Agentforce Specialist study plan

Five weeks at 6–8 hours a week, assuming roughly the year of Salesforce platform experience Salesforce expects. You need an org you can build in — a Developer Edition or a sandbox with Agentforce enabled.

The plan follows the weighting: AI Agents, Prompt Engineering and Data 360 are 75% of the exam between them.

Week 1: prompt engineering (20%)

• When Prompt Builder is appropriate given a business requirement — and when an agent action or a flow is the better answer.
• Prompt template types, including field generation and flex templates, and the considerations for each.
• Creating, activating and executing prompt templates. Activation is a distinct step and it is examinable.
• Access controls governing prompt templates — who can see, edit and use them.
• Writing effective prompts: clarity, structure, examples, constraints.
• Grounding techniques — choosing the right one for a scenario.
• The Trust Layer: its security and privacy features, and what it does to data on the way to and from a model.
• Managing and preventing access to specific models.

Week 2: Data 360 fundamentals (20%)

The section most candidates underestimate, because it is conceptual rather than clicky.

• Agentforce Data Library — what it is and its considerations.
• Chunking — splitting content for retrieval, and why chunk size and boundaries affect answer quality.
• Indexing — making content searchable.
• Retrievers — how relevant content is found and returned to ground a response.

These three terms are the vocabulary of the whole section. Be able to explain each in one sentence and say what goes wrong when each is configured badly.

Weeks 3 and 4: AI Agents (35%)

The largest section, two weeks.

Week 3 — how agents work:

• The basic building blocks of agent script, and how an agent reasons and acts.
• Hybrid reasoning: its components and benefits, and how Agent Script functions in Canvas and Script View.
• Standard and custom topics, and standard and custom agent actions — selecting and configuring each for a use case.
• Deterministic behaviour: filters, variables and template expressions, for when an agent must do exactly the same thing every time.

Week 4 — deployment surface and security:

• Channels: digital experience, email, voice and Slack, and how agents connect to each.
• The security context the agent runs in, and how that affects which actions it can execute. This is a recurring exam theme and a common real-world surprise.
• Employee versus Service agents — choosing by audience.
• When to use the Agent API.

Week 5: testing and governance (20%), multi-agent (5%), then practice

Testing, deployment and maintenance:

• Testing an agent in Testing Center, and how its evaluations work.
• Deploying an agent from sandbox to production, and the considerations involved.
• Deploying a prompt template from sandbox to production — treated separately in the outline, so know both.

Governance and observability:

• Managing and monitoring agents in production.
• Agent analytics and optimisation.

Multi-agent orchestration:

• When a Multi Agent architecture is appropriate for scalability and control.
• The purpose of MCP and A2A as open standards.

About three questions. An hour is proportionate.

Then: a full practice exam under real conditions — 105 minutes — mistakes sorted by section, weakest rebuilt, second practice exam.

Where the hours go

You need 44 out of 60

Salesforce publishes a flat 72% passing score, which on 60 scored questions means 44 correct. That is a concrete target, and it makes practice scores directly meaningful — unlike scaled exams where you cannot convert.

Target the mid-80s in practice to leave margin for question difficulty.

Build one agent end to end

The most efficient preparation: create a prompt template and activate it, ground an agent using a Data Library, build the agent with a custom topic and a custom action, add a filter for deterministic behaviour, test it in Testing Center, and deploy it to another org.

That sequence touches every section on the exam, and Salesforce writes questions from the perspective of someone who has done exactly that.

Check the release. Questions align to Spring ‘26 — material written against an earlier release may describe a different Agent Builder.