Every IT leader I speak to has an artificial intelligence story about customer experience. Almost none have a results history.
This gap is now measurable. Talkdesk Inc.in partnership with NewtonXsurveyed 252 executives at director level and above in customer experience, IT, operations and AI strategy at mid-sized and large organizations in four regions. The resulting benchmark, The State of Agent Automation in CX, concludes that IT should re-plan its next CX investment: adoption is no longer a differentiator and execution is the new battleground.
The headline numbers make it clear. 98 percent of companies use some form of AI for customer experience. Generative AI is at 74%, scripted AI is at 65%. But only 24% use agentic AI, which refers to systems that pursue a goal, reason from it and act across systems. And only 15% combine agent AI with the orchestration needed to consistently meet a customer need. Now 81% are testing or deploying AI agents, but only 19% have scaled beyond a single use case and nearly 80% have fewer than 10 AI automations in production.
Pedro Andrade, vice president of AI at Talkdesk, presented the results in an analyst briefing and outlined the state of the industry. “Adopting AI is no longer a question. Now the question is whether companies can orchestrate all the elements needed to deliver measurable customer outcomes – agents, people, data, knowledge, workflows and governance – all in one package.”
This gap comes with a double cost, as Talkdesk describes in its study announcement: companies pay for isolated AI tools while also bearing the operational costs of customer requests that those tools never fulfill. Talkdesk founder and chief executive Tiago Paiva calls the result “a false sense of progress” and argues that widespread adoption obscures how few companies can quantify the impact of AI.
The routing is not resolved
The most useful distinction in research is between activities that resemble orchestration and true orchestration. 64 percent of companies operate specialized AI agents for functions such as identity verification, billing or technical support. Yet only 35% have AI that can maintain customer context and operate across systems to drive a real solution. If every commit forces you to rebuild context and memory, you’ve built a smarter control panel, not a results engine.
“Rerouting work paths is not the same as resolving them,” Andrade said. He cited a car accident as an example: One event triggers a towing service, a claim, a repair shop and a rental car – multiple departments, multiple systems, one expected result. Fifteen percent of companies can automate this process end-to-end.
The research organizes this into a five-stage maturity curve with two chasms. The first is the agent threshold: the transition from scripted and generative AI to acting AI. 56 percent are below that. The second is CXA unlocking – the transition from isolated agents to autonomous cross-system execution. 29 percent are agentic scalers with one of the two skills; 15% are CXA leaders with both.
It is the spread between these top two levels that should be the focus of budgetary attention. CXA leaders are about four times more likely to see large CSAT or NPS gains than scalers – 22% vs. 5%. 38 percent already solve over 40 percent of customer problems autonomously, while 60 percent of scalers still solve less than 20 percent.
Executives are also about twice as likely to use revenue-focused automations like churn prediction (51% vs. 28%) and personalized recommendations (44% vs. 19%). The difference is not in the number of tools. It depends on how much meaningful work the AI is allowed and able to do.
The blockers are the infrastructure, not the intelligence
This is entirely the responsibility of IT. When asked what limits their ability to automate CX, respondents cited compliance (50%), security (48%), disparate systems (45%), outdated infrastructure (44%), inadequate capabilities (41%) and unclean or inconsistent data (40%). None of the biggest obstacles are due to AI models.
Fragmentation comes with noticeable costs. Human agents waste an average of 28% of their time switching systems, re-entering data, and searching for customer context. This proportion rises to around 35% in the least mature companies and drops to around 25% in the most mature – around 10% of employee capacity is recovered. Respondents estimate that about one in four customer interactions involve incomplete context, repeated explanations and manual escalations. And 94% of organizations do not have AI-powered knowledge management.
This knowledge gap is the one Andrade hears most often in customer conversations: “It’s not about actually having it. It’s more about how do I make it ready to be used by the AI.” Only 64% of CXA leaders have workflow-embedded or AI-powered knowledge – and they are the leaders.
The lesson generalizes. If your environment is too fragmented for humans to work efficiently, it would also be too fragmented for AI agents to work safely.
What IT Pros Should Do
- Check for solution, not deployment. Count how many issues are closed without human intervention and how many need to rebuild the context upon handoff. The number of agents deployed is a vanity metric. Ten agents who surrender are worth less than two who defect.
- Correct data and knowledge before adding agents. Only 2% of organizations report fully consistent data. Move from “some integrations” to “mostly integrated” across all systems that touch your top journeys and move knowledge from static repositories into the workflow. As Andrade noted, “operationalization” – not launching the first pilot – is the real challenge.
- Select travel, not use cases. Select two or three high-volume interdepartmental trips and then map out all the systems, approvals, and exceptions within them. This brings up compliance and legacy issues while the blast radius is still small.
- Incorporate measurement from day one. Only 5% of companies can clearly quantify the business impact of AI. Andrade called measurement the report’s biggest flaw: knowing what to track, which KPIs are important, and having the resources to track them. Link automations to first contact resolution, autonomous resolution rate, cost per contact, and CSAT before go-live, not after.
- Treat agents like employees. Almost one in five companies already view AI agents as a workforce rather than a technology, and 99% believe a hybrid workforce is beneficial – but 52% cite confidence in AI decision quality as their top concern, including 46% of the most advanced companies. The most mature companies are more than 10 times more likely to have AI and humans running in a single, unified operation. Define escalation paths, quality assurance frameworks, observability, and accountability for agents as you would for a new team.
- Skip the general roadmap. Maturity varies greatly depending on the vertical. Retail leads the way, with 24% at the Leader level for repeatable, high-volume travel. Healthcare is leading the way to market but has the most siled data and the lowest returns. Financial services have the most integrated data, yet only 9% are leaders and held back by legacy cores. “CXA roadmaps cannot be generic,” Andrade said. Design your solution according to the real limitations of your industry.
83 percent of companies assume that AI will solve more problems autonomously within two years. This will only happen if someone has done the inglorious work of connecting systems, embedding knowledge and governing autonomy. That someone is IT.
Zeus Kerravala is a Principal Analyst at ZK Research, a division of Kerravala Consulting. He wrote this article for SiliconANGLE.
Image: SiliconANGLE/Gemini
Support our mission to keep content open and free by interacting with theCUBE community. Join theCUBE Alumni Trust Networkwhere technology leaders connect, share information and create opportunities.
- Over 15 million viewers of theCUBE videosto spark conversations about AI, cloud, cybersecurity and more
- Over 11.4k theCUBE alumni — Connect with more than 11,400 technology and business leaders shaping the future through a unique, trusted network
Founded by tech visionaries John Furrier and Dave Vellante, SiliconANGLE Media has built a dynamic ecosystem of industry-leading digital media brands, reaching over 15 million elite tech professionals. Our new proprietary theCUBE AI Video Cloud is a game-changer in audience engagement, leveraging the theCUBEai.com neural network to help technology companies make data-driven decisions and stay at the forefront of the industry conversation.
