Super Data Science: ML & AI Podcast with Jon Krohn · Jon Krohn

956: From Agent Demo to Enterprise Product (with Ease!) feat. Salesforce’s Tyler Carlson

·26 min·2 clips

Building AI agents is easy for demos, but 90% of developers fail at production.

1. Super Data Science: ML & AI Podcast with Jon Krohn focuses on Salesforce's push to move AI agents from demo to enterprise product. 2. Jon Krohn hosts and Tyler Carlson, Salesforce SVP and head of product for AppExchange and Ecosystem, explains why his platform matters for builders. 3. The episode asks what AgentForce 360 is for, and Tyler frames it as a way to ship secure, reliable agentic applications beyond prototype stage. 4. Tyler says AppExchange launched 20 years ago as the first enterprise app store for a SaaS product. 5. He recounts Mark Benioff's story about meeting Steve Jobs and hearing advice to build an app economy around Salesforce. 6. Tyler says Salesforce keeps building an ecosystem of ISVs and developers that extend Salesforce for 150,000-plus customers. 7. He explains that AppExchange includes apps for Sales Cloud, Service Cloud, Marketing Cloud, Commerce Cloud, contract management, and pricing tools. 8. Tyler says some partners build verticalized apps for industries and use Salesforce rails to serve niche workflows. 9. He describes Salesforce's platform as enterprise-grade infrastructure that powers sales, service, marketing, and commerce at massive scale. 10. Tyler says the platform offers low-code tools like WYSIWYG schema building, Flow, drag-and-drop UI customization, Apex, and React. 11. He describes AgentForce 360 as bringing a multi-LLM planner service, deterministic scripting, and Salesforce's data and metadata context into one enterprise platform. 12. Tyler says the platform adds security, permissions, observability, and monitoring so agents can be deployed and iterated on after launch. 13. He says MCP-style access to data still needs enterprise security and context controls. 14. Tyler argues that natural-language rules like "always do this" and "never do this" are not enough for enterprise agents. 15. He says Salesforce built AgentForce Script to handle deterministic logic when LLMs are not reliable enough. 16. Tyler says the last mile from prototype or V1 to production grade is harder than building the prototype itself. 17. The conversation stays technical but conversational, with Jon asking practical follow-ups about app building, enterprise deployment, and commercialization. 18. Tyler speaks in a product-focused, explanatory style that moves from platform history to current agent tooling. 19. Builders of enterprise AI products and SaaS founders using agents will get the most value from this episode. 20. Listeners looking for pure consumer AI commentary or entertainment may skip it.

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from Super Data Science: ML & AI Podcast with Jon Krohn