1.1 Agentforce Specialist (AI-201) Exam Facts & Spring '26 Blueprint

1.1 Agentforce Specialist (AI-201) Exam Facts & Spring '26 Blueprint

Quick Answer: The Salesforce Certified Agentforce Specialist exam (listed on Trailhead Academy as AI-201) has 60 scored multiple-choice questions plus up to 5 unscored questions, a 105-minute time limit, and a 72% passing score. Salesforce's current exam guide lists the Spring '26 release and six domains: AI Agents (35%), Prompt Engineering (20%), Data 360 Fundamentals (20%), Testing, Deployment, and Maintenance (10%), Governance and Observability (10%), and Multi-Agent Orchestration (5%).

This chapter starts with the facts that shape your study plan. The Agentforce Specialist credential has changed several times: it began as the AI Specialist exam, was renamed for Agentforce, was restructured into a five-domain outline in October 2025, and was restructured again for the Spring '26 release. Many older courses and practice tests still teach the five-domain version, so always check a resource against the current outline before you trust its weights.

Who the Exam Is For

Salesforce describes the credential as being for people who design and implement prompt templates and agents. The target audience is administrators, developers, and architects who configure Prompt Builder for grounding, use Data 360 for retrieval, and manage the development lifecycle of agents from testing to production.

A typical successful candidate has about one year of experience with Salesforce platform configuration and standard objects, including Data 360, plus hands-on work in:

• Agentforce Builder – building agents, subagents (formerly called topics), actions, and instructions
• Prompt Builder – creating, grounding, and activating prompt templates
• Testing Center – running batch tests before deployment
• Sandbox-to-production deployment of agents and templates

The exam guide also lists what candidates are not expected to know: fine-tuning large language models (LLMs), coding basics in Apex or Python, external AI tools, transformer architecture, return-on-investment or white-space calculations, and product experience with Marketing Cloud, Heroku, Agentforce Commerce, MuleSoft, Tableau, Slack, Quip, or Industry solutions. If a practice question needs deep Apex syntax or transformer math, it is testing something outside the official scope.

Current Exam Logistics

Registration runs through Trailhead Academy, and Salesforce exams have been delivered by Pearson VUE since the certification experience moved from Webassessor in July 2025. Older blog posts that mention Kryterion or Webassessor describe the retired process.

Salesforce publishes a percentage, not a raw cutoff. If all 60 scored items carry equal weight, 72% works out to 44 correct answers because 43 correct is only 71.7%. Treat that as a planning estimate, not an official number.

The Spring '26 Exam Outline

*Approximate counts are 60 × the domain weight. Salesforce doesn't publish per-domain item counts.

Three changes from the October 2025 outline matter most:

1. Agent Script and hybrid reasoning are now tested. The AI Agents domain asks how an agent works, the building blocks of Agent Script, and how Agent Script appears in Canvas view and Script view.
2. The Einstein Trust Layer and model access moved into Prompt Engineering. Expect questions on Trust Layer privacy features and on preventing specific models from being used.
3. The lifecycle domain split in two. Testing and deployment (including prompt template deployment) is one 10% domain, and monitoring, agent analytics, and agent optimization form a separate 10% domain.

Terminology You Will See in Current Salesforce Docs

Salesforce renames things quickly, and exam questions and product documentation don't always use the same label at the same time.

Building a Study Plan From the Weights

• Spend about a third of your time on AI Agents. Build at least one Service agent and one Employee agent. Write Agent Script with filters, variables, conditional logic, and transitions, and connect an agent to a channel.
• Split Prompt Engineering and Data 360 evenly. Create field generation and flex templates, ground them with records, flows, and retrievers, and build a Data Library so you can see the search index and retriever it creates.
• Don't skip the 10% domains. Testing Center scorers, deployment gotchas, and the observability toolchain produce precise, memorizable questions.
• Finish with Multi-Agent Orchestration. It's only about three questions, but the terms (orchestrator agent, handoff mode, supervisor mode, MCP, A2A) are easy points if you learn them.

The rest of this guide follows those domains in order, with one section per official objective and extra sections where an objective covers several separate skills.