In this episode of the Australian Finance Podcast, Owen Rask speaks with Jon Stahl from Salesforce about how AI agents are moving beyond answering questions and starting to complete real work inside businesses.
Jon explains the difference between a chatbot and an agent, including how agents can reason through multistep tasks, call tools and APIs, update business systems and retain useful context. He shares where the technology is already creating value across customer service, software development, research, order management and commerce.
The conversation also tackles the controls businesses need before giving an AI agent meaningful autonomy. Jon compares an agent to the “world’s smartest but least experienced intern” and explains why permissions, clear identities, audit logs, hard-coded guardrails and human escalation remain essential.
Owen and Jon then explore what must happen before agents can safely buy products or make payments on our behalf. That includes verified authority, spending limits, tokenised payment credentials, merchant recognition and common standards across the payments ecosystem — challenges that cannot be solved by a smarter language model alone.
Finally, Jon separates the hype from the practical opportunity. Fully autonomous shopping is still largely experimental, while brand, trust and storefronts remain important. He also explains why trusted business data can dramatically improve an agent’s usefulness, and how businesses may soon use agents to resolve delayed orders, enrich product catalogues, support sales teams and help employees build their own tools.
