Agentforce for Sales Concepts

About this course

This course has no product screenshots and no click-by-click walkthroughs. It exists because click-paths stop helping the moment something behaves unexpectedly — the cause is almost never on the screen you were shown, and the screen changes every release anyway.

Instead, it teaches what the Salesforce AI Sales Agents documentation actually says: what each agent does, what it reasons over, what it is documented to be unable to do, and which constraints most often derail a project. Every claim traces back to a help page, and the modules follow the documentation's own structure so you can keep reading it yourself long after this course ages.

The spine is one idea introduced early and reused everywhere: an agent is built from layers — situation classification, actions, grounding, instructions, guardrails — and when output is wrong, the cause sits at exactly one of them. Fluent but factually wrong is grounding. Correct but off-brand is instructions. Answering something it should have declined is classification. Naming the layer is most of the fix, and that skill survives every release.

One full episode teaches you to read a Considerations page, because that page structure repeats across every capability and carries the facts most likely to change: required editions, deployment constraints, supported locales, the model in use, and how you are metered. Learn to read one and you can assess a feature this course never covered.

Gotchas are first-class content, not an appendix. You'll learn why agents shouldn't be deployed through change sets or metadata and what that does to your release process; why turning on manual approval changes what the recipient is told; why testing with colleagues' email addresses produces no replies at all; why enabling group-based meeting routing rules out sending in a rep's name; and the reporting feature that stops Lead Nurturing working entirely.

What you'll learn

• Navigate the AI Sales Agents documentation and know which page answers which question
• Describe all five capability families and where each sits in the funnel
• Break an agent into its layers and diagnose which one caused a bad output
• Read any Considerations page and extract the constraints that affect your project
• Explain how agent usage is metered, and how settings change consumption
• Trace the full Lead Nurturing engagement flow from assignment to disengagement
• Explain how seller-voice sending, manual approval, and AI disclosure interact
• Map the email and calendar dependencies that decide whether an agent works at all
• Name the documented send limits, approval windows, and org-level ceilings
• Identify irreversible decisions before you commit to them
• Apply the published ethical guidance on agent naming, disclosure, and outreach volume
• Reason about which fields an agent should be allowed to update autonomously
• Recognise the six failure patterns behind most agent problems
• Assess whether an org is actually ready to configure, before anyone starts

Requirements

• Working familiarity with Salesforce concepts: objects, records, permission sets, and record ownership
• No Salesforce org needed — this is a concepts course with no hands-on configuration and no product screens
• No AI, prompt engineering, or machine learning background required
• Helpful but not required: exposure to a sales process as a rep, admin, or consultant
• Product details change between releases — the course teaches you to verify current limits in the documentation rather than trusting a recording

Who this course is for

• Salesforce admins and consultants who need to understand these agents before configuring them for a customer
• Solution engineers and architects who must design an agent landscape and defend it in a review
• Sales and revenue operations professionals evaluating whether these agents fit their funnel
• Sales leaders who want the capabilities, constraints, and honest limitations before funding a rollout
• Anyone who followed a click-through tutorial, hit a problem it never mentioned, and wants the model underneath
• Not for people wanting a step-by-step configuration walkthrough — this course deliberately contains no product screens

Curriculum