Here’s a simple practical example of how an AI Agent (LLM Agent) works.\n\nProvide ChatGPT with Digital Workforce’s Q3 analysis by Inderes as an attachment.\n\nAsk ChatGPT the question: “Can you read the company’s revenue forecast for 2024 (2024e) from the attached file? After that, can you find on the company’s website what their revenue forecast for 2024 is?”\n\nChatGPT will retrieve Inderes’ forecast and then search DWF’s website for their guidance and report it.\n\nThat is the foundation of an AI Agent. In sema4.ai’s settings, you can specify which language model it uses (e.g., ChatGPT’s GPT-4o).\n\nIn addition to the above, an AI Agent internally calls the LLM multiple times and attempts to break down the task into logical components before execution. Finally, it calls functions (actions, e.g., retrieving customer data from CRM) to complete the task.\n\nOf course, sema4.ai has much more, but as a concept, an AI Agent means just that. There’s nothing exotic about it, and it’s just a logical extension to leverage the continuously evolving reasoning capabilities of current LLM models.\n\nThe question is mainly whether enough useful applications for AI Agents will be found, as has been predicted.\nAnother question is how reliable a probability-based AI Agent is. RPA, based on unambiguous conditions, is naturally reliable.\n\nDWF’s Karli Kalpala seems to be continuously writing articles about AI Agents for international publications. Here for the insurance industry.\n\nPresumably, this is part of the sales strategy.\n\nI also listened to a podcast where DWF’s AI Lead Data Scientist Rami Luisto talked about his tasks at DWF. It’s about 2 hours long and a bit tedious. A pre-planned agenda would have been useful.\nBefore AI Agents, DWF used AI by employing lightweight LLM models that RPA would call when needed for simple text analysis.\n\nDWF has finally turned its results positive this year. Let’s hope they don’t forget business realities in the AI hype and that profit growth continues.\nImplementing AI Agents should not, however, require unreasonable investments.
12 Likes