Foundry Academy · AI Workflow Training · Lesson 2 of 6

Instructions, context, and source grounding

Create a versioned instruction package that defines the task, authorized context, evidence rules, output contract, uncertainty behavior, and review handoff.

Start the lesson

Your learning work, on this device

No signup, cloud storage, cross-device sync or verified completion. Saving is optional. This browser profile is shared with anyone who can use it; private mode, browser cleanup or storage limits may remove work. Use only fictional or non-sensitive material. Export a copy before relying on this device.

Not saved. Worksheets, answers and practice notes currently last only in this tab.

Practice markers are self-reported, never credentials.

01 · Explanation

Instructions, context, and source grounding

Objective: Create a versioned instruction package that defines the task, authorized context, evidence rules, output contract, uncertainty behavior, and review handoff.

A dependable instruction package describes purpose, audience, authorized sources, definitions, constraints, required output, prohibited content, and what to do when evidence is missing. Give the model the minimum context needed and label source boundaries explicitly. If the answer must come from a policy set, require source identifiers and quotations short enough for a reviewer to verify, while respecting copyright and confidentiality. Do not ask the system to invent citations, legal certainty, numerical inputs, customer facts, or current information it cannot access. Examples should demonstrate the rule, not secretly replace it.

Treat instructions as controlled process documentation. Record the model or service, instruction version, source version, key settings where available, output schema, and reviewer role. Test ambiguous language and conflicting sources before release. Define how the system should express uncertainty, abstain, and request clarification. Structured outputs can improve consistency but do not guarantee truth; a well-formed JSON object can still contain fabricated facts. A reviewer should be able to trace material claims to authorized sources and distinguish quoted evidence from model-generated interpretation. Changes to prompts or sources require targeted regression testing, not silent production edits.

Before you begin

  • Confirm F04, F07-F10; record that the source schema, approved route taxonomy, abstention wording, multilingual fidelity test, and change approval are not supplied.
  • STOP. If the source schema, route taxonomy, abstention wording, or multilingual-fidelity test is missing or conflicts with F04/F07–F10, route the instruction to the authorized domain and bilingual reviewers; do not claim change approval or execute deployment or automatic rejection.

Original overview module anchor →

02 · Compare the artifacts

Supported work. Visible uncertainty.

This is a fictional, sanitized AI-governance training case. All records are fabricated, and the exercise provides no legal, privacy, employment, medical, financial, security, or model-performance guarantee.

Redwood Assist governed intake workflow

Redwood Assist is a fictional services company considering an AI workflow to summarize new client intake and draft a routing recommendation. The current form receives about 420 submissions each month. It includes name, business email, phone, company, project description, budget range, requested timing, and an optional attachment. In a sample of 60 fabricated records, seven attachments contain government identifiers, four contain medical details unrelated to the service, nine include third-party personal information, and 16 lack enough context for reliable routing. The proposed model vendor offers a standard account with model-improvement use enabled by default and a 30-day content retention statement. No security, privacy, contract, data-location, deletion, or subprocessors review has been completed. A pilot prompt tells the model to identify serious founders and reject low-quality leads. The output labels applicants high, medium, or low potential, but neither potential nor serious is defined. In a 40-record test, the model routes 31 correctly according to one manager, while a second manager disagrees on 11 of those decisions. Two Spanish-language submissions are summarized with missing budget conditions. The team proposes automatically rejecting low-rated applicants and sending their attachment to a funding partner. There is no consent for partner disclosure, no human-review standard, no protected exception path, and no versioned evaluation set. Leadership wants the automation live next week to save eight hours of staff time. Learners must classify risk, constrain instructions and sources, establish data boundaries, design meaningful human review, build an evaluation set, and control exceptions and changes. The exercise cannot establish vendor suitability, lawful processing, fairness, or permission to automate decisions.

Supported example — reference only

Section or field
Decision boundary
Purpose
Prevent an unsupported consequential outcome.
Supported entry
When required routing context is missing, output NEEDS_REVIEW with the missing fields; never reject automatically.
Source input ID
F04, F10
Gap or learner proposal
Approved required-field list and routing taxonomy are not supplied.
Owner or reviewer
AI workflow owner and legal/privacy reviewer
Status
Draft - review pending

A well-handled evidence gap

Section or field
Multilingual condition preservation
Purpose
Prevent omission of material terms.
Supported entry
Two Spanish-language summaries omitted budget conditions.
Source input ID
F09
Gap or learner proposal
Approved translation/fidelity reference and reviewer are not supplied.
Owner or reviewer
Bilingual domain reviewer
Status
Blocked - reference needed

Flawed approach — do not copy

Marking this versioned instruction “approved and complete” without the required evidence or reviewer is a flawed submission. Stop when the instruction uses undefined labels, permits rejection, omits required terms, or lacks an abstention path.

Repair: Rework the versioned instruction as an evidence-backed draft, not an approved result. Version the instruction and name its bounded classification-and-summary purpose. Define required source fields before describing any output. Replace serious founders and low-quality leads with observable, reviewable values or mark them prohibited. Check the revision against this requirement: Instruction purpose, version, source schema, output schema, and abstention path are aligned. If the required evidence is still absent, keep the decision blocked and identify the missing input or authorized reviewer.

Full case record, ambiguities and all assignments →

03 · Bounded practice

Build the versioned instruction.

Replace subjective labels with an evidence-grounded extraction and routing specification.

Deliverable: A versioned instruction, source schema, output schema, and unsupported-output rule.

Complete a bounded starter and gap analysis using only CB01, F04, F07, F08, F09, F10, and the assignment-scope record below. Populate supported fields, label every unavailable field “not supplied,” and cite the input ID for each material statement. You may design a proposed template, control, question, or decision rule, but must label it as a learner proposal rather than observed case evidence. Do not contact people, access live systems, run tests, sign records, claim approval, or invent names, dates, quotations, transactions, results, or source documents.

Exact supplied inputs for this assignment
  • M02-I01 · F04 — Sixteen sampled records lack enough context for reliable routing.
  • M02-I02 · F07 — The prompt uses undefined labels serious founders and low-quality leads.
  • M02-I03 · F08 — Managers disagree on eleven of thirty-one supposedly correct routing decisions.
  • M02-I04 · F09 — Two Spanish-language summaries omit budget conditions.
  • M02-I05 · F10 — The team proposes automatic rejection using the low-potential label.
  • M02-B01 · CB01 — Use CB01, the full versioned case brief printed once at the start of this packet, as a citable narrative source for details not normalized into F01–F12. Preserve its uncertainty language and do not treat narrative detail as approval, complete operational records, or professional judgment.
  • M02-S01 · F04, F07, F08, F09, F10 — Build a starter version of “A versioned instruction, source schema, output schema, and unsupported-output rule.” from the listed case facts. Treat requested structures, controls, questions, calculations, and templates as learner-designed proposals. Where an operational record or result is absent, add a gap entry naming the missing evidence and authorized owner instead of fabricating it.

Operating procedure

  1. Version the instruction and name its bounded classification-and-summary purpose.
  2. Define required source fields before describing any output.
  3. Replace serious founders and low-quality leads with observable, reviewable values or mark them prohibited.
  4. Write an abstain-and-escalate branch for missing routing context.
  5. Require every budget condition to be preserved in summaries and route disagreements to adjudication.
  6. Prohibit automatic rejection and unsupported inference in the output rules.
  7. Run final QC for version, source-to-output traceability, reviewer boundary, exception path, and pending approval.
Field-by-field guidance
Section or field
Name the reusable template field. Module use: Use the instruction to constrain inputs, outputs, abstention, and human review without treating prompt text as a production control.
Purpose
Explain the decision or control served by the field. Module use: Use the instruction to constrain inputs, outputs, abstention, and human review without treating prompt text as a production control.
Supported entry
Show the packet-supported starter value. Module use: Use the instruction to constrain inputs, outputs, abstention, and human review without treating prompt text as a production control.
Source input ID
Cite the exact Fxx or module input ID supporting the entry. Module use: Use the instruction to constrain inputs, outputs, abstention, and human review without treating prompt text as a production control.
Gap or learner proposal
Write “not supplied” for missing evidence; label any designed rule as a learner proposal. Module use: Use the instruction to constrain inputs, outputs, abstention, and human review without treating prompt text as a production control.
Owner or reviewer
Name an authorized role, never an invented person or completed approval. Module use: Use the instruction to constrain inputs, outputs, abstention, and human review without treating prompt text as a production control.
Status
Use a truthful state such as draft, open—not supplied, review pending, or blocked. Module use: Use the instruction to constrain inputs, outputs, abstention, and human review without treating prompt text as a production control.
Versioned instruction · learning draft
Section or fieldPurposeSupported entrySource input IDGap or learner proposalOwner or reviewerStatus

Start with 6 rows; the complete workbook specifies 9 stable rows for this artifact. Add rows here or use the full download. No action is saved until you explicitly choose saving above.

Download complete six-module workbook (.md) · Structured case packet (.json)

Keep private client data, unpublished inventions, personal identifiers and credentials out of these public learning tools.

Module 2 · 2-item formative check

Instructions, context, and source grounding

Choose an answer and request feedback. Read why each option does or does not fit the evidence. Answers stay in this tab unless you choose device-only saving; they are never submitted.

Question 1 of 2 · MODULE 2 · knowledgeWhat does source grounding contribute to a governed AI workflow?
Question 2 of 2 · MODULE 2 · scenarioF04 confirms insufficient context, F07 confirms undefined labels, F08 is conflicting across manager decisions, F09 confirms omitted budget conditions in Spanish summaries, and F10 is provisional for automatic rejection. Which instruction sequence is evidence-safe?

Answer either question to review its reasoning.

Inspect the artifact, not just your quiz answers

  • Instruction purpose, version, source schema, output schema, and abstention path are aligned.
  • Undefined labels and automatic rejection are prohibited.
  • No instruction is represented as deployed or approved.

Stop: Stop when the instruction uses undefined labels, permits rejection, omits required terms, or lacks an abstention path.

Go: Proceed to offline review when schemas, prohibited outputs, missing-data behavior, and human handoff are explicit.

Escalate: Escalate label disputes and multilingual material-condition failures to the accountable domain and bilingual reviewers.

04 · Evidence to keep

Leave with usable work.

Submit the instruction, source manifest, output contract, abstention rule, two test outputs, and a reviewer traceability check.

Download your artifact CSV and, if wanted, export the learning-work JSON above. Neither export is a reviewed submission or certificate. Device-only saving is optional; you must press Save my work now after edits.

When all six artifacts are ready, compare the full packet against the track rubric. Qualified human review is still required before real-world decisions.

Technology team discussing an AI-assisted workflow and its controls.
Learn the standard. Practice the work.
Professional reviewing an AI-assisted output on a laptop before approval.
Leave with evidence you can inspect.

Sources, scope and review boundaries

Curriculum 2026.10.08-learning-paths-1. External source dates below are record checks, not continuing guarantees. Verify current requirements before consequential use.

nist-ai-600-1 · Official guidance

NIST AI 600-1: Generative Artificial Intelligence Profile

NIST's cross-sector profile describing generative-AI risks and actions aligned with AI RMF 1.0.

Open reviewed external source ↗

ws-ai-workflow-operating-standard · Academy internal operating standard

Wealth Synergy AI workflow internal operating standard

Academy-selected task classification, prompt, evidence, human-review, evaluation, exception, and change controls. This is an internal operating standard selected by Foundry Academy; it is not law, accreditation, licensure, or an external-standard requirement.

Version 1.0 · reviewed 2026-09-01 · owner: Foundry Academy curriculum owner

A future Wealth Synergy private professional-development certificate would be issued only after its assessment, capstone, identity, reviewer, retention, access, deletion, appeal, and issuance controls pass quality review. No credential is currently issued. Any future certificate would not be an accredited academic qualification, professional license, or government certification.