While 70% of CX practitioners say their organization has adopted AI in CX, only a fraction are seeing a return on investment, Forethought AI Agents by Zendesk’s Antoine Nasr said.
Dive Brief:
- AI adoption in CX is accelerating, but few organizations are seeing ROIof their investments, according to a survey of more than 600 CX leaders and practitioners by Forethought AI Agents by Zendesk.
- Only 2% of organizations that have adopted AI in CX are seeing actual value in the form of improving outcomes and ROI, according to Antoine Nasr, head of AI at Forethought AI Agents by Zendesk. “You see the vast majority of people, they have AI deployed, but they’re not really seeing the value,” Nasr told CX Dive.
- Last year, 57% of CX practitioners said their organization had adopted AI. In 2026, that figure has risen to 70%.
Dive Insight:
AI adoption has permeated across CX. The majority of CX leaders — 54% — say their organizations are using purpose-built CX AI operating models, while about one-quarter say they use a help-desk add-on, and 22% say they built their own in-house AI operating model.
However, not all adoption has been successful. In fact, much of it has yet to provide a return on investment.
Three-quarters of enterprises rolled backa customer-facing AI agentafter deploying it, a Sinch survey from earlier this year found. Governance failures were the top culprit, followed by customer data exposure, hallucination or brand risk.
“Taking shortcuts with AI cannot repair a fragmented customer journey, inconsistent knowledge, poor handoffs or a lack of ownership,”Julie Geller, principal research director of Info-Tech Research Group, said in an email. “In fact, it can make that dysfunction faster, less visible and more difficult to unwind.”
Much of the problem stems from a top-down initiative to adopt AI in customer experience.
“It makes sense to your leadership, and it turns out that it’s quite easy to get tools that generate tokens,” Nasr told CX Dive. “It’s a lot harder to get tools that generate business value with those tokens.”
Companies need to start with a problem in need of a solution — not with a general mandate to add AI
“Companies should start by identifying where customers and employees face the greatest friction, then ask whether AI is the right tool to remove it,” Geller said. “Start with a customer problem, not an AI use case. Choose a workflow with enough volume to matter, where the current failure points are understood and where ‘better’ can be clearly defined before anything is deployed. That definition should be tied to a meaningful business KPI.”
Customers and businesses approach AI with high expectations, too.
“Those people have probably tried ChatGPT and they’ve had the magic moment with all those tools, and it’s felt very easy,” Nasr said. “An organization doesn’t have the same ability to just get on ChatGPT and get that magic moment — it’s just not how it works. There’s a lot of context that is in all sorts of different places. The facts on the ground change, and so you need systems that can evolve with how your business evolves.”
For Nasr, success won’t come from meeting management’s request to simply deploy AI. It’s a matter of effort and thoughtfulness.
“The first and foremost ingredient, in my opinion, is the investment of time and effort into setting it up,” Nasr said.
“If you think of a business owner 10 years ago, they weren’t just kind of accepting the first template that they can find on WordPress for their website. The website is the business entity,” Nasr said. “The AI agent — that is the frontline of your customer experience today.”
