AI Agents are piece of software that can act to events or act on its own to take several actions which are normally done by humans. For example, a customer support agent who chats with customer and issues a refund. AI agents are old but modern Gen AI has enabled us to create much smarter agents and especially those who can directly interface with real world humans.

Google demonstrated its impressive abilities few years ago where Google assistant could make a call to a restaurant and reserve a table by talking to an employee there. This feature is now available for users in USA and is reported to have been working pretty well. There is now a gold rush to build AI agents that can do more. This includes shopping on your behalf, investing on your behalf, managing your budget and finances etc.

Banks are looking at AI agents seriously. The idea is to create a shadow employee for every real employee in the system. With time key components and be handed off to AI permanently. Customer service and T1 support is the easiest and generally the first set of jobs that are handed over to AI Agents.

Very likely we will see AI agents being deployed in business environment first where businesses will expand their footprint with the help of AI agents. Bank branches with few employees but full services. AI concierge services in hotels that can help customers far quickly than real people.

This will also open an opportunity for existing businesses to open up their APIs for other AI agents to use. What could be a good interface for AI agents to interact with ? Should it be chat based ? Or should it be more formal with REST api returning fixed JSON responses ?

What we are going to see is more Agent2Agent communication that leads to actions being taken including payments.

Advantages

1. 24×7 availability

2. Higher sales

3. Dynamic pricing and superior inventory management

4. Improved fraud detection