In 2025, customer service leaders are facing a host of familiar problems. Reducing average handling times. Increasing first-time resolutions. Providing a multi-channel experience. Improving customer satisfaction and loyalty.
But as businesses try to blend human support with automated technology, consumers increasingly expect more from these experiences.
A report from the think tank, New Britain, found that 78% of Britons felt frustrated when dealing with customer services and spent between 28 and 41 minutes per week dealing with support teams. Satisfaction is at its lowest point in a decade.
The UK Institute for Customer Service revealed that 24% of 15,000 people surveyed had a poor experience in 2024.
Technology has unlocked new ways for businesses to scale customer service, including implementing task-based automation tools like chatbots.
The aim was to scale their customer service operations by automating specific, repetitive tasks and free up their human staff to deal with more complex problems and high-impact interactions that drive loyalty. But task-based automation has its limitations.
“These technologies were designed to solve single tasks,” says John Kelleher, Zendesk’s VP of Enterprise Sales.
“This could be FAQs, such as retrieving log-in details and returns and refund policies. But what they can’t do is answer unusual queries or more nuanced, multi-part questions. Often, they also don’t pass on information to human operators, meaning customers have to repeat the same query, which might cause frustration.”
Introducing AI agents
Over 100,000 Zendesk AI customers have sought to alleviate these frustrations and elevate their customer service experience by using AI agents alongside human employees.
Agents are software that can independently make decisions and take actions to achieve specific goals.
They are designed to operate like virtual customer service assistants. Rather than just responding to commands, they’re capable of solving complex problems, communicating with other agents and human customer support staff and continuously learning.
For businesses, AI agents create an opportunity to scale their customer service operations intelligently by building teams of agents in the same way they would with human staff. “Different agents have different skill-sets and responsibilities,” says Kelleher.
“One may be responsible for researching information from different knowledge sources, while another will deliver the information to the consumer in the right tone of voice for specific interactions.”
Intelligent orchestration is critical to remove organisational silos that can prevent AI agents from accessing customer information across multiple channels and optimising their ability to resolve queries.
Orchestration refers to the automated coordination of agents, systems, channels, data and actions across multiple tools and teams to achieve a specific outcome efficiently and intelligently.
For example, an AI agent dealing with a customer query on a company website could simultaneously communicate with an AI or human agent operating on WhatsApp and retrieve information from a CRM. This is achieved by positioning Zendesk’s software as a central hub that connects to the rest of the business
Balancing efficiency with empathy
But these increases in efficiencies in the retrieval, routing and processing of information can be undone if it isn’t delivered in a personalised and empathetic manner to the end consumer. The relationship between customer and support teams is a delicate one.
Zendesk benchmark data shows that more than one-half of consumers will switch to a competitor after only one bad customer service experience. In addition, 73% of consumers say they will switch to a competitor after multiple bad experiences.
Personalisation begins with providing a premium multi-channel experience, so AI and human agents can provide the same excellent customer service across every channel and communicate with consumers in their preferred format.
Agents can recall a customer’s preferred form of communication, for example a phone call, email, voice note or knowledge article and even create new channels, such as messaging services, to deliver that information.
But empathy is a bigger challenge. To alleviate frustration, AI agents need to be able to converse in a conversational, empathetic and human-like manner. “Tone of voice and empathy needs to feel individual to the consumer,” says Kelleher.
“A concerned parent inquiring about a health related issue to a provider of consumer goods for kids, requires a very different tone of voice to a teenager wanting the latest updates on tour information from their favourite pop artist.
The Al agent has the intelligence to choose appropriate tones, based on enquiry and consumer context.”
Using AI agents to empower human staff
Early adopters of AI in customer service are already generating ROI. UK ethical cosmetics retailer LUSH supports its human agents with a custom-built AI agent called Marvin. The agent saves them time by solving their most repetitive queries, for example, sales and discounts, donations, order dissatisfaction and discontinued products.
Marvin also gives human agents context to solve issues faster by requesting customer info upfront and adding tags and labels to incoming tickets.
This has helped Zendesk’s customer service team to achieve a 60% first contact resolution (FCR) rate.
This saves LUSH about five minutes per ticket or about 360 agent hours each month. With that time saved, human agents can dedicate more energy to meaningful customer interactions.
These human interactions with consumers remain critical to the brand, who want to replicate the friendly in-store high street experience that generated brand loyalty long before they became a global online success.
Speaking at Zendesk Relate, LUSH’s Customer Care Manager, Naomi Rankin, says agents have been trained to deliver upon SLAs (service level agreements) and sentiment.
“We’re now a global business with millions of customers, but we still want to provide a service that ensures people feel listened to, enables them to have a friendly chat and finds a resolution.”
AI-powered customer service can also be a tool for long-term business growth. LUSH uses AI analytics to listen to customer feedback and spot trends. Positive online sentiment around specific products can help to identify which products to market or anticipate future demand.
Businesses that use AI to empower employees, engage customers in positive interactions and glean insights will turn customer service from problems to profit.
The intelligent automation playbook: 5 steps to evolve your customer support journey
Education: Build AI fluency
One of the biggest barriers to AI implementation is education. Leaders and service teams must be equipped with a strong understanding of how AI works and what it means for their service model. “Education must come before strategy,” says Kelleher.
“Some leaders feel pressured into creating AI strategies before they thoroughly understand what the technology can and can’t do. We work with customers to deliver a lot of that early education.”
Innovation decks, third-party advisors, or ex-CIOs can also bring a real-world perspective and shift the mindset from AI as a tech deployment to AI as a transformation tool.
Map service journeys by complexity, not just channel
Before AI is introduced, businesses must understand the urgency, complexity and sensitivity of each type of customer service interaction.
“Mapping specific journeys can help to decide which interactions can be automated immediately, those that require a hybrid human and AI approach and queries which are best kept human-focused,” adds Kelleher.
Future state journeys should also be mapped with an AI-first mindset, imagining how AI can support or lead each one.
Train AI with real service data, not assumptions
AI agents perform as well as the data they’re trained on. To deliver accurate, helpful, and brand-aligned responses, AI needs to be fed real service data.
“Your tickets, conversations, knowledge bases, and customer interactions should provide this training,” says Kelleher. “This allows agents to learn from actual customer needs rather than hypothetical assumptions.”
The richer and more contextual the data, the faster AI will understand, adapt, and improve. “It’s like training up a digital graduate: the better the training, the better the performance and the fewer risks of bias or miscommunication.”
Equip agents with context, not just tickets
The more context an agent has, the better it can understand and resolve queries. Connecting agents to CRM platforms will enable them to access customer history, preferences, purchase behaviour, and lifecycle stage.
“This helps the agent recognise returning customers, tailor responses based on past issues, and even preempt questions based on current orders or usage patterns,” says Kelleher.
Giving agents access to other knowledge sources, such as user web and app behaviour and conversation history, can also turn them into proactive problem solvers.
Continually refine through feedback loops
AI agents can learn continuously. Feedback loops that enable leaders to continually monitor, evaluate, and improve performance are key.
This can be achieved through (white-box) AI models that explain decision-making, so leaders can see how and why responses are made.
This can also be combined with AI-powered QA tools that analyse 100% of interactions for sentiment, speed, and accuracy.
These insights can then be fed back into training to evolve AI agents over time. With the right loop in place, AI becomes a living system, constantly learning, adapting, and getting better with every interaction.
To find out more information, please visit: www.zendesk.co.uk
In 2025, customer service leaders are facing a host of familiar problems. Reducing average handling times. Increasing first-time resolutions. Providing a multi-channel experience. Improving customer satisfaction and loyalty.
But as businesses try to blend human support with automated technology, consumers increasingly expect more from these experiences.
A report from the think tank, New Britain, found that 78% of Britons felt frustrated when dealing with customer services and spent between 28 and 41 minutes per week dealing with support teams. Satisfaction is at its lowest point in a decade.
