Twelve chapters and sector cases on AI prioritisation, data protection and measurable adoption planning. Synthetic training cases and data; not real ISHRAQA7 performance.
Twelve-chapter curriculum
- Assessing workplace AI use
- Administration and office tasks
- Digital marketing and campaigns
- Communication and professional writing
- Sales and customer support
- Project management and coordination
- Data analysis and reporting
- HR and sensitive decisions
- Education, training and learning
- Technical trades, construction and services
- Governance and measuring value
- Capstone: responsible AI adoption plan
1. Assessing workplace AI use
A fictional SME weighs automated customer replies against internal meeting notes.
Method and workflow — TRACE method
AI adoption starts with mapping work, not buying a product. Rank activities by value, frequency, data sensitivity and risk to identify measurable reversible trials. The outcome of this chapter is observable: Process map and justified priority matrix. A professional result keeps evidence, assumptions and decision owners visible. Any assertion without accessible proof remains pending rather than being inferred from polished language.
Interview four fictional department leads. Map current tasks, bottlenecks, data quality and human decisions. Score six ideas on an impact/risk matrix and select one low-risk pilot. Check scope, input quality, transformations, output format and approval in that order. Preserve the first attempt and justify each edit. Never hide unresolved cases simply to make the presentation look complete.
Example prompt
Independent exercise
Work through this scenario: A fictional SME weighs automated customer replies against internal meeting notes. Deliver process map and justified priority matrix. Create two versions using the same source data, check consistency, log discrepancies and name the required approving role.
Do not claim productivity gains without before-and-after evidence.
Knowledge check
- Inputs can be omitted.
- Deliver: Process map and justified priority matrix.
- An unverified answer is official policy.
- Change evidence during a comparison.
- Ignore missing cases for speed.
- Follow the method: Interview four fictional department leads. Map current tasks, bottlenecks, data quality and human de
- Apply this safeguard: Do not claim productivity gains without before-and-after evidence.
- Run without authorisation.
- Publish the first output immediately.
2. Administration and office tasks
An admin team records three approved decisions with no deadline.
Method and workflow — TRACE method
AI-assisted office work can structure notes and procedures faster, but dates, recipients, commitments and figures must come from approved sources. Sending and filing remain supervised. The outcome of this chapter is observable: Meeting-minutes template and review checklist. A professional result keeps evidence, assumptions and decision owners visible. Any assertion without accessible proof remains pending rather than being inferred from polished language.
From fictional notes, produce minutes separating decisions, actions and open questions. Include owner and deadline fields but keep unsupported values blank. Have a reviewer approve before distribution. Check scope, input quality, transformations, output format and approval in that order. Preserve the first attempt and justify each edit. Never hide unresolved cases simply to make the presentation look complete.
Example prompt
Independent exercise
Work through this scenario: An admin team records three approved decisions with no deadline. Deliver meeting-minutes template and review checklist. Create two versions using the same source data, check consistency, log discrepancies and name the required approving role.
An invented date can become an unintended commitment.
Knowledge check
- An unverified answer is official policy.
- Inputs can be omitted.
- Deliver: Meeting-minutes template and review checklist.
- Follow the method: From fictional notes, produce minutes separating decisions, actions and open questions. Include owne
- Change evidence during a comparison.
- Ignore missing cases for speed.
- Publish the first output immediately.
- Apply this safeguard: An invented date can become an unintended commitment.
- Run without authorisation.
3. Digital marketing and campaigns
A fictional brand launches a free guide without performance statistics.
Method and workflow — TRACE method
AI can propose creative variants, not marketing evidence or image rights. Campaign performance must be measured with approved real campaign data. The outcome of this chapter is observable: Campaign brief, three messages and claims register. A professional result keeps evidence, assumptions and decision owners visible. Any assertion without accessible proof remains pending rather than being inferred from polished language.
Define audience, channel, message, legal constraint and call to action. Produce three variants, fact-check claims and propose an A/B plan without inventing conversions. Check scope, input quality, transformations, output format and approval in that order. Preserve the first attempt and justify each edit. Never hide unresolved cases simply to make the presentation look complete.
Example prompt
Independent exercise
Work through this scenario: A fictional brand launches a free guide without performance statistics. Deliver campaign brief, three messages and claims register. Create two versions using the same source data, check consistency, log discrepancies and name the required approving role.
Never fabricate sales or performance promises.
Knowledge check
- Deliver: Campaign brief, three messages and claims register.
- An unverified answer is official policy.
- Inputs can be omitted.
- Ignore missing cases for speed.
- Follow the method: Define audience, channel, message, legal constraint and call to action. Produce three variants, fact
- Change evidence during a comparison.
- Run without authorisation.
- Publish the first output immediately.
- Apply this safeguard: Never fabricate sales or performance promises.
4. Communication and professional writing
A fictional organisation changes a workshop schedule without a confirmed new time.
Method and workflow — TRACE method
Good communication adapts tone to readers without changing factual meaning. AI provides drafts but does not replace editorial review or accountability. The outcome of this chapter is observable: Three aligned messages with open issues. A professional result keeps evidence, assumptions and decision owners visible. Any assertion without accessible proof remains pending rather than being inferred from polished language.
Prepare one technical note and adapt it into an internal announcement, FAQ and short message. Make an invariant-facts table and check every variation for newly invented commitments. Check scope, input quality, transformations, output format and approval in that order. Preserve the first attempt and justify each edit. Never hide unresolved cases simply to make the presentation look complete.
Example prompt
Independent exercise
Work through this scenario: A fictional organisation changes a workshop schedule without a confirmed new time. Deliver three aligned messages with open issues. Create two versions using the same source data, check consistency, log discrepancies and name the required approving role.
Polite wording must never conceal uncertainty.
Knowledge check
- Inputs can be omitted.
- Deliver: Three aligned messages with open issues.
- An unverified answer is official policy.
- Change evidence during a comparison.
- Ignore missing cases for speed.
- Follow the method: Prepare one technical note and adapt it into an internal announcement, FAQ and short message. Make a
- Apply this safeguard: Polite wording must never conceal uncertainty.
- Run without authorisation.
- Publish the first output immediately.
5. Sales and customer support
A fictional customer disputes a charge not shown on the attached invoice.
Method and workflow — TRACE method
Reliable support recognises intent and prepares drafts from approved policies. Ambiguous requests, disputes and personal data must be escalated to humans. The outcome of this chapter is observable: Triage grid, three draft replies and escalation triggers. A professional result keeps evidence, assumptions and decision owners visible. Any assertion without accessible proof remains pending rather than being inferred from polished language.
Classify synthetic requests as information, support or complaint, plus needs-review. Draft unsent responses and record classification reasons. Compare against a reference-labelled set. Check scope, input quality, transformations, output format and approval in that order. Preserve the first attempt and justify each edit. Never hide unresolved cases simply to make the presentation look complete.
Example prompt
Independent exercise
Work through this scenario: A fictional customer disputes a charge not shown on the attached invoice. Deliver triage grid, three draft replies and escalation triggers. Create two versions using the same source data, check consistency, log discrepancies and name the required approving role.
Do not invent prices, refunds or contractual rights.
Knowledge check
- An unverified answer is official policy.
- Inputs can be omitted.
- Deliver: Triage grid, three draft replies and escalation triggers.
- Follow the method: Classify synthetic requests as information, support or complaint, plus needs-review. Draft unsent re
- Change evidence during a comparison.
- Ignore missing cases for speed.
- Publish the first output immediately.
- Apply this safeguard: Do not invent prices, refunds or contractual rights.
- Run without authorisation.
6. Project management and coordination
A fictional SME schedules training without knowing trainer availability.
Method and workflow — TRACE method
AI-assisted planning turns goals into tasks, dependencies and risks. A generated schedule without capacity and constraints is not a credible commitment. The outcome of this chapter is observable: Project plan with assumptions, dependencies and risks. A professional result keeps evidence, assumptions and decision owners visible. Any assertion without accessible proof remains pending rather than being inferred from polished language.
Break a fictional project into eight tasks with owners, dependencies, effort assumptions and blockers. Identify likely critical dependencies and have an owner approve durations before release. Check scope, input quality, transformations, output format and approval in that order. Preserve the first attempt and justify each edit. Never hide unresolved cases simply to make the presentation look complete.
Example prompt
Independent exercise
Work through this scenario: A fictional SME schedules training without knowing trainer availability. Deliver project plan with assumptions, dependencies and risks. Create two versions using the same source data, check consistency, log discrepancies and name the required approving role.
AI-suggested duration is an assumption, not measured data.
Knowledge check
- Deliver: Project plan with assumptions, dependencies and risks.
- An unverified answer is official policy.
- Inputs can be omitted.
- Ignore missing cases for speed.
- Follow the method: Break a fictional project into eight tasks with owners, dependencies, effort assumptions and blocker
- Change evidence during a comparison.
- Run without authorisation.
- Publish the first output immediately.
- Apply this safeguard: AI-suggested duration is an assumption, not measured data.
7. Data analysis and reporting
A fictional transaction table contains a cancelled row excluded from active totals.
Method and workflow — TRACE method
Data summaries require an explicit scope, period and denominator. AI-generated charts and tables must be recomputed before decisions rely on them. The outcome of this chapter is observable: KPI report with reconciliation and scope notes. A professional result keeps evidence, assumptions and decision owners visible. Any assertion without accessible proof remains pending rather than being inferred from polished language.
Use the twelve-row practice CSV. State units and exclusions, recompute totals and generate two KPIs. Compare with an independent calculation and document missing entries. Check scope, input quality, transformations, output format and approval in that order. Preserve the first attempt and justify each edit. Never hide unresolved cases simply to make the presentation look complete.
Example prompt
Independent exercise
Work through this scenario: A fictional transaction table contains a cancelled row excluded from active totals. Deliver kpi report with reconciliation and scope notes. Create two versions using the same source data, check consistency, log discrepancies and name the required approving role.
Never describe training numbers as ISHRAQA7 performance.
Knowledge check
- Inputs can be omitted.
- Deliver: KPI report with reconciliation and scope notes.
- An unverified answer is official policy.
- Change evidence during a comparison.
- Ignore missing cases for speed.
- Follow the method: Use the twelve-row practice CSV. State units and exclusions, recompute totals and generate two KPIs.
- Apply this safeguard: Never describe training numbers as ISHRAQA7 performance.
- Run without authorisation.
- Publish the first output immediately.
8. HR and sensitive decisions
A fictional model ranks applicants using personal characteristics.
Method and workflow — TRACE method
In HR, AI may draft job descriptions or training rubrics, but individual decisions remain human and subject to applicable rules. Bias and personal information require heightened care. The outcome of this chapter is observable: Job-relevant skills rubric with human-review note. A professional result keeps evidence, assumptions and decision owners visible. Any assertion without accessible proof remains pending rather than being inferred from polished language.
Draft a neutral job description and observable-skill rubric. Remove criteria unrelated to the job and have the rubric reviewed before real-world use. Check scope, input quality, transformations, output format and approval in that order. Preserve the first attempt and justify each edit. Never hide unresolved cases simply to make the presentation look complete.
Example prompt
Independent exercise
Work through this scenario: A fictional model ranks applicants using personal characteristics. Deliver job-relevant skills rubric with human-review note. Create two versions using the same source data, check consistency, log discrepancies and name the required approving role.
Do not delegate sensitive screening or employment decisions to a model.
Knowledge check
- An unverified answer is official policy.
- Inputs can be omitted.
- Deliver: Job-relevant skills rubric with human-review note.
- Follow the method: Draft a neutral job description and observable-skill rubric. Remove criteria unrelated to the job an
- Change evidence during a comparison.
- Ignore missing cases for speed.
- Publish the first output immediately.
- Apply this safeguard: Do not delegate sensitive screening or employment decisions to a model.
- Run without authorisation.
9. Education, training and learning
A fictional academy wants a SQL quiz without knowing learner levels.
Method and workflow — TRACE method
AI tools can vary explanations and formative feedback, but plausible text is not necessarily correct or instructionally sound. Good exercises need objectives, checked solutions and teacher oversight. The outcome of this chapter is observable: Corrected mini-quiz, level tags and educational review sheet. A professional result keeps evidence, assumptions and decision owners visible. Any assertion without accessible proof remains pending rather than being inferred from polished language.
Pick a skill and write three questions of increasing difficulty. Prepare verified model answers and marking rules before testing whether a model follows the syllabus. Check scope, input quality, transformations, output format and approval in that order. Preserve the first attempt and justify each edit. Never hide unresolved cases simply to make the presentation look complete.
Example prompt
Independent exercise
Work through this scenario: A fictional academy wants a SQL quiz without knowing learner levels. Deliver corrected mini-quiz, level tags and educational review sheet. Create two versions using the same source data, check consistency, log discrepancies and name the required approving role.
Never invent an official certification or accreditation.
Knowledge check
- Deliver: Corrected mini-quiz, level tags and educational review sheet.
- An unverified answer is official policy.
- Inputs can be omitted.
- Ignore missing cases for speed.
- Follow the method: Pick a skill and write three questions of increasing difficulty. Prepare verified model answers and
- Change evidence during a comparison.
- Run without authorisation.
- Publish the first output immediately.
- Apply this safeguard: Never invent an official certification or accreditation.
10. Technical trades, construction and services
A fictional inspection notes a missing barrier without the safety file.
Method and workflow — TRACE method
In technical services, assistants may structure reports and checklists, but generated text cannot replace engineering calculations, standards, contractual plans or qualified judgment. The outcome of this chapter is observable: Site report with open issues and sign-off route. A professional result keeps evidence, assumptions and decision owners visible. Any assertion without accessible proof remains pending rather than being inferred from polished language.
From a fictional site visit, draft a report separating observation, risk, owner and action. Flag safety and compliance matters for a qualified professional. Check scope, input quality, transformations, output format and approval in that order. Preserve the first attempt and justify each edit. Never hide unresolved cases simply to make the presentation look complete.
Example prompt
Independent exercise
Work through this scenario: A fictional inspection notes a missing barrier without the safety file. Deliver site report with open issues and sign-off route. Create two versions using the same source data, check consistency, log discrepancies and name the required approving role.
AI alone must not decide critical safety instructions.
Knowledge check
- Inputs can be omitted.
- Deliver: Site report with open issues and sign-off route.
- An unverified answer is official policy.
- Change evidence during a comparison.
- Ignore missing cases for speed.
- Follow the method: From a fictional site visit, draft a report separating observation, risk, owner and action. Flag saf
- Apply this safeguard: AI alone must not decide critical safety instructions.
- Run without authorisation.
- Publish the first output immediately.
11. Governance and measuring value
An executive wants to claim 50% gains without baseline measurements.
Method and workflow — TRACE method
AI initiatives need an owner, quality metrics, data rules and stop procedures. Claimed benefits require a before/after design with comparable periods and populations. The outcome of this chapter is observable: Metric dashboard with assumptions and decision thresholds. A professional result keeps evidence, assumptions and decision owners visible. Any assertion without accessible proof remains pending rather than being inferred from polished language.
Design a fictional four-week pilot. Measure processing time, errors and revisions with sample size and review cost. Define stop criteria before the experiment. Check scope, input quality, transformations, output format and approval in that order. Preserve the first attempt and justify each edit. Never hide unresolved cases simply to make the presentation look complete.
Example prompt
Independent exercise
Work through this scenario: An executive wants to claim 50% gains without baseline measurements. Deliver metric dashboard with assumptions and decision thresholds. Create two versions using the same source data, check consistency, log discrepancies and name the required approving role.
ROI is a hypothesis until measurements exist.
Knowledge check
- An unverified answer is official policy.
- Inputs can be omitted.
- Deliver: Metric dashboard with assumptions and decision thresholds.
- Follow the method: Design a fictional four-week pilot. Measure processing time, errors and revisions with sample size a
- Change evidence during a comparison.
- Ignore missing cases for speed.
- Publish the first output immediately.
- Apply this safeguard: ROI is a hypothesis until measurements exist.
- Run without authorisation.
12. Capstone: responsible AI adoption plan
A fictional organisation wants AI adoption without reducing output reliability.
Method and workflow — TRACE method
A credible strategy starts with bounded business problems, identifies risks and defines a limited pilot with training, governance and independent measurement. It is not a shopping list of AI tools. The outcome of this chapter is observable: Diagnosis, matrices, pilot plan, logs, timeline and approval dossier. A professional result keeps evidence, assumptions and decision owners visible. Any assertion without accessible proof remains pending rather than being inferred from polished language.
Deliver a fictional services firm adoption plan: six-task inventory, impact/risk matrix, pilot design, privacy log, baseline metrics and conditional scale-up decision. Check scope, input quality, transformations, output format and approval in that order. Preserve the first attempt and justify each edit. Never hide unresolved cases simply to make the presentation look complete.
Example prompt
Independent exercise
Work through this scenario: A fictional organisation wants AI adoption without reducing output reliability. Deliver diagnosis, matrices, pilot plan, logs, timeline and approval dossier. Create two versions using the same source data, check consistency, log discrepancies and name the required approving role.
A negative pilot result is still valuable evidence and must be recorded.
Knowledge check
- Deliver: Diagnosis, matrices, pilot plan, logs, timeline and approval dossier.
- An unverified answer is official policy.
- Inputs can be omitted.
- Ignore missing cases for speed.
- Follow the method: Deliver a fictional services firm adoption plan: six-task inventory, impact/risk matrix, pilot desig
- Change evidence during a comparison.
- Run without authorisation.
- Publish the first output immediately.
- Apply this safeguard: A negative pilot result is still valuable evidence and must be recorded.
