Salesforce Agentforce Specialist Practice Test | 300+ Free Questions

Salesforce Agentforce Specialist Practice Test: Current Spring ’26 Preparation

The Salesforce Agentforce Specialist practice test helps you prepare for questions about prompt templates, trusted business data, AI agents, testing, deployment, monitoring, and multi-agent solutions. This page turns the official Spring ’26 objectives into plain-language lessons, original questions, and hands-on activities. You can also Access free exam practice questions while building a study routine that fits your current Salesforce experience.

Quick answer: The Salesforce Certified Agentforce Specialist exam validates your ability to design and implement prompt templates and agents on the Salesforce Platform. The current Spring ’26 exam contains 60 scored multiple-choice questions and up to five unscored questions. You have 105 minutes, and the passing score is 72%. AI Agents is the largest domain at 35%.

What Is the Salesforce Agentforce Specialist Certification?

The Salesforce Certified Agentforce Specialist credential is designed for people who build and manage AI agent solutions in Salesforce. Salesforce describes certified specialists as professionals who manage and optimize Agentforce and understand both Salesforce platform configuration and Agentforce capabilities.

The exam is especially relevant to Salesforce administrators, developers, and architects who work with Prompt Builder, Agentforce Builder, Data 360, Testing Center, security controls, deployment processes, and agent monitoring. Some registration pages and candidate searches refer to the exam as AI-201, but the complete official credential name is Salesforce Certified Agentforce Specialist.

This is not mainly an AI mathematics or machine-learning exam. Salesforce says candidates are not expected to understand transformer architecture, perform advanced large language model fine-tuning, or possess deep knowledge of external AI tools. The emphasis is on selecting and configuring Salesforce features to meet a business requirement safely and effectively.

According to the current Salesforce Agentforce Specialist exam guide, a successful candidate typically has about one year of experience with Salesforce platform configuration and standard objects, including Data 360. Hands-on experience with Agentforce Builder, Prompt Builder, Testing Center, and sandbox-to-production deployment is also useful. This experience is recommended, not a formal registration prerequisite.

Who Should Take the Agentforce Specialist Exam?

This credential may suit you if you already understand basic Salesforce administration and want to prove that you can turn an AI-agent idea into a governed production solution. Typical candidates include:

• Salesforce administrators who configure agents, permissions, prompts, and automation
• Developers who connect agents to actions, APIs, and other systems
• Solution architects who design secure agent, data, and integration patterns
• Consultants who translate business requirements into Agentforce solutions
• Platform App Builders who want to expand into agentic automation

There is no required certification before you register. Salesforce does, however, recommend the Platform Administrator and Platform App Builder credentials as useful related preparation. If you are completely new to Salesforce, first learn objects, records, relationships, permissions, Flow, sandboxes, and change management. Agentforce questions often assume that foundation.

Current Salesforce Agentforce Specialist Exam Outline

The Spring ’26 blueprint has six weighted domains:

The first three domains represent 75% of the exam, but the smaller sections still matter. Governance, Testing Center, deployment, MCP, A2A, or SOMA knowledge can separate a passing result from a near miss.

Many older resources show five sections such as Data Cloud for Agentforce, Agent Development Lifecycle, and Multi-Agent Interoperability. Those outlines do not reflect the current six-domain Spring ’26 guide. Use the live official certification exam guide as your source of truth.

Domain 1: AI Agents (35%)

AI Agents is the largest exam domain. It measures whether you understand how an agent receives a request, reasons about the request, selects appropriate capabilities, acts, and returns a useful response.

Understand Agent Script and hybrid reasoning

The current outline includes the basic building blocks of Agent Script and the benefits of hybrid reasoning. Agent Script gives builders a structured way to define agent behavior. You should understand the relationship between the visual Canvas view and the more direct Script view.

Hybrid reasoning combines flexible language-model reasoning with deterministic control. In simple terms, the agent can interpret natural language while filters, variables, template expressions, and defined paths keep important behavior predictable. Exam scenarios may ask how to reduce ambiguity, control routing, or apply a business rule consistently.

Configure subagents and actions

Official objectives still refer to standard and custom topics in some places. Salesforce documentation notes that, beginning in April 2026, agent topics are called subagents, with no functional change during the terminology transition. You may therefore see both words in training material and interfaces.

A subagent describes a job or area the agent can handle. An action gives the agent an operation it can perform. Actions can retrieve information, run automation, update a record, or connect to another capability. Practice deciding whether a requirement needs clearer instructions, a new subagent, a standard action, or a custom action.

Control deterministic behavior

Study filters, variables, and template expressions. These mechanisms help the agent make consistent decisions based on known values. For example, a filter may make an action available only for a particular customer type, while a variable can preserve information needed later in the interaction.

Connect agents to channels

The blueprint includes digital experiences, email, voice, and Slack. A channel is where the user interacts with the agent. Channel choice affects identity, context, security, handoff, and experience. Learn the general connection process and the business reason for selecting each channel.

Understand the agent’s security context

An agent does not automatically receive unlimited access. The security context in which it runs affects which records and actions are available. A well-designed agent should not expose data or perform operations beyond the permissions that apply to its runtime context.

Choose an Employee or Service agent

An Employee agent supports internal users and workplace tasks. A Service agent handles customer-service interactions and service processes. Identify the intended user, channel, records, escalation path, and actions before selecting the agent type.

Know when to use Agent API

Agent API is relevant when an external system or custom application must interact programmatically with an agent. The exam is more likely to ask when the API fits a scenario than to require lengthy code.

Domain 2: Prompt Engineering (20%)

Prompt engineering in Salesforce means more than writing a clever sentence. You must select the right template, provide approved context, control access, activate the template, and consider privacy and security.

Decide when to use Prompt Builder

Prompt Builder is appropriate when a business needs a reusable prompt template that can combine instructions with Salesforce data or other grounding sources. It provides a governed approach that is easier to manage than letting every user type an unrestricted prompt.

Compare prompt template types

The current outline specifically mentions field generation and flex templates. A field generation template is useful when generated content should populate a supported record field. A flex template provides a more flexible prompt structure for broader business tasks. Read every scenario carefully: the destination and expected output often reveal the right template type.

Select a grounding technique

Grounding provides relevant business context to the model. It can make a response more accurate and useful because the output is based on approved records, flows, data, or retrieved content instead of only the model’s general knowledge.

Create, activate, and execute templates

Know the basic lifecycle: create the template, define instructions and resources, preview and test the result, correct problems, activate the correct version, and execute it through the intended Salesforce feature. A draft template is not automatically ready for production use.

Apply access controls and Trust Layer protections

Users need appropriate access to prompt templates and the data used for grounding. The Einstein Trust Layer provides security and privacy capabilities for generative AI interactions. The blueprint also expects candidates to understand how administrators can manage or prevent access to specific models.

Effective prompts should state the task, supply relevant context, define limits, and specify the expected output. Avoid vague instructions, conflicting rules, excessive context, and data that the user should not access.

Domain 3: Data 360 Fundamentals (20%)

Salesforce renamed Data Cloud to Data 360, so older resources may use the previous name. The exam tests how Data 360 supports grounded agent responses, especially through Agentforce Data Library, chunking, indexing, and retrievers.

Agentforce Data Library

An Agentforce Data Library connects approved information to an agent. The source may include structured or unstructured business content. Your design should consider data quality, scope, permissions, update frequency, and whether the content actually answers the intended questions.

Chunking

Long unstructured documents are divided into smaller units called chunks. Chunk size and boundaries affect what can be found and how much useful context reaches the model. Poor chunking can split related information or return a passage that lacks the details needed for an accurate answer.

Indexing

Indexing prepares content for efficient retrieval. Think of it as organizing information so that a search or retriever can locate useful pieces when an agent needs them. Updated or removed source content may require the retrieval layer to stay synchronized.

Retrievers

A retriever selects relevant information from the prepared data source and supplies it as grounding context. In exam scenarios, connect the retriever design to the type of data, the user’s question, access requirements, and expected response quality.

Do not treat grounding as a guarantee. Trusted results still depend on accurate source data, suitable chunking and indexing, relevant retrieval, clear prompts, and correct permissions.

Domain 4: Testing, Deployment, and Maintenance (10%)

This domain follows the solution from a safe test environment into production.

Test with Agentforce Testing Center

Testing Center helps teams evaluate an agent across repeatable test cases. Study how test inputs, expected outcomes, and evaluation results help reveal routing, response, action, or grounding problems. One successful conversation is not enough evidence that an agent is ready.

Create tests for ordinary requests, unclear wording, missing data, permission restrictions, unsupported requests, action failures, and escalation. Review failures by category instead of changing several settings at once.

Deploy agents and prompt templates

Understand sandbox-to-production considerations. Check dependencies, permissions, data sources, actions, activated versions, environment-specific settings, and post-deployment testing. An agent can deploy successfully but still fail if its production dependencies or access controls are incomplete.

Agents and prompt templates have related but distinct deployment considerations. Practice identifying which components must exist first and which configuration should be rechecked after deployment.

Domain 5: Governance and Observability (10%)

Governance defines how an organization keeps agent behavior controlled, secure, and accountable. Observability gives teams the information needed to understand how the agent behaves after deployment.

The blueprint focuses on managing and monitoring agents, agent analytics, and optimization. Useful measures can include adoption, completion, escalation, errors, latency, feedback, and task resolution. The purpose is not to chase one attractive number. It is to determine whether the agent reliably completes approved business tasks without creating unnecessary risk.

When an agent performs poorly, review evidence before editing instructions. The issue may come from routing, missing actions, poor source content, weak retrieval, permissions, an integration failure, or unclear prompts. Change one major factor at a time and retest.

Domain 6: Multi-Agent Orchestration (5%)

Multi-agent orchestration appears small in percentage, but its terms can be easy to confuse.

SOMA

Single-Org Multi-Agent (SOMA) is an architecture in which multiple specialized Agentforce agents collaborate within one Salesforce org. An orchestrator provides a unified conversational entry point and delegates work to connected subagents while maintaining context. The official Salesforce multi-agent guidance describes SOMA as useful when specialized agents in the same org need coordinated routing and shared governance.

MCP

Model Context Protocol (MCP) is an open standard that helps agents connect to tools, data sources, and APIs through a consistent interface. Think “agent to tools and resources” when comparing it with A2A.

A2A

Agent-to-Agent (A2A) is an open protocol for communication and collaboration between agents. Think “agent to another agent,” including agents that may come from different platforms.

The exam can give you a business architecture and ask which protocol or pattern fits. First, determine whether the need is tool access, agent collaboration, or coordinated agents inside one org.

Your Agentforce Specialist Readiness Toolkit

Strong preparation should help you turn product knowledge into safe implementation decisions. Our study approach gives you:

• Coverage matched to the six current domains: Practice follows the Spring ’26 outline instead of the older five-domain version.
• Original scenario questions: Each question tests a business decision rather than recalling a copied sentence.
• Plain-language explanations: Answers explain the requirement, the correct feature, and why the alternatives do not fit.
• Practice across the full lifecycle: Questions cover planning, prompts, data, agent behavior, testing, deployment, observation, and improvement.
• A practical correction method: Missed answers become focused tasks such as testing a prompt template, checking security context, or comparing MCP with A2A.
• Hands-on activities: Suggested exercises connect the blueprint to actions in Agentforce Builder, Prompt Builder, Data 360, and Testing Center.
• Ethical exam preparation: The material is designed for learning and does not claim to contain live, leaked, or confidential Salesforce questions.

If you are preparing for another role in the Salesforce ecosystem, View the complete Salesforce practice-test collection and compare each credential’s intended audience before choosing your next exam.

How to Use an Agentforce Specialist Practice Test

Start with a diagnostic set

Take 15–25 mixed questions before studying deeply. Record the domain for every wrong answer, uncertain answer, and correct guess. A correct guess still reveals a knowledge gap.

Use a decision journal

For each scenario, write four short notes:

1. What is the business objective?
2. What security, data, channel, or lifecycle constraint matters?
3. Which feature best meets the complete requirement?
4. Why do the other options fail?

This method is more useful than memorizing answer letters because Salesforce questions often include several real features, but only one fits the stated requirement.

Reproduce the concept in a practice environment

If you miss a question about prompt activation, build and test a small prompt template. If you confuse subagents and actions, create one of each and describe their roles. If MCP and A2A seem similar, draw a simple “agent-to-tool” versus “agent-to-agent” comparison.

Retest with new wording

Do not immediately repeat the identical question. Wait, study the concept, and answer a different scenario that tests the same objective. You are ready when you can apply the concept without recognizing the wording.

Add timed mixed sessions

The official exam allows 105 minutes. Practice reading the requirement first, eliminating clearly unsuitable choices, and moving on when a question is consuming too much time. Leave review time for marked items.

Hands-On Agentforce Specialist Practice Checklist

Use a suitable Salesforce learning or practice environment for these activities:

1. Identify one clear employee or service-agent business use case.
2. List intended users, channels, approved data, actions, and escalation rules.
3. Create a basic agent in Agentforce Builder.
4. Compare Canvas view and Script view.
5. Create clear instructions for one subagent or topic.
6. Add a standard action and describe its inputs and outputs.
7. Design a custom-action scenario without granting excessive access.
8. Use a filter, variable, or template expression to control behavior.
9. Create a field generation prompt template.
10. Create or inspect a flex prompt template.
11. Test how grounding changes a response.
12. Review prompt-template access and model-access settings.
13. Add approved content to an Agentforce Data Library.
14. Explain how chunking, indexing, and retrieval affect the result.
15. Create normal, boundary, and failure test cases in Testing Center.
16. Review a failed test and identify the most likely cause.
17. List the dependencies needed for sandbox-to-production deployment.
18. Review agent monitoring or analytics and choose one improvement.
19. Compare MCP, A2A, and SOMA using one business example each.
20. Document how the agent should behave when data or an action is unavailable.

Salesforce recommends hands-on experience together with training and self-study. The official Design and Implement AI Agents with Agentforce trail walks learners through planning, building, testing, deploying, and monitoring an agent. The Agentblazer learning path provides additional progressive Agentforce learning.

Six-Week Salesforce Agentforce Specialist Study Plan

Week 1: Learn the platform and current outline

Read the entire Spring ’26 exam guide. Review Salesforce records, permissions, Flow, sandboxes, and basic deployment if these are not familiar. Learn the purpose of Agentforce Builder, Prompt Builder, Data 360, and Testing Center. Take a diagnostic practice set.

Week 2: Prompt Engineering and Trust

Study Prompt Builder use cases, access controls, field generation and flex templates, grounding, activation, execution, effective-prompt practices, Trust Layer protections, and model-access controls. Build and test small templates instead of only reading definitions.

Week 3: Data 360 Fundamentals

Work with approved structured and unstructured content. Learn the role of Agentforce Data Library, chunking, indexing, and retrievers. Practice diagnosing poor responses caused by missing, outdated, inaccessible, or poorly prepared data.

Week 4: AI Agents

Give this domain the most time because it represents 35%. Study Agent Script, Canvas and Script views, hybrid reasoning, filters, variables, template expressions, subagents, actions, channels, security context, Employee and Service agents, and Agent API.

Week 5: Lifecycle, governance, and orchestration

Use Testing Center, review evaluations, and map sandbox-to-production dependencies. Study monitoring, analytics, and optimization. Compare SOMA, MCP, and A2A until you can select each from a short business scenario.

Week 6: Timed practice and final review

Complete mixed timed sets. Review every incorrect answer and correct guess. Revisit one hands-on task from each weak domain. Confirm the live exam guide, registration details, identification requirements, and testing instructions before exam day.

Review the official guide again shortly before your appointment because Agentforce changes quickly. When you want to continue with another credential, Find your next certification practice test and choose a path that builds naturally on your current Salesforce skills.

Common Agentforce Specialist Preparation Mistakes

Using the old five-domain blueprint

This is the biggest content risk. Current preparation should cover six Spring ’26 domains, including separate Testing, Deployment, and Maintenance plus Governance and Observability sections.

Confusing a subagent with an action

A subagent defines an area of responsibility or capability. An action performs an operation. A scenario may require clearer classification, a new action, or both.

Treating grounding as automatic accuracy

Grounding quality depends on the source, permissions, chunking, indexing, retrieval, prompt, and current data. Always test whether the returned information actually supports the answer.

Ignoring security context

An agent may be correctly configured yet fail because its runtime context cannot access required data or automation. Avoid “solve it by making the agent an administrator” thinking. Select the minimum access that supports the approved task.

Testing only ideal questions

Real users provide incomplete, misspelled, ambiguous, and unsupported requests. Include edge cases, permission failures, missing data, action failures, and escalation paths.

Memorizing acronyms without architectures

Remember the relationship: MCP connects agents to tools and resources, A2A supports communication between agents, and SOMA coordinates specialized agents inside one Salesforce org.

Believing outdated free-exam promotions

Salesforce’s current guide lists a registration fee. Check official registration for the amount that applies to you and do not plan around an expired 2025 promotion.

Using claimed live questions

Leaked or memorized exam content is unreliable, may be outdated, and conflicts with certification security rules. Use original questions, official learning, and hands-on practice.

How to Know You Are Ready

You are approaching exam readiness when you can:

• Recall all six domains and their current percentages.
• Explain Prompt Builder, field generation templates, flex templates, and grounding in plain language.
• Describe how the Trust Layer and access controls support safe use.
• Connect Data Library, chunking, indexing, and retrieval as one grounding process.
• Explain Agent Script and hybrid reasoning without relying on memorized definitions.
• Choose subagents, actions, filters, variables, and template expressions from scenarios.
• Select between Employee and Service agents.
• Explain why an agent’s runtime security context matters.
• Build a repeatable Testing Center plan and interpret failed evaluations.
• List the main checks for deploying agents and templates to production.
• Use analytics to investigate behavior before changing the solution.
• Distinguish SOMA, MCP, and A2A quickly.
• Complete unfamiliar mixed practice questions within a realistic time limit.

Aim for consistent reasoning across new questions. A high score on a familiar set may only measure memory.

Salesforce Agentforce Specialist Frequently Asked Questions

What is the Salesforce Agentforce Specialist exam code?

The common registration and search code is AI-201. When scheduling, confirm that the full title says Salesforce Certified Agentforce Specialist.

How many questions are on the Agentforce Specialist exam?

The current official guide lists 60 scored multiple-choice questions and up to five additional unscored questions. The unscored items are mixed into the exam and do not affect the result.

How long is the Agentforce Specialist exam?

You have 105 minutes to complete the exam.

What score is required to pass?

The current passing score is 72%. Salesforce bases the result on the total number of scored questions answered correctly.

Is the Agentforce Specialist exam still free?

The official Spring ’26 guide lists a USD 200 registration fee, plus applicable taxes. The earlier free-attempt promotion is no longer the current standard price. Verify the amount shown during registration.

How much does an Agentforce Specialist retake cost?

The current guide lists a USD 100 retake fee, plus applicable taxes. Prices and policies can change, so confirm them before booking.

Are there prerequisites?

No formal prerequisite certification is required. Platform Administrator and Platform App Builder are recommended related certifications, not mandatory requirements.

How much Salesforce experience should I have?

Salesforce says a successful candidate typically has about one year of Salesforce platform configuration and standard-object experience, including Data 360, plus hands-on exposure to Agentforce Builder, Prompt Builder, testing, and deployment.

What are the current exam domains?

The six domains are AI Agents, Prompt Engineering, Data 360 Fundamentals, Testing/Deployment/Maintenance, Governance/Observability, and Multi-Agent Orchestration.

Which domain is the most important?

AI Agents is the largest at 35%. Prompt Engineering and Data 360 Fundamentals are 20% each. Together, those three areas account for 75% of the blueprint.

Do I need to know Apex or Python?

The official audience description says candidates are not expected to know the basics of coding languages such as Apex or Python. You should still understand when an action or API-based integration fits a business scenario.

What happened to “topics” in Agentforce?

Salesforce began renaming agent topics to subagents in April 2026. The functionality did not change during the terminology transition, so current and older resources may use both terms.

What is the difference between MCP and A2A?

MCP provides a standard way for an agent to connect to tools, resources, and APIs. A2A supports communication and collaboration between agents.

What is SOMA in Agentforce?

SOMA means Single-Org Multi-Agent. It coordinates multiple specialized Agentforce agents in one Salesforce org through a unified entry point and orchestrated task delegation.

Can I use notes or Salesforce Help during the exam?

No. The current exam guide says hard-copy and online reference materials are not permitted.

How do I maintain the certification after passing?

Salesforce currently requires completion of the Agentforce Specialist maintenance module on Trailhead once each year. Missing maintenance requirements can cause the certification to expire.

Are practice questions enough to pass?

Practice questions are useful for checking decisions and timing, but they should not replace the official exam guide, Trailhead study, Salesforce documentation, and hands-on building and testing.

How long should I prepare?

There is no universal duration. Candidates with recent hands-on Agentforce experience may need a focused review, while learners new to the platform may need several weeks or months. Start with a diagnostic set and adjust your plan by domain.