The consultancy Capgemini argues that companies should prepare to deploy AI agents, as this technology will soon be available in nearly every business function where it can be applied. According to its research, 82% of organisations expect to integrate AI agents into their operations by 2027.
In its report Multi-agent AI – 21st century automation revolution, Capgemini recommends combining human empathy with technological efficiency to excel in areas such as customer service. This entails using multiple communication channels and interaction styles to engage different consumer segments. AI agents make this possible by drafting automated responses to user inquiries based on historical interaction data.
An artificial intelligence (AI) agent, the report notes, is a software application that interacts with its environment, gathers data and uses them autonomously. “As an evolution from technologies like robotic process automation (RPA) and machine learning (ML), AI agents can perceive, reason and act in changing environments to achieve their goals. How they reach them is largely left to them to decide.”
To interact with users and complete tasks independently, AI agents often rely on large language models (LLMs) as their interface. Trained on vast amounts of data, LLMs encode knowledge through pattern recognition, enabling AI agents to reason, make decisions and communicate effectively.
Capgemini identifies six core traits that distinguish AI agents:
- Autonomous
- Goal-oriented
- Context-aware, using relevant data to make decisions
- Adaptive, adjusting behaviour and responses as data or interactions change
- Proactive, initiating action independently without user prompts
- Language-aware, interpreting and responding in human language
Common use cases for AI agents include customer service, predictive analytics, real-time monitoring, task and inventory management, scheduling, orchestration in manufacturing, content recommendation and fraud detection.
Autonomy in problem resolution
Recent advances mean AI agents can now handle customer incidents on their own. As Capgemini points out, chatbot refund requests used to be escalated to a service representative. Today, an AI agent can request proof of purchase or a photo of the product, analyse the case and propose a solution — all without human involvement.