This week Gartner released a number every contact center leader should pause on. In a survey of 275 service and support leaders, only 4% said their organization has achieved the ROI it expected from AI.

Four percent. After years of pilots, budgets, and board presentations.

I’m not surprised. Too many AI projects in the contact center start from the wrong question: how do we take people out of the interaction? Christopher Sladdin, Senior Director Analyst at Gartner, put it well: “Leaders must stop asking how they remove humans from interactions and instead ask where their involvement creates disproportionate value for customers.”

That is the question Zappix AI Self-Service is built to answer.

Why so few contact centers see AI ROI

As CX Dive reported, executives often expect AI to cut costs compared with human-handled interactions. Tools that help agents resolve calls faster or spend less time on post-call notes do save time. But those gains don’t necessarily turn into financial wins.

The reason is simple. If the same calls still reach live agents, the cost structure doesn’t change. You made each call a little shorter. You didn’t change which calls need a person at all.

Gartner also advises pursuing fewer AI use cases and tying each one directly to a strategic objective. For most contact centers, the objective that matters most is cost to serve. And the biggest driver of cost to serve is the number of calls that reach live agents.

Live agents and AI are not fit for the same job

Sladdin said it directly: “Humans and AI are not fit for the same job. Each can deliver value but in different ways.”

Think about what fills a live agent’s day. Checking an order status. Updating an address. Rescheduling an appointment. Making a payment. These calls matter to the customer, but they follow the same steps every time. They don’t need judgment, empathy, or negotiation. They need a clear path to done.

Now think about the calls that do need a person: a billing dispute that crosses three departments, a frustrated customer on their third attempt, a patient who is anxious about a diagnosis. That is where a live agent creates disproportionate value, to use Gartner’s words.

When repetitive calls fill the queue, agents spend the day on work AI should handle, and customers with complex needs wait longer to reach them. Nobody wins.

The fix is not to replace agents. It is to send each call to the right resource: AI self-service for the routine work, live agents for the moments that matter. That is how you lower cost to serve without lowering the quality of service.

How AI self-service takes repetitive calls off the queue

The trouble with traditional self-service is that voice alone is a poor way to finish a task. Callers listen to a menu, forget the options, misspeak a long account number, and press zero. The call lands with an agent anyway.

Even when voice AI understands the caller perfectly, some tasks are simply easier on a screen: reviewing details, choosing from a list of options, or entering information.

Zappix AI Self-Service works differently. Our multimodal AI agent pairs a voice conversation with a synchronized visual screen on the caller’s phone, in the same call, in real time. Customers hear what to do, see it on their phone, and act right there. They confirm an order, pick an appointment slot, or make a payment by tapping, not by repeating themselves to a speech engine.

The visual screen is what lets a caller resolve the reason they called. That is the difference between a call that is deflected on paper and one that is actually resolved.

Zappix deflects more than 75% of calls from live agents to visual self-service. Those are the repetitive, process-driven calls that don’t need a person. The calls that do still reach one, and the agent who picks up has the time and attention to handle them well.

What visual self-service looks like in practice

Hibbett Sports shows what happens when a contact center moves digital customers from voice IVR to visual self-service. With Zappix:

• Call containment reached 80%+.
• Average handle time is 3+ minutes shorter than with voice IVR.
• Customers rated ease of use with an NPS of 57.

That combination is what moves ROI. Fewer calls reach agents, the calls handled in self-service go faster, and customers like the experience. Lower cost to serve and a better customer experience, at the same time.

Getting the foundations right without an AI team

Gartner found that AI-forward organizations invest nearly 1.8 times more in foundations, like clean, current data and the knowledge sources AI relies on, than in the AI use cases themselves. Driving value also still takes people: to decide which use cases come first, and to train staff on how and when to use the tools.

Sladdin also said, “AI’s superpower is that it enables leadership to learn from customer interactions at scale.” Every self-service session shows where callers finish the task and where they get stuck. That is the data you need to keep improving.

That is a lot to ask of a contact center team that is already stretched. It’s why we deliver Zappix turnkey: we design, we implement, we launch, and we continuously improve. Clients don’t need an internal AI team to get results. For Genesys customers, Zappix runs on Genesys Cloud as the preferred AI Visual IVR partner of Genesys. Enterprise security and compliance are built in from day one.

The takeaway

The 4% figure is not a verdict on AI. It’s a verdict on how AI gets applied. Point AI at the repetitive calls it is built to handle, give customers a visual way to finish the task, and let live agents focus on the conversations where they make the difference.

See. Hear. Resolve.

If you’d like to see Zappix AI Self-Service on a live call, let’s set up a demo.

Sources: Matt Vartabedian, “Only 4% of customer service teams have achieved expected ROI from AI. Here’s what needs to change.” (CX Dive, Nov. 4, 2026): https://www.customerexperiencedive.com/news/only-4-of-customer-service-teams-have-achieved-expected-roi-from-ai-here/873732/