Agentic AI in the contact centre: hype, added value or risk?

The concept of agentic AI is gaining traction in the contact centre industry. In vendor pitches, innovation roadmaps and debates about the future of customer service. But how often have we seen this before? New labels promising “autonomy”, while in real life you’re still struggling with fragmented customer data, system dropouts and mounting pressure on the floor.

So it’s worth taking a step back. Not to dismiss agentic AI as the next buzzword, but to clarify when it genuinely adds value for the customer experience (CX) and employee experience (EX) – and when you risk drifting into over-automation. Let’s dive in!

What is agentic AI?

Agentic AI refers to AI systems that don’t merely respond to input, but can independently and purposefully carry out tasks by planning their own steps and orchestrating tools. In other words, this type of AI goes beyond generating text or analysing data: it can make decisions and take action to achieve a defined objective.

What does an AI agent do?

The difference between an AI agent and a chatbot or generative AI is very tangible. An AI agent goes beyond drafting responses, summarising conversations or retrieving knowledge articles. It can, for example, create a case, trigger follow-up actions, initiate a process or combine multiple actions – all within predefined guardrails.

While many AI implementations today remain largely reactive – such as chatbots answering questions – agentic AI can act proactively. It chains together multiple steps, determines what’s required to reach a goal and uses different systems and tools to get there. In doing so, it comes much closer to the day-to-day work of your agents.

Is agentic AI mainly a buzzword or does it offer genuine added value?

The answer is nuanced. On the one hand, “agentic AI” is certainly a popular term that is currently widely used – and sometimes misused – as a marketing label. Gartner explicitly warns of agent washing: solutions that are essentially chatbots or smart workflow automation, but are marketed as agentic AI.

On the other hand, the underlying technology enables something that many previous generations of AI could not: autonomous planning, decision-making and action across multiple systems. The real added value lies in workflows that combine numerous steps, data sources and systems – particularly in contact centres where speed, consistency and error reduction are critical. It’s in precisely these environments that agentic AI makes a difference, because it understands customer intent and can act on it independently.

1) Gartner Newsroom

Where does agentic AI deliver the most impact in the contact centre?

In the contact centre, agentic AI proves most valuable where customer service is currently under strain. Not in neat, clearly defined use cases, but rather in environments where volumes fluctuate, systems intersect and decisions influence one another.

In many contact centres, we still see familiar patterns: significant manual coordination, limited time to think ahead, and tooling that provides insight but offers little real operational support.

It’s exactly within this field of tension – high dynamics, multiple dependencies and pressure on people – that the role of AI shifts. From analysing and advising to actively collaborating, anticipating and adjusting in real time. This shift is particularly visible in two domains where the impact is immediately tangible.

Agent support: from co-pilot to task partner

Many contact centres already use co-pilots, often powered by generative AI (GenAI). They assist agents with knowledge retrieval, summaries and suggestions. They accelerate the work, but responsibility and execution remain entirely with the agent.

Agentic AI begins where the system itself takes on tasks based on the customer’s objective. For example, by:

• recognising the type of query and the desired outcome
• independently determining which systems and steps are required
• combining information from multiple sources
• preparing cases, forms and follow-up actions in parallel
• initiating back-office processes without explicit instruction

The agent remains accountable, but no longer has to orchestrate every step. The AI agent collaborates, monitors progress and requests input only where genuinely necessary.

Workforce management: from planning to continuous optimisation

Workforce management (WFM) is an area where automation has long been present. Forecasting, scheduling and intraday management are supported by tools, BI and RPA. In that form, however, WFM remains largely rule-based: deviations are flagged and people decide what action to take.

Within workforce management, we only speak of agentic AI when the system itself is given responsibility for achieving an objective – such as safeguarding SLAs or managing workload. In that case, the system can:

• not only detect intraday deviations, but interpret them in context
• weigh up multiple interventions (re-routing, flex pools, overtime, channel steering)
• model scenarios before intervention is required
• factor in previous actions and their impact
• independently execute proposals within agreed parameters

The supervisor’s role then shifts from manual intervention to governance and exception management. The system is no longer merely advisory, but an active participant in day-to-day operations.

Concrete use cases of agentic AI in customer service

Because the boundary between agentic AI and other forms of automation is not always crystal clear, here are a few practical examples:

Insurance provider – claims handling

Traditional automation: a chatbot assists with completing a claim form; a co-pilot supports the agent in responding to the claim.

Agentic AI: independently gathers policy data, previous claims and photos, requests missing information, builds a complete case file and initiates the claims assessment or payout workflow.

Parcel delivery company – delivery issue

Traditional automation: a chatbot provides track & trace information and offers standard redelivery options.

Agentic AI: identifies recurring delivery issues, determines the best alternative, automatically adjusts delivery planning and customer preferences, implements changes in logistics systems and proactively informs the customer.

Healthcare provider – rescheduling an appointment

Traditional automation: a chatbot offers available time slots; a staff member handles the rest.

Agentic AI: checks diaries, constraints and file requirements, independently books a suitable new slot, sends confirmations and ensures all necessary information is available in the patient file in advance.

The risk of over-automation

This may well be the most important point. In many organisations, automation has long been driven primarily by cost reduction. But customer service is not a factory: the customer experience is shaped by control, trust and convenience.

Over-automation often leads to the opposite. Customers have fewer choices, get lost in self-service journeys or are connected to a human too late. The result: repeat contact increases, escalations rise, NPS declines and agents are left dealing mainly with frustrated customers.

There are also genuine risks around autonomy and security. If misconfigured or misused, agentic systems could carry out unintended actions or expose sensitive data. That’s why agentic AI requires clear guardrails.

The real challenge is not to automate blindly, but to introduce agentic AI in a smart and controlled way – with human intervention where necessary and clear boundaries around what can and cannot be automated.