How agentic AI reshapes customer support

From answering to acting

Customer support is entering a new phase. During the UBA 60 minutes session on 19 May 2026, Hannah Patronoudis, Head of Experience at Raccoons, and Michiel Vandendriessche, Co-founder and keynote speaker at Raccoons, showed how agentic AI can change the role of AI in service teams.

Until now, many AI applications in customer support have focused on helping humans work faster: summarising messages, suggesting answers or improving FAQ search. Agentic AI goes further. It allows AI systems to decide which steps are needed, use tools, carry out actions and keep working until a goal is reached.

This shift matters for brands. A support interaction is often one of the most direct moments of contact between a customer and a company. If an AI agent responds quickly, accurately and in the right tone of voice, it becomes part of the brand experience. If it fails, that failure also reflects on the brand.

The foundation: trusted knowledge

Before thinking about autonomous agents, companies need a strong knowledge base. The session highlighted the role of knowledge assistants: AI systems that answer customer or employee questions based on validated company information.

These systems can use product documentation, policies, FAQs, intranet pages or support articles to generate reliable answers. They can work in several formats, such as a chatbot, an email drafting tool or a search experience. The most important point is that answers are grounded in company-approved sources, not in generic AI knowledge.

For marketing and brand managers, this means the quality of the content behind the AI becomes essential. Clear product information, consistent policies and up-to-date support content directly influence the quality of the customer experience.

A customer support team of AI agents

Raccoons illustrated how a multi-agent customer service team could work. One orchestrator agent receives and classifies incoming requests, then decides whether to route them to a specialist agent or escalate them to a human. A knowledge agent retrieves product, policy or FAQ information from a validated knowledge base. A CRM agent checks customer profiles, order history and previous interactions.

In practice, a customer might write: “My parcel hasn’t arrived.” The orchestrator recognises this as a delivery issue. The CRM agent checks the order status. The knowledge agent retrieves the relevant policy. Together, the agents prepare the next step or customer response, while a human can still intervene when needed.

This is not just faster support. It changes who, or what, is doing the work.

Trust, integration and human judgement

The main challenge is not only technical capability. It is implementation, integration and trust. AI agents need access to tools such as CRM systems, email platforms, ticketing tools or order databases. That access must be designed carefully, with clear guardrails and decisions about where autonomy is appropriate.

The speakers also stressed the difference between using a consumer AI platform and using a large language model through an API in a controlled business environment. For sensitive customer or CRM data, architecture, hosting, terms and conditions, and data governance matter.

Humans remain essential where creativity, judgement and responsibility are needed. The role of the organisation is to define where AI may act autonomously, where approval is required and where escalation to a person is mandatory.

What brand managers should ask now

Agentic AI makes customer support a strategic brand topic. The question is not simply: “How much can we automate?” A better question is: “What experience do we want an autonomous system to deliver on behalf of our brand?”

That includes tone of voice, service standards, escalation rules and the moments where human empathy must remain central. Brands that prepare now can build AI support systems that are not only efficient, but also consistent, useful and aligned with customer expectations.