Are AI agents fixing customer experience, or are they becoming the latest layer in a stack that already asks too much of agents, customers, and IT teams?
That is the question facing many enterprises as customer relationship management (CRM) platforms, contact center systems, ticketing tools, and AI applications converge. For Vasili Triant, CEO of UJET, the issue is that many companies are applying AI to the wrong part of the customer experience operation.
As Triant told CX Today in an interview, the industry’s current focus on automating frontline interactions risks misdiagnosing the real issue.
“The industry is experimenting primarily with voice and chat agents at the front end trying to automate on this premise that the actual human or the the actual contact center agent is the problem. Do I believe that we have it wrong? Yes, I one hundred percent believe we have it wrong.”
As enterprises race to deploy virtual agents, agent assist tools, and AI copilots, many are still operating on a software foundation that has grown increasingly hard to manage. AI can intensify friction if businesses simply attach new intelligence to old workflows, Triant said, citing Salesforce’s recent partnership with Anthropic to embed Salesforce in Claude.
“Right now, everything that’s happening is ‘let’s bolt an AI solution onto some stack.’ So Claudeforce is ‘let’s bolt Claude onto Salesforce’ in some aspect.”
The risk is that products positioned as productivity enhancers become another interface, another workflow and another operational dependency. For contact center agents already navigating multiple screens and systems, that can turn AI into a burden rather than a breakthrough.
Triant traces much of this complexity back to the rise of software as a service (SaaS) buying patterns, where new tools are easy to procure and difficult to unwind.
“The challenge that’s happening within customer experience, within the SaaS world in general, is this complexity has been built up over the last 26, 27 years.”
Triant compared the problem to the shift from cable bundles to streaming services: each individual subscription seems manageable, but the combined environment becomes expensive, fragmented and hard to control.
“It’s like, ‘let’s buy another thing and another thing’ to keep adding other features on. And next thing you know, there’s this massively complex stack that you and I have to deal with as business users. And the customers are just getting the outcomes from that and complaining that this is a frustrating process.”
The impact on customers is familiar. They hear agents say a screen is slow, an application needs to load, or another system has to be checked. For agents, the complexity becomes part of the job and for customers, it becomes the experience.
Triant believes this points to a deeper question: whether the CRM and ticketing layer should remain at the center of service operations.
“Ticketing platforms, CRM platforms… have created this complexity. And they started in a world where the idea of data centralization and data warehouse didn’t exist.”
Many enterprises now have a more mature data strategy than they did when those systems became dominant. Customer data platforms, data lakes and data warehouses can provide a different foundation for customer context if contact center systems are built to use them effectively.
Triant went further, taking the position in the ongoing debate around the future of CRM that some parts of the legacy stack—including CRM platforms—may well disappear.
“I fully believe that in the future we will remove ticketing a hundred percent. There’s not many people that like their ticketing platform. It’s a necessary evil and it’s become this center of all these integrations and so people feel like you have to have it. But the reality is you don’t need it anymore.”
The alternative is intelligence applied enterprise-wide. “AI has to deal with a lot more than contact center tools. And that is one of the problems in the current construct of AI and customer experience,” Triant said.
Rather than confining AI to a virtual agent at the front door or a narrow agent-assist use case, Triant argued it should work across enterprise systems and surface the right customer context at the right moment.
“A key thing is you wanna keep driving that single data lake, data warehouse, because then the AI has one place to go to deal with information about the customer and the history and the context and bring what’s relevant up to the agent, not the entire three-page record.”
Central to the CRM complexity debate is the goal of removing the need for human agents to search across disconnected applications during live conversations, rather than drowning them in more information. It also reframes the role of human agents. Triant does not see them disappearing.
“I think the human becomes part of the orchestration layer, but they are the connection for that emotive context between the end customer and the business,” which may become more important as customers push back against poorly designed automation. Triant is already hearing examples of that fatigue from enterprises.
“Customers, if they hear a virtual agent, they just hang up. They don’t even want to deal with it.”
Virtual agents still have a role for simple tasks. Password resets, order updates and routine requests can be automated effectively. “But the idea that a voice virtual agent to replace humans altogether literally divorces you from that emotive context and divorces that customer relationship,” Triant warned.
The strongest return on investment may come from removing software, not people. “There is a model where you can do all of this and actually still save money, but you have to now start removing pieces of that complexity and that saves you money,” Triant suggested.
The issue for customer experience leaders is whether the stack underneath the agentic AI tools they choose can support simpler, faster and more human customer experiences.
As Triant puts it, the companies that get this right will give customers a more personal relationship with the brand, solve questions the first time and reduce operational costs. The ones that get it wrong may discover that AI has not solved CRM complexity and has made the problem harder to ignore.
