Understanding Agentic AI: Innovation Exec and Ford Scientists Share Tips

The generative AI trend based on large language models (LLMs) like ChatGPT has for years dominated the conversation about artificial intelligence. When something like agentic AI enters the conversation it’s easy to conflate terms, confuse agentic with generative, and not get the full picture.

On this episode of Production Pulse we spoke with Ron Norris, former Director of Innovation at Georgia Pacific, and Nagadithya Nookala, product manager for data analytics and AI-related products and Sanjay Ahire, data scientist, both from Ford Motor Company.

AI Agents Versus Agentic AI

Nookala explains the difference between AI agents and agentic AI using the metaphor of a Google search. If you ask Google about a topic it provides you with links and materials, after which you’re on your own. Agentic AI would not only identify the links and materials but also digest the material to provide summaries, recommendations, essentially do the research and draw the appropriate conclusions for you.

Ahire uses the metaphor of GPS systems used for navigation. Google maps provides directions on how to get somewhere. Agentic AI would not only identify the best route, it would also drive the car itself, using multiple data streams for guidance and navigation.

Norris says the ability to learn differentiates agentic AI from regular agents. Agentic AI not only suggests action but also understands cause and effect and can therefore explain why to take an action.

Are You Ready for Agentic?

Companies that adopted generative AI have already purchased software, on-prem or cloud servers and other infrastructure to serve the new technology. Do those investments also prepare companies to adopt agentic AI tools?

The answer, thankfully, is yes you’ve invested in compute and software to support generative AI you have a leg up on adopting agentic AI. Agent development kits can integrate with existing systems and applications. You may need more stringent guardrails, however, to make sure humans always provide final authority before giving agentic AI permission to act.