Artificial intelligence has made its way into every contact centre. Chatbots, voicebots, generative AI to assist advisors, automated analysis of customer verbatims — the promises are real. But so is the risk of over-automation. Should you automate everything? Preserve everything human? Or find a balance — and if so, what does that look like in practice?
Let’s start with the facts. AI delivers measurable, documented benefits across customer operations:
These benefits are real. But they have limits.
The enthusiasm around AI in customer service often obscures a more nuanced reality. Certain situations resist automation — and forcing them into a bot degrades the customer experience significantly.
A customer calling because a gift they ordered did not arrive in time for Christmas. A policyholder discovering their claim is not covered. A patient trying to understand a complex medical bill. These situations require empathy, flexibility, and human judgement. A bot that responds with polite but irrelevant messaging does not neutralise dissatisfaction — it amplifies it into anger.
As soon as a request requires cross-referencing multiple sources of information, exercising judgement on an unusual situation, or negotiating a tailored solution, the limitations of automated systems become apparent. Generative AI is advancing rapidly on this front but remains fragile on edge cases.
A human advisor can build a connection, recall a previous conversation, and adapt their tone to the customer’s personality. This relational capital is a powerful retention lever that automation cannot replicate — at least not yet.
When a customer is angry, the instinct of many companies is to offer a form or a chatbot. This is often the worst possible decision. A trained human advisor de-escalates the situation; a bot frequently makes it worse. Customer rage directed at a bot rarely ends well — for the brand or the customer.
In 2026, the question is no longer “AI or human?” It is “AI and human: how do you combine them intelligently?”
The hybrid model assigns each type of interaction to the most appropriate handling channel:
AI does not replace the human — it frees the human from repetitive tasks so they can focus on what they do best: creating relational value.
Deploying a chatbot “to see what happens” — without clearly defining which interaction types it should handle, to what depth, and with what limits — is a reliable recipe for a tool that either fails to add value or actively damages the experience.
Every bot must have an accessible route to a human advisor — easy to reach, without requiring the customer to repeat their entire request. The “bot tunnel” with no exit is one of the most significant sources of customer frustration in 2026. Any deployment without a clear, well-signposted escalation path is incomplete.
A high containment rate (the proportion of interactions handled without human intervention) is not inherently a success indicator if customer satisfaction is simultaneously declining. AI must be measured against the same KPIs as human advisors: CSAT, FCR, NPS. Operational efficiency metrics that are not correlated with satisfaction improvements are not evidence of value creation.
AI changes the advisor role — it does not eliminate it. Teams must be trained to work alongside AI tools: how to validate suggestions, understand system limitations, and know when to override an automated recommendation. Without change management and adoption support, even the best technology fails to deliver.
At Armatis, AI is not an end in itself — it is a lever in service of customer experience quality and operational performance. Our hybrid frameworks combine:
This model allows us to deliver superior service quality at controlled cost — while preserving what makes the real difference in customer service: human intelligence.
Not within the foreseeable future — and probably never entirely. AI takes over repetitive, low-value tasks, which shifts the advisor role toward greater complexity, expertise, and relationship management. The overall volume of employment in customer service is stabilising rather than declining, with a structural evolution in the skills required. The advisors of 2026 are more expert, more empathetic, and more commercially capable than those of ten years ago.
Generative AI primarily enables two things in a contact centre environment: real-time advisor assistance (suggesting responses, summarising history, proposing next-best actions) and automating the drafting of written responses on certain channels. Early real-world deployments show advisor productivity improvements of 20% to 40% on the interactions where these tools are applied.
Yes — if and only if it is well-designed, properly trained on precise use cases, and paired with an effective human fallback. A chatbot that genuinely resolves the customer’s issue scores better than an advisor with a 10-minute wait time. The problem is that many chatbots do not actually resolve the issue — they deflect it temporarily, creating frustration that surfaces through a different channel shortly after.
Key indicators to track: containment rate, cost per interaction before and after deployment, CSAT and NPS compared across bot and human interactions, first contact resolution rate, and advisor productivity (interactions handled per hour). A successful deployment simultaneously improves cost efficiency and customer satisfaction — if only one of these improves, the deployment is incomplete.
Armatis designs hybrid AI + human frameworks tailored to each client’s specific CX objectives. Our approach: automation where it creates value, human expertise where it makes the difference. Let’s discuss your optimisation project.
Contact our teams to discuss your challenges.
Join the leaders who trust our multilingual and technological expertise.