AI Agents in Customer Service: Introduction
The introduction of AI agents in customer service holds the potential to automate routine inquiries, but also the concern that humans could be replaced. Sinch's forecasts indicate that AI agents and Voice AI can increase the volume of digital customer interactions three- to fivefold by 2026. Companies such as Essity are launching programs to anchor AI agents in core processes. The central question is how benefits and risks can be kept in view, rather than being guided by hype.
Source: Custom Representation
Fundamentals of AI Agents
AI agents in customer service are more than traditional chatbots. They are software-based 'team members' who, with the help of large language models, can understand text and speech, make decisions, and perform actions in other systems, such as creating tickets, triggering orders, or writing data into the CRM. Microsoft beschreibt sie als virtuelle Teamkollegen.
Unlike earlier bots that worked strictly by rules, agentive AI systems combine multiple capabilities: they plan multi-step workflows, access internal knowledge sources, call tools such as databases or APIs, and know when to hand off to humans. Microsoft bezeichnet KI-Agenten als virtuelle Teamkollegen, that work with humans.
Concrete examples in customer service include digital frontline agents that handle common inquiries about invoices, delivery status, or password resets, and only escalate complex cases to humans. Internally, IT helpdesk assistants look up policies, prioritize tickets, and propose solutions, or back-office agents that read contract data from PDFs and transfer it into ERP systems. Diese Beispiele zeigen die Bandbreite der Anwendungen.
Platforms such as Microsoft Copilot Studio offer building blocks to configure dialog-oriented agents, autonomous process agents, and voice agents, connect them to data sources, and deploy them across channels such as Webchat, Microsoft Teams, or telephony. This is largely done via configuration and low-code.
Source: agorum.com
AI Agents: A look at how the intelligent helpers in customer service work.
Source: YouTube
Current Status and Developments
The latest figures show strong momentum in the AI agents space. Sinch, ein Anbieter für Kommunikationsplattformen, They handle over 900 billion interactions per year and forecast that AI agents and Voice AI could lead to a three- to fivefold increase in global messaging traffic by 2026. This enables more customer contacts to be conducted economically.
Voice AI plays an increasingly important role in this. Moderne Sprachsysteme reagieren mit einer Latenz von rund 800 Millisekunden, which makes complex telephone conversations practically feasible. Prognosen von Marktforschern indicate that by 2026 around three-quarters of customer service interactions could be supported or fully taken over by AI-based voice agents.
In parallel, Essity, ein globaler Anbieter von Hygiene- und Gesundheitsprodukten, In collaboration with Accenture and Microsoft, a cloud-based platform based on Azure, Copilot Studio and Power Platform is being implemented. AI agents will first be tested and optimized in procurement and finance processes, in order to roll them out gradually to further core processes. The platform will be operated under a Responsible AI framework.
Technologically, it has gained new capabilities, Microsoft Copilot Studio such as Computer Use, which allows agents to interact directly with desktop applications and websites, even if no APIs are available. This enables the automation of back-office activities such as invoice processing.
The current „State of AI 2025“-Studie von McKinsey, shows that 88 percent of companies use AI in at least one area of their business. In agentic AI, 23 percent scale up corresponding systems, another 39 percent experiment with it. However, less than 10 percent of companies have rolled out AI agents broadly, indicating an experimentation phase.
In customer service, examples range from augmentation strategies to automation programs. Verizon setzt auf ein KI-System, which supports service staff in real time, leading to shorter conversation times and an increase in sales performance. Salesforce hat 4.000 Support-Jobs abgebaut and replaced by AI agents.
Roland Berger kommt zu dem Ergebnis, that a substantial portion of standard tasks in customer service can be automated by AI, while complex cases remain with human agents.
Motives and Interests
The push toward AI agents is driven by several motives. Technology providers such as Microsoft, Salesforce, ServiceNow, or Sinch pursue a clear platform strategy. They want ganze Agenten-Ökosysteme anbieten, in which companies can obtain everything from data connectivity to governance from a single source. Sinch betont, dass KI-Agenten sich von reinen Kostensenkungs-Werkzeugen zu Wachstumsmotoren entwickeln sollen.
For companies, there is often a twofold strategy: reduce costs and enable new revenue. McKinsey berichtet, that companies with the greatest measurable AI impact not only aim for efficiency but also explicitly want to use AI for growth and innovation. Laut EY Global CPO Survey 80 percent of Chief Procurement Officers plan to use generative AI in their procurement processes. Essitys Start des KI-Agentenprogramms in Beschaffung und Finanzen That aligns with it.
In many industries, a new AI-first, but not AI-only paradigm is taking shape. Eine Analyse zur Zukunft des Customer Experience Managements describes AI agents as the basis of an 'AI-first' customer service, where virtual agents resolve many issues but deliberately collaborate with human staff. The reality is more nuanced, ranging from augmenting tools like those at Verizon to job reductions such as at Salesforce.
Media and platform dynamics shape perception. Success stories spread widely, while implementation problems rarely come into the spotlight. McKinsey weist darauf hin, that only a small portion of companies see a clear impact on results, and many still struggle with risks such as lack of transparency and susceptibility to errors.
Leaders in customer service or back office operate in a tension field between potential, commercial interests and media-amplified success stories. It pays to look beyond headlines and soberly assess what problems AI agents can really solve.
Facts and Myths
Evidence: That AI agents and related technologies massively change the workload in customer service is supported by multiple sources. Sinch erwartet eine drei- bis fünffache Zunahme digitaler Interaktionen bis 2026. McKinsey-Studien zeigen, that almost nine out of ten companies are already using AI. Roland Berger verweist darauf, that standard tasks are increasingly automated. Case studies such as those of Verizon or Salesforce demonstrate measurable effects on conversation duration, revenue, and headcount.
Unclear: Whether AI agents will replace the majority of service jobs in a short time is not substantiated. Die McKinsey-Umfrage zeigt ein gemischtes Bild: A third expect a declining headcount, nearly half expect no major effect, and a small portion even expect increases in staff. Eine Analyse von Roland Berger emphasizes that many activities will be redesigned and new roles will emerge. Die Berichterstattung von AP The call center sector also shows a mixed picture.
False or misleading: The claim that AI agents are 'just a new name for chatbots' does not fit with the technological developments. Microsoft and McKinsey define agentive AI as systems that can autonomously plan, execute multiple steps, and act. Also misleading is the notion that companies that do not automate everything immediately will be left behind. McKinsey betont, that the most successful companies carefully redesign workflows, build governance-capable structures, and manage risks.
Reactions and Counterarguments
Reactions to AI agents in customer service are not uniform. Software vendors paint a positive picture. Sinch spricht von einer „Explosion der Kommunikationsvolumina“. Microsoft beschreibt, how companies can build and control Copilot Studio Agents.
On the corporate side, efficiency gains and better customer experiences are emphasized. Verizon stellt heraus, that AI makes the work of service employees easier. Salesforce rechtfertigt einen drastischen Abbau von Supportstellen with the argument that AI features can take over a large share of interactions.
Criticism comes from employee representatives, ethicists, and consumer advocates. The use of voice-modification AI by the call center operator Teleperformance to neutralize accents is viewed critically. Kritiker fürchten zusätzlichen Druck auf Arbeitskräfte. Policy, such as the US-Kongress, discusses statutory regulations that aim to ensure a right to personal contact with a human agent.
Analysten wie BCG classify AI agents as the beginning of a 'golden era of the Customer Experience', but emphasize that the added value only emerges if companies take governance, monitoring and human supervision seriously.
Practical Implementation
A sensible first step is mapping customer journeys and back-office processes. Where do you spend a lot of time on repetitive, structured tasks? There AI agents are especially strong. Microsofts Szenariobibliotheken für Copilot und Copilot Studio provide examples of how self-service agents can be built.
Second, it pays to leverage the existing technology base. If Microsoft 365, Dynamics 365 or Power Platform are used, AI agents can be integrated into familiar interfaces. Dies gilt auch für andere Ökosysteme, that offer their own agent frameworks.
Third, governance is mandatory. McKinsey-Daten zeigen, that more than half of companies have experienced negative incidents with AI. Essity nennt ein Responsible-AI-Rahmenwerk as guardrails. Clear rules must be defined for when decisions can be made fully automatically and when a human should be involved.
Source: tiq-solutions.de
The path to successful implementation of AI agents in customer service in five steps.
Fourth, the involvement of teams determines success or failure. AI agents change the roles within service. Eine Untersuchung der Callcenter-Branche zeigt, that employees who experience AI as an assistant report positive effects. Employees should be involved early as 'co-designers' of the agents.
Finally, you need clear success criteria. Cost per contact and handling times are important, but also first-contact resolution rate, re-contact rate, and customer satisfaction. Analysten warnen davor, To see AI only as a deflection machine. Langfristig gewinnen diejenigen, to align automation with service quality and trust.
Source: YouTube
Open Questions and Outlook
Despite the many activities, key questions remain open. One is the right degree of autonomy. McKinsey spricht von agentischer KI as systems that act in the real world and can autonomously execute multi-step workflows. But at what point is 'Human in the Loop' indispensable, especially for financially or legally sensitive decisions?
Another open issue is the long-term employment effect. The range spans from scenarios where AI agents replace a large portion of simple service jobs to models where they serve as productivity boosters. Die McKinsey-Studie zeigt, that 32 percent of respondents expect a significant staff reduction due to AI in the next twelve months. Examples such as Salesforce show drastic downsizing, while Verizon shows how AI assistance can be without job cuts. Reliable long-term data is still missing.
Regulatory developments remain dynamic. While some countries consider rights to access to human contacts and apply existing consumer-protection rules to AI-assisted services, companies and standardization bodies are working on best practices for transparency, labeling, and liability for AI agents. Dies erfordert flexible Strategien, that can support regulatory adjustments.
Introducing AI agents in customer service is not a cosmetic IT project, but an intervention at the heart of customer relationships and workflows. The state of play shows clear momentum: communication volumes are rising, Voice AI is becoming everyday usable, and companies such as Sinch, Essity, Verizon or Salesforce are already using agentive AI to achieve real effects on productivity, revenue, and headcount.
The most successful organizations do not treat AI agents as a quick savings lever, but as an opportunity to rethink processes, take governance seriously, and bring their teams along. Dies erfordert eine sorgfältige Planung. Es lohnt sich, klein, fokussiert und transparent zu beginnen: with clearly defined use cases, well-informed employees, and measurable objectives. Then AI agents can move from a fear topic to a true tool.