer service organizations have achieved the return they expected from their AI projects. It comes from a survey of 275 customer service and support leaders in July and August. The conference wraps up today.

What Gartner found

In the opening keynote, Gartner analyst Christopher Sladdin said the average organization spends 12% of its customer service budget on AI, “yet service leaders report positive returns only a quarter of the time.” His diagnosis: “If we keep pursuing AI primarily for cost reduction, the math simply doesn’t work.”

Gartner says the most successful organizations, which it calls “AI-forward,” share three habits:

• They use AI to create value for customers, not just to cut costs, including studying calls and messages to see where customers get stuck.
• They choose fewer, connected uses, and spend nearly 1.8 times as much on data and knowledge management, the foundation AI works from, as on the AI projects themselves.
• They redesign the work. “Humans and AI are not fit for the same job,” Sladdin said, urging leaders to ask where a person adds the most value instead of how to remove people.

Another Gartner analyst, Daniel O’Sullivan, put it bluntly: “AI alone is not a strategy.”

The 4% reflects leaders’ own expectations, and the full report is for Gartner clients.

Why it matters for your business

Most small businesses don’t have a customer service department, but the trap is the same: buy an AI tool to save money, switch it on, and wonder months later why nothing got better.

Customers notice. In a separate Gartner survey of 3,566 customers earlier this year, 87% said it’s essential for companies using generative AI for service to offer a way to reach a person.

Closer to home

Picture a Dothan heating and air company that puts AI on the phones to save the front desk time, but the AI can’t see the schedule, so callers still wait for a callback. Or an Enterprise veterinary clinic whose chat assistant quotes last year’s prices, or an insurance agency anywhere in the Wiregrass whose chatbot keeps a customer with storm damage going in circles. Each bought AI to cut costs, and none gave it what it needed to help customers.

Five ways to make AI customer service pay off

1. Aim at what customers want, not only at what you can cut. Start with a request customers wait on, like booking a visit, and measure what changes for them: appointments booked, questions answered the first time, calls no longer missed.
2. Get your information in order first. AI answers only as well as what it reads. Put current prices, hours, service areas, and policies in one place, and make someone responsible for keeping it current.
3. Use the simplest tool that works. O’Sullivan’s rule for AI agents that act on their own: “If a problem can be solved without agentic AI, it probably should be.” Sometimes a booking link or a clearer FAQ page does the job. When AI is the right tool, connect it to your calendar so it can finish the task, not just take a message. Here’s what an AI phone agent really costs.
4. Keep a person within reach. Let customers ask for someone at any point, and send upset customers and big jobs straight to your team, with the conversation attached.
5. Treat launch day as the start. O’Sullivan called go-live “only the first 10% of the journey.” Each week, read a sample of conversations, note what the AI keeps missing, and fix the cause, whether that’s its instructions, your website, or a policy nobody wrote down.

None of this means AI can’t help your customers. It means the payoff comes from the work around the tool.

Where we can help

Our AI agents service starts with one conversation or task that happens often and costs real time, and an agreement on what a great outcome looks like. The agent works from your approved content, connects only to the systems it needs, and has to handle your real questions and tricky cases before a customer ever sees it. After launch, uncertain cases go to a person, and we review the results with you and keep improving. Book a free strategy call.

Sources

• Gartner: Gartner Survey Reveals Just 4% of Customer Service Leaders Have Achieved the Expected ROI of Their AI Initiatives, November 4, 2026
• Gartner: Gartner Customer Service & Support Conference: Day 1 Highlights, November 4, 2026
• Gartner: Gartner Survey Finds 87% of Customers Say Companies Using GenAI for Customer Service Must Provide Access to a Human Agent, August 4, 2026