Improve NPS and save time with an AI-powered Call Center

The generative AI revolution is underway, and it will radically transform our customer service centers. This transformation isn’t just a technological upgrade: AI-powered Call Centers promise to redefine the customer experience while ultimately optimising business productivity.

This article is taken from Dimitri Cabaud’s talk during the Microsoft AI Tour Paris.

In August 2022, BCG published a revealing figure: 95% of customer service center managers anticipated the intervention of an AI bot in their customer journey within three years. This prediction is coming true at lightning speed, even exceeding initial expectations.

Download Now: 50+ AI Reference Cases
[Updated January 2025]

Customer service center costs and issues

To understand the magnitude of this transformation, let’s dive into the current reality of customer service centers. Today, a simple phone call costs companies an average of $13 ( source Gartner ). Even more worrying, only 14% of customers resolve their problems independently by consulting FAQs and available documentation. Even for problems considered “easy,” the rate of autonomous resolution peaks at 36%. These figures reveal considerable room for improvement.

Call centers also face a significant challenge: turnover. The financial impact is considerable, with a staff turnover rate of between 18 and 25% and an estimated replacement cost of $14,000 per agent. These often stressful environments require a profound transformation, where agents must manage complex and sometimes tense situations.

$3.5: This is the average cost of a customer contact on an assisted channel (call, email or chat).

$14,113: This is the cost of replacing an agent, the turnover rate is on average between 18-25% in call centers.

Lower inference costs, an opportunity

Technological evolution is accelerating dramatically. In just one year, we have gone from the early days of video generation with Stable Diffusion to mind-blowing achievements like OpenAI’s Sora.

At the same time, the costs of using these technologies are falling rapidly: between GPT-3 and GPT-4, then between GPT-4 and GPT-4o, costs have been halved with each iteration. The arrival of NVIDIA’s new Blackwell GPUs even promises a 30-fold increase in inference performance (GB200 vs. H100), making use cases that were previously too expensive accessible.

Voice remains a preferred communication channel, particularly in France, where 15% of the population is illiterate. Beyond accessibility, voice communication has undeniable advantages: It allows 200 words per minute to be transmitted, compared to 40 to 100 in writing, while conveying essential non-verbal information such as emotions, intonations, or speed of speech.

Elizabeth Fresnel, speech therapist and director of the Voice Laboratory in Paris, said, “I compare the voice to a painter’s palette.”

Download Now: 50+ AI Reference Cases
[Updated January 2025]

Generative AI use cases in customer service centers

Enter GPT-4o (o for omni), OpenAI’s first truly multimodal model. It can process text, audio, and images simultaneously. In its recently released real-time version (October 2024), it responds in just 320 milliseconds on average—the natural response time of a human conversation.

Even more impressive, it understands and adapts to emotions, can interact with different APIs, and supports multiple languages. At just 6 cents per minute of input audio and 24 cents per minute of output audio, it makes it economically viable to automate many interactions.

The impact of this technology on customer service centres revolves around three principal axes:

• First, it can significantly improve productivity while reducing costs. In 2022, Gartner stated that: “By 2026, the deployment of conversational artificial intelligence in customer service centers will reduce agent labour costs by $80 billion.”
• Second, it increases customer satisfaction through always-on availability, multilingual support, and faster, more relevant responses. BCG says companies are already seeing improvements in customer satisfaction by integrating generative AI into text-based chatbots. Companies like Octopus Energy, using generative AI to handle email queries, have seen an 18% increase in satisfaction scores compared to purely human responses.
• Finally, it generates additional revenue by reducing agent turnover and enabling a more proactive approach to sales. According to a 2023 Salesforce study: “88% of customers say that good customer service makes them more likely to buy again.”

This revolution is not limited to large call centers. Small businesses and local shops can all benefit from these technological advances. We are no longer talking about chatbots but real voice assistants capable of natural and effective interactions.

The transformation of customer service centers through generative AI is not a simple technological evolution: it is a revolution that redefines customer service standards. It promises a future where customer satisfaction and operational efficiency are no longer contradictory objectives but natural allies in creating value for companies and their customers.