Course 3.3 · Curriculum 2026.1
Prompt Builder and Agent Builder in practice
Build a grounded prompt template and an agent with real topics and actions. Then test it against the cases that break it.
- Estimated effort
- 6 hr
- Lessons
- 50
- Modules
- 8
- Material
- 6 hr
Estimated effort is the sum of the lesson and lab times in this version of the course. It is not video runtime, which is lower.
Canadian dollars. One learner, lifetime access to this course version. Tax extra.
1 lesson free to preview
This course does not issue a certificate on its own. It counts toward the paths listed below, and those do.
01What you will learn
7 capabilities, stated as things you can do
Each one is what you should be able to do at the end, not what the course covers. If an outcome is not testable, it is not an outcome.
- Build a prompt template grounded in record data, with the fields chosen deliberately.
- Write instructions that survive a user who does not follow the happy path.
- Define agent topics and actions, and say what each action is allowed to write.
- Wire a Flow or an Apex action into an agent, with input and output types that hold.
- Test an agent against a set of cases, not by trying it once and shipping.
- Read the trace of an agent turn and say why it chose the action it chose.
- Set the boundaries: what the agent must escalate rather than answer.
02Curriculum
8 modules, 50 lessons
Every lesson shows its length and its type. A SCORM lesson looks like any other lesson, which is the point of ingesting packages rather than linking out to them.
8 modules · 50 lessons · 6 hr of material
Prompt Builder and Agent Builder are configuration surfaces, so they deserve the same review discipline as a validation rule.
- VideoGetting an org with generative AI enabledPreview6 min
- VideoTemplate types, and choosing by where the output lands6 min
- VideoPrompt Builder against Agent Builder: two tools, one model6 min
- VideoReading Salesforce’s own prompt gallery for the trick, not the copy6 min
- ReadingThe org you need, and the trial path if you don’t have one5 min
- LabLab: build your first template and run it once14 min
The fields you send are the fields the model can leak. Choose them the way you would choose what goes in a report.
- VideoGrounding: picking the fields, and the ones you must not send6 min
- VideoMerge fields, related lists, and flow-based grounding6 min
- VideoThe field you added because it was there6 min
- VideoPII minimisation inside a grounded template6 min
- ReadingA grounding checklist before you save5 min
- LabLab: build a grounded summary template14 min
Instructions written for the happy path fail the first time a user does not follow it. Write for the second user, not the first.
- VideoWriting instructions that hold under an unusual record6 min
- VideoThe user who goes off the happy path6 min
- VideoNegative instructions: what the model must never do6 min
- VideoReal-data-only: no placeholder names, no round invented figures6 min
- ReadingDate grounding, so “this year” means this year5 min
- LabLab: break your own instructions, then fix them14 min
A topic is a boundary and an action is a capability. Getting the boundary wrong is how an agent answers a question it should have escalated.
- VideoTopics, classification, and the boundary between them6 min
- VideoStandard actions, Flow actions and Apex actions6 min
- VideoWhat each action is allowed to write, decided before it is built6 min
- VideoInput and output types that do not drift6 min
- ReadingNaming topics and actions so the model picks correctly5 min
- LabLab: define two topics and the actions under each14 min
Actions that write need more discipline than actions that read. This module wires both, and adds the guardrail that keeps a raised exception from killing the whole run.
- VideoWiring a Flow action with input and output variables6 min
- VideoWiring an Apex-invocable action the same way6 min
- VideoA safe-tool wrapper: never let a raised exception kill the run6 min
- VideoEvery write wrapped in a required confirmationThe user gets an approve button. Nothing writes without one, no exceptions for convenience.6 min
- ReadingAuth errors marked non-retryable, with a hint to stop and ask5 min
- LabLab: build an agent with two topics and one write action14 min
Every one of these recurs across four production agents. That repetition is the evidence they are not incidental.
- VideoQuery coaching inside the error hint6 min
- VideoThe real-data-only instruction, and what happens without it6 min
- VideoDate grounding on every run6 min
- VideoBulk edits as one approval, not N6 min
- VideoNever claim success from memory: re-query by id after a write6 min
- ReadingFour agents, one guardrail list: what they all share5 min
- LabLab: add a safe-tool wrapper to a failing action14 min
What the agent must not answer is as much a design decision as what it may. Then learn to read the trace that shows why it chose what it chose.
- VideoEscalation: what the agent must refuse to answer6 min
- VideoReading a trace of an agent turn6 min
- VideoWhy the agent chose that action, and why it chose the wrong one6 min
- VideoAuth errors reaching the user, not swallowed by a retry loop6 min
- ReadingDrafting the refusal message so it does not read as a bug5 min
- LabLab: read three traces and explain each decision14 min
The gap between a demo and production is a test set. This module builds one, runs it, and grades it without grading the wording.
- VideoBuilding a test set before you shipTwenty cases with expected outcomes, including five that must be refused. Run it on every change.6 min
- VideoTwenty cases, five of which must be refused6 min
- VideoThe accuracy drop from test to production, and where it comes from6 min
- VideoRunning the case set on every change to the instructions6 min
- ReadingGrading a transcript without grading the wording5 min
- LabLab: build and run a twenty-case test set15 min
- QuizCheck: Prompt Builder and Agent Builder9 min
03Before you start
What you need first
Assumed knowledge and setup
- Course 3.1, and course 1.4 for the Flow actions.
- An org with Einstein Generative AI enabled. A trial org path is documented in lesson one.
- No Apex required. Course 3.4 covers the Apex route.
Courses that come first
- 3.1 · The three places AI can run, and how to choose3 hr · Intermediate
Part of these paths
- AI on SalesforceShareCo Certified AI Practitioner
- The Full StackShareCo Certified Platform Engineer
04Who teaches it
Ege Taskent
Salesforce Developer, Certified Platform Administrator
The credential, and the rollout. The half of an implementation where people actually start using the thing.
Salesforce Certified Platform Administrator
05Reviews
No learner reviews yet
This course has not been taken by enough people to publish an honest rating, and we will not print invented quotes on a page that sells verification discipline. Here is what we can evidence instead.
- Assessment
- Lesson checks onlyThis course carries lesson checks. The certificate is issued by the path assessment, not by this course.
- Curriculum version
- 2026.1You enrol into a version. Content changes do not move you mid-course, and your required-lesson count is snapshotted at enrollment.
- Refunds
- 14 daysFull refund within 14 days if you have completed under a quarter of the required lessons. Stated here rather than in a footer.
Start 3.3
Enrol as an individual, or buy seats and assign this course to your team. Progress is reported per lesson, per module and per path.
- 6 hr estimated effort, labs included
- Lifetime access to the version you enrol in
- A ShareCo certificate is not a Salesforce certification