Practical guide

AI academy curriculum map template

Use this template to connect an AI academy to real work outcomes. The map forces each module to name the work decision, observable objective, practice case, evidence artifact, feedback route, assessment, transfer check, and accountable owner. It helps programme designers find attractive sessions that do not build or test usable capability. Copy the blank map into a spreadsheet or document..

By Two Prune

Use this template to connect an AI academy to real work outcomes. The map forces each module to name the work decision, observable objective, practice case, evidence artifact, feedback route, assessment, transfer check, and accountable owner. It helps programme designers find attractive sessions that do not build or test usable capability. Copy the blank map into a spreadsheet or document. Complete the work outcomes and evidence plan before choosing speakers or filling a calendar.

Evidence behind the design

The UK government's employer guide describes effective AI upskilling as practical, reachable, integrated, modular, expandable, and sustainable. It recommends real tasks and decisions, safe practice, organizational alignment, follow-up, and support for transfer. CDC's Quality Training Standards begin with a needs assessment, align measurable objectives and methods, call for accurate current content and subject-matter review, include practice and feedback, and require usability and accessibility. The OECD's work on an AI-ready public workforce also distinguishes broad foundational literacy from role-specific and specialist needs. These sources guide the map's structure; they do not establish that a particular academy will improve organizational outcomes. UK Employer Guide CDC Quality Training Standards OECD: Building an AI-ready public workforce

Sources: [1] [2] [3]

Blank curriculum map

Create one row for each assessable capability step, not one row for every lecture.

| ID | Work outcome and user | Decision or task | Observable objective | Prerequisite | Practice case and conditions | Required evidence artifact | Feedback and reviewer | Assessment and pass rule | Transfer check | Risk boundary and escalation | Owner and review date | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | 1 | | | | | | | | | | | | | 2 | | | | | | | | | | | | | 3 | | | | | | | | | | | |

Add a small programme-level register beside the map:

| Field | Entry | | --- | --- | | target population and access needs | | | business or service need | | | workflows included | | | workflows excluded | | | approved tools and data | | | baseline evidence | | | programme owner | | | content reviewers | | | evidence owner | | | version and effective date | | | next content review or expiry | |

How to complete the map

### 1. Start from work outcomes

Name the person who uses the output and the decision it supports. “Understand prompting” is not a work outcome. “Produce a source-linked variance explanation that a finance partner can approve or return” is. Confirm that a capability gap exists and that training is part of the remedy. Missing access, unusable data, absent policy, or a broken process will not be fixed by a course alone.

### 2. Write observable objectives

Use a verb, conditions, output, and standard:

> Given an approved monthly ledger extract and commentary pack, produce a variance explanation that separates observed values, calculations, assumptions, and unresolved questions; all decision-driving figures must reconcile to the supplied source.

One objective may combine several skills, but its evidence must remain inspectable. Avoid objectives based only on confidence, attendance, or tool usage.

### 3. Design representative practice

Specify the case, tool access, time, collaboration, source set, and review conditions. Include an ordinary case and at least one meaningful difficulty: contradictory sources, missing data, a plausible unsupported claim, a tool failure, or a boundary that requires escalation. Practice should resemble the work without exposing live confidential data.

### 4. Name the artifact before the session

The artifact is what the learner produces: a comparison table, checked analysis, decision memo, prompt-and-verification log, corrected model, escalation note, or final deliverable. It provides evidence for feedback and assessment. If a module has no artifact, decide whether it is orientation rather than assessable capability.

### 5. Separate feedback, assessment, and transfer

Feedback helps the learner improve during practice. Assessment checks a defined standard under controlled conditions. Transfer asks whether the behavior appears later in a new case or workplace setting. Use different columns because a well-coached practice artifact is not independent evidence of transfer.

### 6. Add boundaries and ownership

Record prohibited data, required human review, permitted uses, escalation triggers, retention, and tool limitations. Name the person who owns content accuracy, the reviewer who can challenge it, and the date when it must be updated or retired. Changing the model, data access, policy, workflow, or decision standard can trigger an earlier review.

Worked example: four-week finance variance academy

This synthetic academy supports finance partners who explain monthly operating-cost variances to budget owners. It does not train forecasting, accounting policy decisions, or approval authority.

| ID | Work outcome and user | Decision or task | Observable objective | Prerequisite | Practice case and conditions | Required evidence artifact | Feedback and reviewer | Assessment and pass rule | Transfer check | Risk boundary and escalation | Owner and review date | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | F1 | clean issue list for finance partner | find material data problems before analysis | identify duplicates, sign errors, period mismatches, and missing support in a controlled extract | ledger structure and materiality rule | week 1; ordinary extract plus five seeded defects; spreadsheet allowed | issue log with row IDs, consequence, and proposed treatment | peer check, then data steward | all material seeded defects found; no valid record removed | new extract two weeks later without hints | no live personal data; escalate unreconciled opening total | finance data lead; quarterly | | F2 | traceable variance bridge for budget owner | explain change against budget | calculate and label price, volume, timing, and one-off effects; reconcile bridge to reported variance | F1 pass | week 2; incomplete commentary and one conflicting invoice date; AI may draft formulas | calculation workbook and evidence register | facilitator feedback; finance controller reviews logic | bridge reconciles; every material driver has evidence or an open-question label | novel cost center with different driver mix | AI output cannot override governed ledger; material conflict goes to controller | academy owner; after tool change | | F3 | bounded narrative for decision meeting | communicate what changed and what remains unknown | draft a 250-word explanation separating facts, interpretation, and action; cite decision-driving evidence | F2 pass | week 3; plausible but unsupported manager explanation included | narrative, source links, and verification log | budget-owner role play plus editor feedback | no unsupported material claim; action and uncertainty are clear | manager reviews a later workplace draft using the same rubric | no named employee inference; sensitive issue uses established route | finance communications lead; six months | | F4 | release-ready variance pack | decide whether the pack can be issued | integrate analysis and narrative, respond to challenge, and choose release, release-with-limitations, or return | F1–F3 | week 4 timed simulation; source conflict and late adjustment | final pack, exception log, disposition, and revision record | independent assessor | calculations reconcile, material claims trace, conflict treated, disposition justified | delayed unfamiliar case with reduced support | assessor reports observed evidence only; approval remains with finance owner | controller and learning lead; each cohort |

### What the example reveals

The map shows a progression from data integrity to calculation, communication, and release judgment. It also exposes a dependency: F3 should not be taught as fluent writing before learners can build the traceable calculation in F2. The transfer column prevents the programme from treating a coached final simulation as proof of workplace use. The ownership column makes content refresh a named responsibility rather than a vague annual intention.

Practitioner QA before launch

Run this check with the workflow owner, a subject-matter reviewer, a learning designer, and the relevant risk or accessibility reviewer.

  • [ ] Each row serves a named work outcome and user.
  • [ ] Objectives describe observable performance under stated conditions.
  • [ ] Practice cases include representative evidence and a meaningful failure condition.
  • [ ] Every assessable row creates an inspectable artifact.
  • [ ] Feedback occurs before the corresponding independent assessment.
  • [ ] Pass rules identify critical failures that cannot be offset by a total score.
  • [ ] Transfer checks use a new case or later setting with reduced support.
  • [ ] Tools, data, human review, escalation, retention, and permitted-use boundaries are explicit.
  • [ ] Access needs and alternative participation routes have been reviewed.
  • [ ] Sources, content owners, version, expiry, and change triggers are recorded.
  • [ ] Programme measures separate participation, learning, transfer, and operational outcomes.
  • [ ] Claims stay within the evidence actually collected.

Version history template

| Version | Effective date | Rows changed | Reason and evidence | Reviewed by | Reassessment needed? | | --- | --- | --- | --- | --- | --- | | 0.1 | | initial draft | needs assessment and workflow map | | | | 1.0 | | approved curriculum | pilot review completed | | | | | | | | | |

Keep old versions with their assessment artifacts when policy permits. If a changed pass rule would alter interpretation, do not silently compare results across versions.

Two Prune boundary

Two Prune can supply realistic assessments in which participants complete workplace simulations, navigate evidence, use AI support, verify outputs, and submit a business deliverable. Assessment evidence can inform an academy's baseline or transfer check when the design is aligned. Two Prune does not design every learning programme, certify a curriculum, prove workplace transfer, issue professional accreditation, or make employment decisions.

Sources

UK Government: Employer guide—what works for AI upskilling

  • CDC: Quality Training Standards
  • OECD: Building an AI-ready public workforce

Sources: [1] [2] [3]

Sources

  1. 1.UK Employer Guide · UK Government
  2. 2.CDC Quality Training Standards · Centers for Disease Control and Prevention
  3. 3.OECD: Building an AI-ready public workforce · www.oecd.org