Recognize where generative AI helps and where human judgment is needed.
Choose five recurring work tasks. Explain which could use AI, which should stay human-led, and why.
Find the right AI tools for any job, task or industry.AI SKILLS
49 practical learning guides. Search your role and choose an industry for a suggested learning order.
These are editorial learning suggestions, not mandatory job requirements. How we choose skills →
Recognize where generative AI helps and where human judgment is needed.
Choose five recurring work tasks. Explain which could use AI, which should stay human-led, and why.
Turn a vague request into a clear goal, useful context, constraints, and a reviewable output.
Write three versions of a prompt for the same work task. Compare the outputs against a checklist you define first.
Check facts, numbers, citations, and omissions before relying on an AI response.
Audit an AI-generated summary against the source document. Flag unsupported claims and write a corrected version.
Decide what information an AI tool may access and who may see its outputs.
Create a data-sharing checklist for one workflow and test it using fictional records.
Use AI to help explore data while checking definitions, calculations, and missing values.
Analyze a small public dataset, reproduce one AI-assisted calculation manually, and explain one limitation.
Connect an AI step to a repeatable process with checks, approvals, and a recovery path.
Build a sandbox workflow that drafts an update from sample records, pauses for review, and logs failed inputs.
Use a coding assistant to implement changes, then inspect and test the result.
Implement a small feature in a practice repository. Review the diff and add a regression test that fails before the fix.
Define representative tests and measure the errors that matter for a specific workflow.
Create a small labeled classification dataset. Compare two approaches and explain false positives and false negatives.
Review how untrusted input, tools, and permissions can affect an AI application.
Threat-model a fictional internal assistant. Test malicious document instructions with dummy data and record defenses.
Organize approved sources so an assistant can retrieve relevant information and cite it.
Build a small knowledge assistant over public documents. Test answerable, ambiguous, and unanswerable questions.
Select a focused business problem and define what a successful AI pilot would change.
Write a one-page pilot proposal for one workflow with a baseline, owner, cost estimate, and stop condition.
Help a team redesign work, learn new habits, and review the effects of AI adoption.
Design a small team pilot with a training exercise, feedback channel, and review meeting.
Define who approves AI use, reviews risks, and responds when a system causes problems.
Draft a lightweight use-case register for three fictional AI projects, including owners and review triggers.
Use public evidence to prepare account briefs and useful, accurate sales messages.
Prepare a brief for a fictional prospect using public company information, then write an outreach draft with checked claims.
Create support flows that use approved knowledge and hand over when they cannot help.
Create a sample FAQ assistant and test ordinary questions, missing information, and a request requiring a human.
Use AI for drafts and visual concepts while retaining editorial and brand control.
Produce three campaign concepts for a fictional product. Select one and document factual, accessibility, and rights checks.
Review demand forecasts and distinguish model output from planning assumptions.
Use historical sample demand data to compare a forecast with a naive baseline and describe the biggest planning risk.
Assess inspection or maintenance models against the conditions in which they will operate.
Design a defect-detection pilot using sample images. Separate test conditions and document how errors would be reviewed.
Translate engineering requirements into constraints and review generated alternatives.
Set up a non-production design study and compare two outcomes against documented requirements.
Use AI to organize evidence while inspecting the original material and its limitations.
Build an evidence table from three public documents. Compare an AI summary with the original claims.
Use AI to prepare learning activities while preserving teacher judgment and learner agency.
Adapt one lesson with an AI-assisted activity and a follow-up task that requires the learner’s own explanation.
Check extracted fields and summaries against the original document and retain a review trail.
Extract dates and obligations from a fictional agreement or administrative form and verify them against the original.
Examine whether recruitment workflows rely on appropriate evidence and meaningful human review.
Audit a fictional screening workflow using synthetic records. Identify questionable signals and redesign the review step.
Understand imagery, location, and time when preparing or reviewing spatial analysis.
Compare public imagery for one area over two dates and explain what cannot be inferred reliably.
Draft SQL with an assistant, then check it against the real schema and known results.
Ask an assistant for a monthly revenue query on a sample database. Then check row counts, null handling and duplicate joins against a hand-calculated total.
Use AI to interpret execution plans, then prove each suggested change with measurements.
Capture EXPLAIN ANALYZE output for a slow query on a test database, ask an assistant for three options, and measure each one.
Use AI to draft data models and migrations, then review constraints, locks and rollback.
Draft a schema for a small ordering system with AI, then run its migration on a test copy with sample data and undo it.
Connect AI tools to databases with limited permissions, and treat generated SQL as untrusted.
Create a read-only role on a test database, connect an assistant, and confirm that write attempts and restricted tables are refused.
Use AI to draft backup and recovery runbooks, then prove them with a real restore.
Back up a test database, delete a table, restore it by following an AI-drafted runbook, and correct whatever the runbook got wrong.
Use AI to draft transformations and data tests, then confirm they catch real faults.
Build one small transformation with tests on sample data, then add duplicate and missing rows and check that the tests catch them.
Use AI to draft Verilog, SystemVerilog or VHDL, then check it with lint, testbench simulation and synthesis reports. Passing your checks shows the design met your tests, not that it is correct in every case.
Solve ten HDL practice problems yourself, then ask an assistant for the same designs, compare results in simulation, and lint and synthesise the designs that pass.
Apply AI to testbenches, assertions, coverage and debug without weakening sign-off.
Take a small open-source block, have AI draft its testbench and assertions, then measure the coverage and bugs it actually finds.
Use AI for component selection, schematics and routing, then verify against datasheets, ERC and DRC.
Design a small sensor board with AI help in KiCad, then check every pinout, footprint and rail against the datasheets.
Use AI to write and debug embedded C/C++, then confirm it on target hardware and under real timing.
Have an assistant write a driver for a sensor on a dev board, then debug it on the board and record every correction.
Train, compress and deploy small models to microcontrollers, and measure accuracy, latency and power on the device.
Build a keyword-spotting or vibration model, deploy it to a dev board, and compare on-device results with the training metrics.
Use AI to draft bench-instrument scripts (SCPI and VISA) for simulated or low-energy, non-hazardous setups, with limits, read-only modes and human approval. Not a plant-control or functional-safety course.
Write a measurement script with AI help against a simulated instrument, add limits and logging, and stop there unless a qualified supervisor approves a low-energy bench test.
Define an ideal customer profile and validate company and business-contact evidence.
Build a ten-company sample prospect list from permitted public company data and document why each account qualifies.
Separate account-level signals from assumptions about an individual buyer.
Rank fictional accounts using fit and intent, then explain two plausible non-buying reasons for every signal.
Turn a commercial question into a research plan with defined population, method and evidence limits.
Write a research brief for a fictional manufactured product and triangulate three public demand indicators without inventing market size.
Use dated evidence to compare competitors and make reviewable positioning claims.
Compare three fictional suppliers using public product facts and draft a battlecard with an explicit unknowns section.
Review AI-generated questions and results for bias, sampling limitations and unsupported generalizations.
Draft a ten-question customer survey, manually revise biased wording, and audit an AI summary against a synthetic response set.
Check pipeline summaries and forecast claims against historical deal evidence.
Evaluate a forecast on synthetic historical opportunities and trace each deal-risk summary back to its input records.
Check AI-matched transactions, suggested journal entries and variance explanations before they reach the ledger.
Reconcile a synthetic bank statement against a sample ledger with an AI tool, then review every suggested match and record which you rejected and why.
Use AI for sampling, tie-outs and anomaly flags while keeping evidence sufficient, appropriate and documented.
Run an AI anomaly check on a fictional general ledger, re-perform a sample of the flagged and unflagged items, and write the workpaper.
Use AI to find and summarise tax positions, then confirm every answer against current law, rules and official guidance.
Ask an assistant three GST, VAT or income tax questions on a fictional client, then verify each answer in the official law and guidance and correct it.
Use AI to draft three-statement, DCF and budget models, then test their structure, formulas and assumptions.
Have an AI tool draft a three-statement model for a fictional company, then audit it for hard-codes, broken links and balance errors before running sensitivities.
Use AI for filings, transcripts and screening while keeping sources traceable, inputs lawful and conclusions your own.
Use an assistant to summarise a public annual report and earnings call, then verify each figure and claim and write a short note with sources.
Use AI meeting notes, plan drafts and onboarding tools while learning to review the suitability, record-keeping and disclosure duties that apply in your jurisdiction.
With every participant’s permission, record a role-played client review, generate AI notes and a plan summary, then correct them and mark which records your regulator may require you to keep.
Validate AI credit scores and spreading, explain decisions to borrowers and manage model risk.
Using synthetic data only, score a sample loan book with a simple model, explain five decisions in plain language and compare approval rates across groups.