AI Customer Service Needs Everyone in the Room

AI is now woven into customer service in a way that would have felt experimental only a few years ago. Chatbots answer common questions. Automated systems triage requests. Natural language tools help route customers, summarise conversations, and suggest replies. In electronic marketplaces, where service is often the main point of contact between a platform and its users, those tools can shape whether a customer feels helped, dismissed, or willing to come back.

The study "Aligning stakeholders in AI-enabled customer service: Toward human-centric adoption in electronic marketplaces" looks at this issue from a useful angle. Rather than treating AI adoption as a technology decision alone, it asks what happens when different people involved in the same service system see that change differently. The researchers surveyed end users, frontline agents, and supervisors, then added interview insights to understand the gaps between them.

Those gaps matter: AI in customer service can look successful from a dashboard while feeling clunky to customers and frustrating to the staff expected to work with it. A faster response time means little if the customer has to fight the bot to reach someone. A new tool may look efficient to management while adding extra steps for agents. The article's main point is simple enough, but often missed: AI adoption works best when the people affected by it are properly considered.

Customers are open to AI, but they still want human fallback

The customer findings are less anti-AI than some businesses might assume. Many users are willing to accept automation when it improves the service experience. They value speed and access, especially when the system can answer something clearly without creating more effort.

Still, the human element remains central. In the study, 88.3% of end users said they wanted the option to speak to a human even if AI was being used. That figure says a lot. Customers may tolerate automated service for simple requests, but they do not want to feel trapped inside it. When the issue is sensitive, confusing, emotional, or time-consuming, the ability to reach a person becomes part of the service itself.

The article also found that users placed high value on empathetic treatment and first-contact resolution. These are not decorative parts of customer service. They are often the difference between a customer feeling that their problem has been handled and feeling that they have been processed.

For marketplaces and digital platforms, this is especially important. Customers may never meet a staff member. They may not know who is responsible when something goes wrong. Customer service becomes the place where trust is either repaired or weakened. A useful bot can help. A bot that blocks access to help can damage the relationship quickly.

Frontline agents see the problems before the dashboard does

One of the strongest parts of the research is its attention to frontline agents -  intrinsic people within customer service who are increasingly being overlooked, but shouldn't be. These are the people who deal with customers after automation has failed, confused someone, or pushed a more complex issue into the queue. Their perspective is often less optimistic than the view from management.

In the study, 58.1% of agents said automation had worsened service quality. Many described increased workload, unclear changes to their roles, and limited training. That is a serious warning for any organisation introducing AI into service teams.

The problem is not necessarily the technology itself. The issue is how it lands inside the working day. If AI creates summaries that need to be corrected, escalates cases without enough context, or gives customers wrong expectations before they reach a human, agents inherit the mess. They become responsible for smoothing over the gap between what the system promised and what it actually delivered.

Training was another clear concern. The article found that agents often felt underprepared for the changes automation introduced. This makes adoption harder because staff are expected to trust and support tools they may not understand, influence, or benefit from.

There is also an emotional layer here. If agents believe AI is mainly being introduced to reduce headcount or make their work less valued, resistance is understandable. A service team cannot be expected to champion a tool that feels like a threat. Human-centric adoption means involving agents early, explaining what is changing, and giving them a real role in shaping how AI is used.

Managers need a wider definition of efficiency

Supervisors and managers in the study were more positive about AI's impact. Most believed automation had improved service efficiency, either moderately or substantially. That view makes sense from an operational perspective. AI can reduce volume, standardise answers, and improve consistency when applied well.

The difficulty is that efficiency can mean different things depending on where someone sits. For a manager, efficiency might mean fewer tickets reaching human agents. For a customer, it might mean getting the right answer without having to repeat the problem. For an agent, it might mean having better context and fewer avoidable escalations.

The article shows that supervisors sometimes underestimate the friction felt by frontline staff. They may acknowledge that training is needed, but still frame success mainly through productivity and operational metrics. That creates a risk. A service operation can become faster on paper while trust erodes in practice.

A better approach would treat efficiency as something broader than speed or cost reduction. It should include customer confidence, employee workload, escalation quality, and the ability to resolve complex cases well. If AI helps agents focus on the work that needs judgement, care, and problem-solving, it can strengthen the service model. If it simply pushes volume around, the benefits will be fragile.

The article's findings point toward a more realistic way to think about AI-enabled customer service. Start with the people who use it, work beside it, and manage it. Give customers a clear route to human help. Give agents training and influence. Give supervisors better measures than speed alone.

AI can improve customer service, but only when adoption is treated as a service design and workforce issue as much as a technology rollout. In electronic marketplaces, where customer service often carries the weight of trust, that distinction becomes hard to ignore.

Sources

Lopez-Lopez, D., Fondevila-Gascón, J.-F., Torres-Peregrina, B.-M., & Marco-Simó, J.-M. (2026). Aligning stakeholders in AI-enabled customer service: Toward human-centric adoption in electronic marketplaces. Electronic Markets, 36, Article 42. https://doi.org/10.1007/s12525-026-00892-1