Speakers
AI agents can do far more than answer questions about Salesforce data – but connecting an AI model to your org is only the beginning. The real challenge is giving it the right context, keeping its answers reliable, and making sure access to sensitive Salesforce data remains secure.
Join us for a practical workshop exploring how one sales team built an AI sales agent on top of Salesforce, the mistakes they uncovered along the way, and four rules you can apply when building AI into your own org.
What We’ll Cover:
- Start With Read-Only Access: Learn why the team deliberately limited its AI agent’s Salesforce access, and how least privilege, source-level permissions, audit logs, and human approval can provide additional safeguards.
- Teach AI How Your Salesforce Org Actually Works: See how seemingly obvious fields and record types produced confident but incorrect answers – and how a plain-English guide to the org helped the agent understand which data to trust.
- Make Salesforce the Data Spine: Explore how connecting calls, contracts, pricing, and deal-desk requests back to Salesforce gave the agent the context to support a real pricing workflow from opportunity through approval.
- From Pricing Question to Useful Answer: See the agent work through a real deal-desk scenario using account history, calls, contracts, and previous decisions to recommend pricing with its reasoning and sources – plus how the same context can support proposals, follow-ups, renewals, and more.
- Treat Free-Text Fields as Untrusted: Discover how data entered into Salesforce can become a prompt-injection risk, including a real internal test, and the practical controls admins can put in place to reduce exposure.
- What We Learned After a Year in Production: Hear the unexpected lessons uncovered by connecting the full workflow, including how the team discovered that packaging – rather than pricing itself – was creating problems, alongside measurable results from the system where available.
Walk away with four practical rules for connecting AI to Salesforce safely and effectively, plus a real-world framework for building an agent that understands the context behind your data – not just the record in front of it.
