This week, we dive into the world of AI agents.
While generative AI has already transformed our usage patterns, AI agents represent the next step: more than just a chatbot or an automated workflow, they learn and adapt in real time.

But how do they work? What are their concrete use cases in businesses? And most importantly, what role does the human still play in this equation?

What Is an AI Agent?
Unlike automated workflows that follow predefined rules, AI agents are designed to handle more complex and evolving tasks. While a workflow executes a fixed set of actions (e.g., an automatic refund in three steps), an AI agent can:

• Analyze various information sources to make a decision,
• Interact in real time with a user,
• Adapt to new scenarios without specific programming.

Example: Instead of a bot that simply executes a refund policy, an AI agent can identify exceptions, engage in conversation with the customer, and propose a personalized solution.

5 Use Cases of AI Agents in Businesses
Major companies are already deploying these agents for various tasks:

1. Pharmaceutical : Accelerating drug discovery with AI agents that optimize chemical synthesis processes.
2. Banking : 35 AI agents dedicated to financial analysis, facilitating the review of complex regulatory documents.
3. Ecommerce : Automating the creation of marketing campaigns and code writing through collaborative AI agents.
4. Telecom : Implementing an intelligent internal assistant (“askT”) to help employees navigate HR policies.
5. Services : AI agents handling part of customer service, reducing the need for human intervention in standardized requests.

What Is the ROI for Businesses?
Companies adopting these technologies see productivity gains, better personalization of customer experiences, and cost optimization.

However, the widespread adoption of AI agents also raises challenges:

• Reliability and bias: How can we ensure that the agent’s decisions are relevant and fair?
• Cybersecurity: Gartner predicts that by 2028, 25% of security breaches will be due to AI agent-related abuses.
• Impact on jobs: While some roles are automated, new ones emerge to supervise and train these models.

Humans Remain Essential
At isahit, we believe that the future of AI agents lies in a smart hybridization between machines and humans. Through approaches like Reinforcement Learning from Human Feedback (RLHF) and Human-in-the-Loop (HITL), we enable companies to:

• Compare an AI agent’s performance before and after deployment,
• Ensure quality control over its responses,
• Test different configurations via supervised A/B testing.

Far from entirely replacing human intervention, these new AI agents redefine roles and require an approach where humans remain at the heart of control and optimization. The challenge is to integrate these tools wisely to maximize their impact while ensuring a smooth and ethical user experience.

Want to learn more?
Visit isahit.com/agentic-ai!

‍