AI in customer service: voicebots and automation

Home Page AI in customer service means that an AI application independently understands and processes incoming customer enquiries: by telephone via an AI voicebot, via text via AI chatbots, or in the background through automated prioritisation and summarisation. Unlike traditional menu systems, today’s AI systems recognise natural language and the intent behind a query, rather than following a rigid set of rules. The expectations placed on artificial intelligence in customer service are high, and so is the complexity of its implementation. That is why this article focuses less on the theory and more on the practice: how an AI voicebot works from a technical perspective, what it has achieved in a real-world project, what AI cannot do, and what you should bear in mind when implementing it. According to a Bitkom survey conducted in March 2026, 41 per cent of companies in Germany with 20 or more employees now use artificial intelligence, whilst a further 48 per cent are planning to use AI or are discussing its implementation. Customer service is one of the areas where the benefits of AI are most quickly apparent, because the volume of enquiries is measurable and a large proportion of customer enquiries are recurring. How does AI work in customer service? In customer service, two types of AI tools are used, which are often lumped together. AI chatbots process text – in website chats, via messaging apps or by email. An AI voicebot processes spoken language, usually over the telephone. Both AI tools can access the same data, but the voicebot solves the problem that most support teams actually face: the telephone helpline. Technically, several components work together in an AI voicebot: Real-time speech recognition (speech-to-text) converts spoken words into text whilst the caller is still speaking; an LLM-powered dialogue control system identifies the issue at hand and determines the next step in the conversation; a workflow engine carries out the actual task: checking status, completing a task, updating data; and speech synthesis (text-to-speech) returns the response in natural language. AI voicebot versus traditional IVR system The key difference between AI and a traditional IVR system based on the „Press 1 for …“ model lies not in the voice, but in the freedom of input. An AI voicebot allows free-form voice input, recognises intentions and context, and works with customisable workflows. The traditional system guides users through a rigid menu structure, operates on a rule-based system and offers only limited customisation. For customers, this makes the difference between getting direct answers and experiencing frustration due to lengthy navigation. Why system integration is key to success An AI is only as good as the data it is allowed to access. A stand-alone solution can provide information, but cannot carry out any actions. Only integration with the core system – such as an ERP system like Microsoft Dynamics 365 Business Central – enables the AI to recognise the caller by their phone number during the conversation, assign an order status, retrieve invoicing details or view master data. What starts as a simple enquiry then becomes a completed transaction. Case study: 200 calls a day – and what AI made of them The starting point With over 1,000 vehicles in more than 40 towns and cities, Carl and Carla are among the leading providers of flexible van hire in Germany. As the business grew, so did the strain on customer service: particularly at weekends and outside business hours, the volume of calls rose dramatically, reaching up to 200 calls a day. Most customer enquiries were recurring: return procedures, vehicle usage and general queries. The service team was constantly stretched to the limit; genuine 24/7 availability was unfeasible, and waiting times and operating costs were rising. The AI solution Rather than automating the entire customer service process, a single procedure was singled out: vehicle return, one of the most common and time-sensitive use cases. The AI service agent ILAI identifies callers via their telephone number or registration number, retrieves location-based vehicle data – such as GPS position and fuel level – and guides them through the return process via natural language dialogue. Once completed, ILAI can even lock the vehicle remotely. The result: within six weeks, around 90 per cent of return processes were fully automated. The average call duration was halved, and the cost per call fell to less than one euro. Since then, customer service has been available round the clock, in multiple languages and without overburdening the team. „Thanks to the AI-based call centre agent, our customers can now complete the vehicle return process in just a few minutes and with absolutely no waiting time. For our team, this takes a huge load off their shoulders, as routine enquiries are handled automatically and reliably.“ Gregor Wendt, CTO at Carl and Carla. The lesson to be learnt from this AI project is less about the percentage figures and more about the specific nature of the task: the initiative was successful because a clearly defined, high-frequency process was selected – rather than customer service as a whole. Where AI delivers the greatest benefits in customer service The proven areas of application for AI vary more according to the type of process than by sector. Customer enquiries that occur frequently, are clearly structured and require system access are particularly well suited: E-commerce and wholesale: placing orders, initiating returns, checking delivery status; Production and manufacturing: helpdesk for internal IT enquiries, supplier communication; Healthcare: booking appointments and responding to frequent patient enquiries – in other words, administrative processes, not medical advice. Four effects in everyday practice: Availability without waiting times. AI handles enquiries even outside business hours, without a shift rota and without a hold queue. Reduced workload for routine tasks. Recurring processes tie up fewer staff, leaving more time for complex issues. This is particularly relevant where service centres are difficult to staff. Parallel processing during peak periods. An AI can handle several conversations simultaneously, so that peak loads at weekends or following a campaign no longer automatically lead to long waiting times. Consistency and traceability. Every process follows the same steps and is documented. This facilitates evaluation and quality assurance and measurably improves service quality. What AI cannot do in customer service AI does not replace a service team, and providers who promise this are selling an expectation that does not hold up in practice. Three limitations are important in practice. Emotional and exceptional issues require human intervention. Complaints, claims and exceptional situations: here, sound judgement counts, not strict adherence to processes. A well-designed AI system recognises this and escalates the matter rather than attempting to handle it itself. A smooth handover is a necessity, not a convenience. When the AI hands over, the context of the conversation must be passed on. Otherwise, the customer has to start from scratch – and the result is worse than without automation. Without access to data, it remains merely an information service. Anyone who fails to integrate AI tools with their core systems ends up with a better voice response system, not true automation. What you should bear in mind when introducing AI: Consider data protection from the outset. What matters is not the ‘GDPR-compliant’ label, but what the AI actually does: In which data centre is the processing carried out? What data is actually collected – are a telephone number and reference number sufficient, or is a profile being created? Are the actions of the

Regulated industries: Introducing AI and cloud in compliance

AI and the cloud offer regulated companies real opportunities - but only if governance, data quality and the operating model are considered from the outset. What pharma, biotech and medical technology companies need to consider when introducing AI.