AI can make customer service faster, more efficient and easier to scale. But what happens when automation starts getting in the way of genuine customer relationships?
For operations leaders, the challenge isn't choosing between AI and people. It's designing customer experiences where technology handles the repetitive work, employees focus on what matters, and customers never feel trapped in a process.
Most of us have experienced the frustration of dealing with an automated support system that simply doesn't understand what we're asking. You explain your problem to a chatbot, receive an irrelevant response and try rephrasing your question, only to find yourself going around in circles. Eventually, you manage to reach a human agent, who asks you to explain everything again.
While the technology may be functioning exactly as intended, the customer experience has clearly failed.
Customers aren't necessarily opposed to interacting with AI. In many cases, they appreciate the convenience of getting an immediate answer without waiting for an available agent. Problems arise when automation becomes an obstacle rather than a solution, particularly when customers cannot easily reach a person, receive generic responses that ignore their history or have to repeat information they've already provided.
For businesses, this highlights an important distinction between handling an enquiry and resolving it. A chatbot might successfully process hundreds of conversations in a day, but that metric means little if a significant portion of those customers leave without getting the help they need.
As we explore in our article on the importance of human connection in a digital-first AI era, technology should strengthen customer relationships rather than create distance between businesses and the people they serve.
The objective, therefore, shouldn't be to automate as many interactions as possible, but to use automation where it genuinely improves the experience.
One of the biggest misconceptions surrounding AI adoption is that automation must come at the expense of human involvement. In reality, some of the most valuable applications of AI are those that support employees behind the scenes, allowing them to spend less time on repetitive administration and more time helping customers.
Consider a customer who wants to check the status of an order. An AI assistant connected to the appropriate system can retrieve the information and provide an accurate update within seconds. The interaction is straightforward, and involving a human agent would offer little additional value.
Now consider a customer who has contacted your business several times about an unresolved billing issue and is considering cancelling their contract. Although AI can retrieve previous conversations, identify the account and summarise the problem, resolving the situation requires a different approach. A human agent is better positioned to acknowledge the frustration, exercise judgement and take responsibility for finding a resolution.
Both scenarios demonstrate the value of AI, but they also illustrate why businesses need to be selective about where automation takes the lead.
A useful starting point is to assess customer interactions according to their complexity, sensitivity and potential consequences.
This doesn't mean AI has no place in more complex interactions. It can still help employees retrieve information, prepare summaries and identify relevant next steps. The difference is that responsibility for the outcome remains with a person who can assess the situation in context.
Rather than asking how many customer interactions can be automated, businesses should consider which tasks are predictable enough for AI to handle reliably and where human involvement will create a better outcome.
The aim is to automate repetitive processes without automating away the relationship.
Even a sophisticated AI assistant will struggle to deliver meaningful customer support if it operates without access to relevant customer information. This is why CRM integration plays such an important role in creating connected, personalised experiences.
Imagine a long-standing customer contacting your business about an issue that has already been discussed with your support team. If the AI assistant only has access to the message typed into the chatbot, it has no understanding of the customer's history, previous conversations or outstanding support tickets. As a result, the interaction begins from scratch, regardless of how much information the business already holds.
When AI is appropriately integrated with a CRM, however, it can draw on relevant customer records to provide better context, recognise recurring issues and help route enquiries to the right people. This information can also accompany a conversation when it is transferred to a human agent, allowing the employee to pick up where the previous interaction ended.
For businesses using HubSpot, this could involve combining AI capabilities with Service Hub, CRM records, help desk and automation workflows. Depending on the configuration, these systems can support ticket categorisation, customer identification, escalation rules and access to previous interactions.
Our guide on how to create a customer agent in HubSpot explores how businesses can configure AI-assisted conversations using trusted knowledge sources and defined handover processes.
However, connecting AI to a CRM doesn't automatically guarantee a better customer experience. The information needs to be accurate, appropriately structured and accessible within the correct permissions. Incomplete records, duplicate contacts and disconnected processes can quickly undermine the benefits of automation.
We experienced the importance of this foundation firsthand, as discussed in our case study on cleaning up CRM data with AI. Better data governance creates more reliable information for both employees and automated systems, making it easier to deliver consistent customer experiences.
Before investing in sophisticated AI tools, businesses should therefore assess whether their existing data and processes are ready to support them.
One of the most overlooked aspects of AI implementation is what happens when automation reaches its limits. Businesses often invest considerable time designing chatbot conversations but pay less attention to how those interactions transition to human employees.
Yet the quality of that transition can have a significant impact on the customer's overall experience.
A well-designed handover should feel like a continuation of the same conversation rather than an entirely new interaction. Customers should know how to request human assistance, understand when someone will respond and feel confident that the information they've already provided won't be lost.
Achieving this requires several practical considerations.
Make human assistance accessible. Customers shouldn't have to navigate multiple unsuccessful chatbot responses before being offered an alternative. A clear escalation option should be available, with realistic expectations about response times when agents are unavailable.
Establish meaningful escalation triggers. Repeated misunderstandings, unresolved complaints, sensitive requests and explicit requests for human support are all potential indicators that automation should step aside. These triggers should reflect the organisation's customer service priorities and the consequences of an incorrect response.
Preserve the conversation context. When an interaction is transferred, the employee should receive the relevant conversation history, customer details and reason for escalation. This reduces repetition and allows the agent to focus on resolving the issue rather than reconstructing what happened.
Assign clear responsibility. Every escalated conversation needs an accountable team or individual, along with an appropriate response expectation. Without clear ownership, even a technically successful handover can leave customers waiting for assistance.
These principles are relatively straightforward, but they require coordination between customer service teams, CRM administrators and the people responsible for implementing AI.
The ultimate measure of a successful handover isn't simply how quickly the conversation reaches an agent. It's whether the customer feels understood and supported when that agent responds.
For many businesses, the appeal of AI lies in its ability to improve efficiency, reduce operational costs and help teams manage increasing volumes of customer enquiries. While these are legitimate objectives, they shouldn't be the only measures used to evaluate success.
A reduction in support tickets, for example, might initially suggest that automation is working well. However, if customers are abandoning chatbot conversations, contacting the business repeatedly or expressing greater dissatisfaction, those apparent efficiency gains may be masking a larger problem.
To understand whether AI is genuinely improving customer service, businesses need to measure operational performance alongside customer outcomes.
Five particularly useful metrics include:
These metrics are most valuable when considered together. A high automation rate accompanied by declining satisfaction or increasing repeat contacts should prompt further investigation rather than celebration.
Businesses should also involve their customer service teams in reviewing AI performance. Employees often identify recurring problems, inaccurate responses and gaps in the knowledge base before those issues become obvious in reporting.
By combining customer feedback, operational data and employee insights, organisations can make more informed decisions about which processes to improve, expand or return to human ownership.
While customer experience is often the focus of AI implementation, employee adoption is equally important. Introducing automation can create uncertainty about changing responsibilities, performance expectations and the future role of customer-facing teams.
If employees view AI primarily as a threat, they're unlikely to embrace it enthusiastically. Businesses therefore need to communicate clearly about why the technology is being introduced, what it will be used for and where human judgement will remain essential.
One of the most effective ways to build confidence is to involve employees in identifying automation opportunities. Customer service teams understand which enquiries are repetitive, where processes create unnecessary administration and which situations require specialist attention.
Rather than approaching AI implementation with the question of which roles can be replaced, businesses can ask which tasks prevent employees from spending more time helping customers.
The answers may reveal opportunities to automate conversation summaries, ticket categorisation, knowledge retrieval and routine CRM updates. These improvements can reduce administrative workloads while allowing employees to focus on more complex, valuable interactions.
Training and ongoing feedback are equally important. Employees need to understand how AI tools work, when to question their outputs and how to intervene when automation produces an unsuitable response.
Ultimately, successful AI adoption depends on more than choosing the right technology. It requires employees who understand its purpose, trust the processes surrounding it and feel equipped to use it effectively.
AI has enormous potential to improve customer service, but the most successful implementations won't necessarily be those that automate the greatest number of interactions. They'll be the ones that make it easier for customers to get help, give employees better information and create more consistent experiences across every touchpoint.
Achieving that balance requires a clear understanding of the customer journey, reliable CRM data and thoughtfully designed processes that connect automation with human expertise.
At Velocity, we help businesses assess their AI readiness, optimise CRM environments and implement automation strategies that support both operational efficiency and customer experience. Rather than introducing technology for its own sake, we focus on identifying the opportunities where it can deliver meaningful business value.
The future of customer service isn't about choosing between AI and people. It's about using AI to help people deliver better service.
If you're exploring AI or looking to improve an existing implementation, let's discuss how Velocity can help you build a more connected, human-centred customer experience.
Businesses should begin by identifying predictable, repetitive tasks suitable for automation while keeping human assistance accessible. Connecting AI to relevant customer information and establishing clear escalation processes helps ensure that customers receive efficient support without sacrificing personal attention.
Human-in-the-loop AI involves employees reviewing, approving or taking over specific automated processes. It's particularly valuable when decisions involve complex complaints, sensitive information, financial consequences or situations where human judgement and accountability are essential.
Complex complaints, sensitive personal matters, significant billing disputes and interactions requiring empathy or professional judgement generally benefit from human oversight. AI can still support these processes by retrieving information, summarising conversations and helping employees prepare responses.
A combination of customer satisfaction, first contact resolution, repeat contact rates, escalation performance and customer effort provides a more complete picture of AI effectiveness. These metrics help businesses understand whether automation is genuinely resolving issues or simply reducing the number of interactions handled by employees.
Yes. AI can reduce repetitive administrative work, assist with routine enquiries and provide employees with faster access to relevant information. When implemented thoughtfully, this allows customer service teams to spend more time resolving complex issues and building stronger customer relationships.