Free
- Untimed practice across all five sections
- Every question explained against the requirement that decides it
- Sections by name and weight
Salesforce
Four answers that would all technically run
That is what this exam is. A business requirement, and configurations that differ by fit rather than by whether they work. 60 questions, 105 minutes, 73% to pass. Practice with original items, free.
- 60 scored questions plus up to 5 unscored, 105 minutes, no breaks
- Five sections, from AI Agents at 35% down to Multi-Agent Interoperability at 5%
- Readiness shown against the 73% cut score as a range, not a single number
Agentforce SpecialistAI AgentsHard
An agent handles both order status and returns. In testing, a customer saying their item arrived damaged and they want to send it back is answered by the order status topic instead of the returns topic. Both topics have correct scopes and instructions. What should you change first?
Select one.
1 of 5
Original items, one per published section. Not a real exam question.
The loop
Practice, sit, review, repeat
Step 1 is free. Steps 2 to 4 need a subscription.
certs.acedify.com/agentforce-specialist
Practice · untimed
HardAn agent handles both order status and returns. In testing, a customer saying their item arrived damaged and they want to send it back is answered by the order status topic instead of the returns topic. Both topics have correct scopes and instructions.
What should you change first?
Rewrite the returns topic’s scope so it describes damaged goods explicitly.
Sharpen the topic label and classification description on both topics so the boundary between them is explicit.
Add an instruction to the order status topic telling it not to handle returns.
Merge the two topics into one so there is no choice to get wrong.
Topic selection happens before anything else is read. Label and classification description decide which topic a request lands in; scope and instructions only shape what happens once it has landed. Domain: AI Agents.
Click a step to see it.
What you get back
A sitting is a number. Ability is a range.
Finish a sitting and you get a score, because it happened. Your ability estimate is different: it is an inference from every answer you have given, so it is reported as a range that narrows as you do more work. Until we have enough answers to say something useful about a section, we show nothing for it rather than a range so wide it means nothing — which is why Multi-Agent Interoperability, at three questions a sitting, takes the longest to fill in.
Crosses 73%
A range, because an estimate without one is a guess dressed up as a fact. We report it against the cut score. Until a section has enough answers, we show nothing for it.
- Overall
- Range still crosses 73%
- AI Agents
- Enough answers to show a range
- Prompt Engineering
- Enough answers to show a range
- Data Cloud for Agentforce
- Enough answers to show a range
- Deployment Lifecycle
- Enough answers to show a range
- Multi-Agent Interoperability
- Not enough answers yet
Sections
What Agentforce Specialist actually asks
Three different blueprints for this exam are circulating on prep sites, and the older ones still split the paper by Service Cloud and Sales Cloud. These are the weights Salesforce publishes.
Building and customising agents for a stated business need, choosing between employee, service and sales agents, controlling behaviour with topics, instructions and actions, and validating what you built in Testing Center.
- Employee, service and sales agents
- Topics, instructions and actions
- Deterministic behaviour with filters and variables
- Testing Center
When Prompt Builder is the right tool, who can manage and run templates, what constrains a template design, and which grounding technique a given scenario calls for.
- When Prompt Builder fits the requirement
- Template creation, activation, execution
- Roles that can manage templates
- Choosing a grounding technique
Grounding responses with retrievers, using the Agentforce Data Library to raise answer accuracy, and personalising what an agent says with data the org already holds.
- Retrievers
- Agentforce Data Library
- Response accuracy
- Personalisation
The product was renamed Data 360 in October 2025. The exam objectives still print "Data Cloud".
Moving agents from sandbox to production, monitoring adoption once they are live, and keeping user security and responsible-use controls intact through the move.
- Sandbox to production
- Adoption monitoring
- User security
- Responsible use
How agents hand off to and coordinate with one another. The smallest section on the exam, and the newest.
- Agent-to-agent handoff
- Coordination patterns
Pricing
Practice free. Pay when you want to know where you stand.
Pro
- Timed 60-question sittings at 105 minutes
- Full mock exams drawn against the published weights
- Your ability estimate and readiness
- Analytics and coaching
One subscription covers all our IT certification tracks.
Start practising for freeFAQ
Agentforce Specialist questions
Find out where you actually stand
Start with untimed sets across all five sections. When you want the clock, the mock and the estimate, they are one subscription away. Every performance claim on this site is listed in our claims register.
Start practising for free