AI agent management: The new CX role you can’t ignore

AI agents are changing the work of customer service leaders and customer experience practitioners. We've all felt it coming for a while, but now the job boards – and surveys – are starting to validate and quantify the trend. 

Research fielded by Twilio in Q2 concluded that by the end of 2027, as many as 63 percent of customer service interactions in Europe, the Middle East and Africa (EMEA region) could be handled by AI, including AI agents. 

Twilio surveyed 7,652 consumers and 660 business leaders globally and also found that among consumers, 64 percent feel undervalued when their AI-driven service experience falls short. 

The broader application of AI will remove many repetitive tasks from the daily workloads of human agents – but it will also change the work of their team leaders, who now find themselves tasked with managing the AI agents. As the 64 percent stat demonstrates, getting this right is critical to customer retention. 

Service isn't the only function experiencing a transition to hybrid workforce management. CX is feeling it, too. And across the business world – from software developers to accountants and even some engineering disciplines – reliance on AI agents is building. And everybody who uses AI in their work must ensure human oversight of the outputs.

In fact, Microsoft CEO Satya Nadella said in early 2025 that "every employee will soon become a manager… of AI teammates".

Today, that prediction is visible on CX job boards. In recent months organizations including edtech platform Preply, online recruitment platform Bullhorn, consultancy Accenture, and, of course, Amazon have all advertised for service and CX jobs with AI management titles and requirements. 

Others – including Coinbase, Jane App, Vodafone and Chubb – have hired for senior execs to take ownership of AI for CX and service more generally. These job titles include: senior lead, AI and automation, customer support; senior PM, help center (CX automation); and VP, head of AI and digital experience.

These job descriptions are not confined to senior roles. Associate and coordinator roles are also starting to include requirements to manage AI. 

In this article, CX Network speaks to AI specialists from JPMorgan Chase & Co., New York Life and Planhat to find out: 

• Why technical roles now fall to CX instead of tech, engineering or IT
• The basics practitioners who are new to this area of management need to know
• The skills required of those who want to transition to such a role 

Don't miss any news, updates or insider tips from CX Network by getting them delivered to your inbox. Sign up to our newsletter and join our community of experts. 

Learn More

Why are technical roles now falling to CX and service instead of tech, engineering or IT?

AI isn't just changing how CX and service work – it's re-writing the job description for practitioners. The experts say this is because organizations already have somebody to own the technology – they now need somebody to take responsibility for the outcomes. 

On the reasons, Shalini Sivasamy, senior AI/ML engineer for New York Life, explains: "Once an agent is talking to customers, the hard questions are less about the model and more about tone, policy, escalation and trust. CX teams own those outcomes, so they are increasingly owning the agents. Engineering still builds the platform, but CX now decides how it behaves."

Sudarshan Reddy Gurampatti, senior machine learning engineer and AI architect for JPMorgan Chase & Co. agrees that managing AI agents is less about owning a piece of technology and more about "owning the experience that technology creates".

"Before, technology teams would build the platform and hand it over to the business. With AI agents, that handover is much harder," he says. "An AI agent talks directly to customers, uses company knowledge, makes decisions and sometimes decides when a human needs to step in. All of that is customer experience," he adds. 

"These new roles sit in between. They need enough technical understanding to work with AI systems, but they also need to understand customers, operations and how the business actually runs," Gurampatti adds. 

Another reason for the shift, according to Kaveh Rostampor, co-founder and CEO of Planhat, is that "the people closest to the customer are the ones who know what the agent needs to do". 

But, he adds: "These titles are the early signs of something bigger." 

Rostampor says that software has traditionally followed the org chart across sales, support, customer success and services, which due to AI, can now work without the constraints of lost handoffs, shift patterns and headcount ceilings. 

"AI removes those constraints, and the roles are starting to merge," Rostampor explains. "A 'senior manager of AI platforms' in CX is a preview of what I call the commercial operator: one person who understands the customer, the revenue, the product and delivery, and who manages both people and agents."

And it's critical this work isn't split across multiple people or functions. 

"The mistake I see is companies buying a sales agent, a support agent and a renewals agent, each seeing only its own slice of the customer. Five islands are five islands, and five agentic islands are still five islands. Someone has to own the whole picture, and that person sits closest to the customer. That's why the role is showing up in CX," Rostampor adds. 

What do CX practitioners need to know about managing customer-facing AI?

As many software engineers joke, managing AI is more akin to managing human children than adult service reps: errors are due to bad instructions. 

On the basics practitioners need to know about this work, Gurampatti and Sivasamy say practitioners should start with a specific, small task. 

Sivasamy says: "Treat an agent like a new team member: it needs clear instructions, approved knowledge and supervision."
Practitioners should also define what success looks like; set clear permissions; monitor continuously; ensure the knowledge base is up to date; and plan for failure.

On the final point, Gurampatti explains: "What does the agent do when it gets confused? What happens if an API goes down or it lacks confidence? Figure out those fallback scenarios beforehand, not after a customer runs into trouble."

He adds: "The goal on day one isn't 100 percent automation. It's building enough trust in the system to expand it safely over time."

Rostampor says practitioners should start with one question: If tomorrow you had one agent (or human assistant that was highly capable) what job would you give them? 

"Then pick one real process in your company, not a grand transformation, and think about how you can instruct your new human assistant (or agent) to do that work," he advises. "Choose repetitive work where one mistake is cheap, like the long tail of smaller accounts," he adds.

When Planhat took this approach, within weeks, 30 percent of human admin work had moved to its multiplayer customer agent Pi, exposing data problems the business didn't know it had. "Across our customers, 90 percent automated at least one process within the first five months of adopting our AI platform," he adds. 

Rostampor also advises tracking the agent's work and having visibility into its actions. This is now required for some organizations under the EU AI Act, but should be best practice for all organizations. 

"Real commercial work is done by teams, so agents should operate where the team can see and correct them," Rostampor says.

"That's why we built Pi to be multiplayer. Your role shifts from doing the work to directing it, and the payoff should be more time with customers, not less."

Key skills for CX practitioners who want a role managing AI agents 

According to Harvard Business Review, the best AI agent managers are not backend engineers. Instead, their expertise lies in domain, quality assurance and operations, and they have a deep understanding of both business processes and customer intent. 

Sivasamy says the key hard skills are:

• Conversation design.
• Knowledge and data management.
• Evaluation and agent testing.
• Governance.

Soft skills, however, are just as important. Sivasamy says practitioners need "enough technical fluency to work closely with engineering, without needing to code". And Rostampor agrees.

"I'd frame it as three skills, and none of them are coding," he says. 

He says practitioners should: 

• Become exceptional at understanding customers: The practitioner's abilities in empathy, trust and reading the room become the advantage and cannot be automated. 
• Learn to build your own workforce: Like engineers, practitioners must become adept at process thinking: breaking a job like a renewal into what information you need, what decisions get made, what action follows and what gets updated. "If you can't describe the work clearly, you can't hand it to an agent," he says. It also takes judgment about guardrails. He adds: "What can an agent decide on its own, what needs a human sign-off, and what does a mistake cost?"
• Learn the business, not just your function: Practitioners are already specialists in customer intent, but a deep understanding of the other emotions that can surface during a journey is also important. They must also understand how customer success, sales, support, product and services connect.

Gurampatti says the biggest mistake a practitioner can make is thinking they need to become a machine learning engineer. A working knowledge of retrieval-augmented generation (RAG), APIs and how an agent uses tools is desirable, but "process and problem-solving skills matter just as much," he says. 

"You need to understand the customer journey, spot where automation genuinely helps and know when an agent should stop and bring in a person," Gurampatti explains.

"Monitoring and governance are part of the role as well. You need to know what to watch for when the system starts giving wrong answers, taking unexpected actions or creating more work than it saves. And communication is essential." 

He adds: "These roles sit between CX, engineering, data, operations, risk and business teams. Being able to translate between those groups is a large part of the job."

The CX role is changing – and practitioners must change, too

Managing the AI agents is one of many new responsibilities in CX. The function also needs AI trainers, systems architects and data specialists.  

These trends are pointing toward a future where CX is embedded across multiple technical and customer-focused roles – but regardless of what the job title says, all practitioners need to understand how AI works and have the ability to manage the outcomes it delivers.

As Rostampor said, the discipline needs a commercial operator, "one person who understands the customer, the revenue, the product and delivery, and who manages both people and agents". 

He's not the only one with this outlook. As explained by Bill Staikos in this article for CX Network, a gap is emerging in how customer outcomes are owned end-to-end, and it calls for a "CX engineer", who can diagnose, understand and execute. 

On the vendor side, the use of forward deployed engineers is changing how CX co-creates solutions and applies them to customer problems. According to CX Network Advisory Board member Andrae Kirkland, it's a role "designed for CX in the age of AI". 

As in all professions, CX practitioners are also required to keep their technical and soft skills sharp. AI will not change that – but it does place a greater emphasis on the need for a senior leader with influence to advocate for the customer and own their experience end-to-end.

Quick links

• Responsible AI vs. AI governance: What CX professionals need to know
• How to improve AI accuracy with better enterprise knowledge
• CX budget insights for 2027: 4 spending tips and 11 questions for your vendors