The Challenge
Many teams want to deploy AI agents in Salesforce — but quickly hit a wall: "We don't actually know how our processes really work, or how we can make agents work on top of all of this complexity."
Without process clarity, the risks are significant:
Agents automate broken, inconsistent, or undocumented processes
AI instructions become vague, over-scoped, or unsafe
Leadership invests in AI pilots that stall due to lack of structural clarity
Agents create compliance, data, or trust issues instead of driving efficiency
AI readiness isn't a model problem — it's a process clarity problem.
Getting Started: Connect Your Salesforce Org
Before you can assess AI readiness, you need an accurate view of your org's metadata.
Feature: Salesforce Sync (Salesforce Connect)
Support: Connect and Sync Salesforce Org
💡 Important: The metadata dictionary is only as accurate as the latest sync. Always confirm sync status before analysing metadata.
Step-by-Step: Preparing Your Org for AI Agents
Step 1 — Understand the Real Process
Most processes suffer from automation overlap and overprovisioning. Before building agents, consider optimising the process — locking permissions, removing automation sprawl, and improving data quality.
Step 2 — Identify Agent Opportunities
Using the Agent Finder, analyse your process to identify all automation and agentification opportunities.
Review findings with business stakeholders, agree on expected business logic, and formally approve required automations and agents.
Step 3 — Design Agent Architecture
For identified agents, create an Agent Interaction Map to architect expected logic and interaction boundaries.
For AI workflows, create an Interaction Flow Diagram to design the expected logic.
Step 4 — Design Agent Logic
Capture an Agent Instruction Diagram to map out expected AI reasoning logic for each agent topic or prompt template.
Use Agent Checker to review and improve instruction design, catching patterns and anti-patterns before build.
Step 5 — Get Agent Design Formally Approved
Get your agent design formally approved by the business before moving to build. This creates a clear audit trail and ensures stakeholder alignment.
What You'll Learn Along the Way
Going through this process shifts your team from "Let's just build an agent" to a much deeper understanding:
We don't fully understand our own business processes at the configuration level
There are automation overlaps we didn't know about
Some steps are deterministic — AI is unnecessary and adds risk
Some steps are high-friction and perfect for agentification
Agents need explicit interaction boundaries to prevent overreach
Good agent instructions are structured — not conversational
AI readiness is a process clarity problem, not a model problem
What You Can Do Next
Once you've completed this process, your team will be able to:
Prioritise AI use cases based on actual process friction, not hype
Fix broken or overly complex automation before agent rollout
Define clear human-in-the-loop checkpoints
Create structured, safe agent instruction frameworks
Prevent agent overreach with explicit guardrails
Present a justified, ROI-backed AI roadmap to leadership
Move from "AI experiment" to a sustainable AI operating model