FEAR & GREED | Business News Q+A: ChatGPT was just the beginning. Now the real disruption starts.
Mar 3, 2026
Sanjana Parulekar, SVP of Product Marketing for Agentforce at Salesforce, who builds agentic AI for enterprises. She explains how AI has moved beyond LLMs into agentic systems that combine business logic, data and workflows. They cover balancing creativity with correctness, solving the last-mile to production, and how agents will become ordinary tools in a few years.
AI Snips
Chapters
Transcript
Episode notes
LLMs Are Raw Intelligence Not Business Logic
- LLMs provide raw intelligence but lack company-specific context and workflows.
- Sanjana Parulekar explains that only a company's systems and staff know processes like returns, so you must combine LLMs with business logic.
Combine Creativity With Deterministic Workflows
- Design agents that combine LLM creativity with deterministic workflows for critical tasks.
- Use Agentforce to route order returns or regulated financial tasks to workflow logic to ensure correctness, not chance.
Agentic AI Marries Chatbot Predictability And LLM Dynamism
- Agentic AI blends chatbot predictability with LLM dynamism to cover both simple and complex cases.
- Sanjana notes chatbots deflect predictable queries while LLMs add conversational breadth, so agents marry both strengths.