Salesforce AI certification path
The Agentforce Specialist exam assumes you already know Salesforce. It is not an AI exam that happens to run on Salesforce; it is a Salesforce exam about building agents, and the platform knowledge is the harder prerequisite for most candidates.
What the exam actually weights
| Section | Weight |
|---|---|
| AI Agents | 35% |
| Prompt Engineering | 20% |
| Data 360 Fundamentals | 20% |
| Testing, Deployment, and Maintenance | 10% |
| Governance and Observability | 10% |
| Multi-Agent Orchestration | 5% |
Over a third of the score is agent construction: topics, actions, scope, escalation and how the security model constrains what an agent can reach. Another fifth is data — identity resolution, harmonisation, ingestion — which is where people from an AI background usually lose marks.
Who should take it
Salesforce administrators, consultants and developers whose organisations are adopting Agentforce. You already have the platform model, the security model and the data model; the agent layer is the new part. That is a four-to-six week exam.
Salesforce partners who need the credential for practice requirements. Same reasoning.
Who should not
People coming from general AI work with no Salesforce background. The prompt engineering section will be easy and the rest will not. Sharing rules, record access, object relationships and the Data 360 model are assumed, not taught, and they carry more weight than the AI content.
If you understand LLMs well but have never configured a Salesforce permission set, the exam will punish the gap you did not expect.
The part that catches people out
Agent actions run within the platform’s security model. Questions repeatedly test whether you understand that an agent cannot return data the executing context is not entitled to see, and that instructing a model to respect permissions is not a control.
That principle — enforce entitlement at retrieval, not in the prompt — recurs across the exam and generalises well beyond Salesforce. It is worth internalising rather than memorising.
A realistic preparation plan
- Weeks 1–2: agent construction. Topics, classification, actions, scope, escalation paths. This is 35% of the exam and where the depth is.
- Week 3: Data 360. Identity resolution, mapping to a harmonised model, ingestion modes. Do not skip this because it sounds like plumbing — it is a fifth of the score.
- Week 4: prompt templates and grounding, then testing, deployment, governance and orchestration. These four sections together are 45%, but they are shallower.
- Week 5: practice tests, then back to whichever section scored worst.
Hands-on time in an org matters more than reading here. Building one agent end to end, including an action that writes data and an escalation to a human, teaches more than any summary.
How long it takes
Four to six weeks for someone working with Salesforce daily. Considerably longer if the platform itself is new to you — and in that case, learning Salesforce first is the faster route overall.
What to do next
Read the Agentforce Specialist exam format guide, then take the free 20-question practice test in that section. If the Data 360 questions go badly, start there — it is the section most people underestimate.
Building agents on another platform too? See AB-620 vs Agentforce Specialist. The exam changes tracker lists the current version of each exam.