The new shiny toy are AI agents. Reminds me of the LLM hype from about two years ago. Many ‘experts’ are now touting AI agents to be the endgame in AI, just like they were touting LLMs to be the rise of sentient machines.
I was amused to read such an article claiming the rise of unicorn AI startups with no employees! If you believe that, I have a bridge to sell in NYC. Contact me for further details 🙂
Don’t get me wrong. I too am very bullish about AI agents, but only if the agentic workflow and supporting cognitive architecture is implemented correctly. The additional challenge would be to neatly integrate the AI agents, workflow and cognitive architecture into the existing enterprise architecture. All this will not be easier even for the folks who are willing to start afresh.
If your current enterprise architecture is based on microservices, then an easy hook would be to integrate the existing APIs and microservices using function calling or as ‘tools’ for the AI agents. At the very least, this will get you started on integrating the agentic workflow in your enterprise.
What would be the best place to start with AI agents then? Glad you asked.
Nightly batch jobs or Cron jobs would be a good place to start. In fact, any task that runs unattended can be optimized and enhanced with AI agents. Other activities that require data and information gathering for decision making – e.g., support ticket triaging – are also great fit for AI agents and agentic workflow. The trick, as always, is to start small and gradually increase the footprint of the solution.