AI Agents35% of the exam
Hybrid reasoning in Canvas and Script view
How Canvas view and Script view edit one Agent Script, where hybrid reasoning draws the line between deterministic code and LLM calls, which lifecycle points spend an LLM call, and when a workflow should stay LLM-led.
Explain the components and benefits of hybrid reasoning, including how Agent Script functions in Canvas and Script View.
~7 min read · 12 practice questions ·
What the exam expects
- Explain that Canvas view, Script view and chatting with Agentforce all edit the same Agent Script, and what each surface offers.
- Explain how hybrid reasoning splits work between deterministic execution and LLM reasoning, and what triggers an LLM call.
- Identify which lifecycle points invoke the LLM and which never do, to budget model usage.
- Justify hybrid reasoning (auditability, cost, latency, predictability) and recognise when LLM-led reasoning is the better fit.
- Key terms: hybrid reasoning (deterministic code plus LLM reasoning in one agent); logic instruction (
->, runs as code); prompt instruction (|, natural language the LLM interprets); Canvas view (block-based editor); Script view (direct code editor).
Key facts
One artifact
- Detail
- Canvas and Script view edit the same Agent Script; configuration stays consistent
Canvas shortcuts
- Detail
/inserts expressions such as if/else;@inserts subagents, actions and variables
Committing
- Detail
- Warnings allow a commit; errors block it
Invalid script
- Detail
- Can make Canvas view unavailable until fixed in Script view
LLM trigger
- Detail
- A node's prompt instructions trigger the LLM call; other nodes run as code
Deterministic path audit
- Detail
- Results still pass through the trust layer with full audit logging
Always deterministic
- Detail
- Graph traversal and state transitions
Always an LLM call
- Detail
- Agent reasoning (choosing the next action)
Action billing
- Detail
- Actions bill whether deterministic or LLM-chosen; utilities don't bill as actions
Builder assistant
- Detail
- Not billed or metered; doesn't consume credits
Simulate vs Live Test
- Detail
- Simulate uses mock data and changes nothing; Live Test runs actions and updates data
Sources for this section (6)
- Building Agents in Canvas Viewhelp.salesforce.com (opens in a new tab)
- Agent Scriptdeveloper.salesforce.com (opens in a new tab)
- Agentforce Builder Considerationshelp.salesforce.com (opens in a new tab)
- Hybrid Reasoningarchitect.salesforce.com (opens in a new tab)
- Learn About New Agentforce Builderhelp.salesforce.com (opens in a new tab)
- Simulate and Live Test Modeshelp.salesforce.com (opens in a new tab)
- Why can moving a business rule from a prompt instruction into Agent Script logic lower cost and latency?
- When should an agent lean on LLM reasoning instead of deterministic logic?
- When does Canvas view become unavailable?
- Which agent behaviour never involves the LLM?
The questions are free to read. Answers with their sources and self-grading are included with the pack.
Keep reading with the Agentforce Specialist pack
- Core concepts
- Exam traps
- Worked scenario
The pack includes:
- All 26 lessons of the Agentforce Specialist guide, with reading progress
- Interactive decision walks and step-by-step sequence steppers
- Inline practice questions and one-click topic practice (12 questions for this lesson)
- An official source cited for every claim
- An AI tutor that answers from the guide
- Unlimited timed practice exams