Country
You are in the Nigeria edition. Type your country to switch.
Nigeria
Loading every country…
No country matches. Check the spelling, or choose "Any country" for the Global edition.
Your learning profile
Learning path
AI at work
Practical, tool-neutral AI skills for work: prompt engineering, everyday productivity, and using AI responsibly in finance, accounting, design, content, marketing, data analysis and small business, with ethics, security and agents.
Start the pathBeginner
- AI at workAI ethics matters at workRead nowWhy AI ethics matters at work
AI tools can cause real harm at work: unfair decisions, privacy loss, misinformation and safety problems. This lesson sets out the principles that are widely published around the world, shows how ethics differs from law, and explains the part every employee plays.
- AI at workBias and fairnessRead nowBias and fairness
Bias can enter an AI system through its data, its labels, its design and the way people use it. This lesson shows simple fairness measures with worked numbers, explains why different definitions of fairness can conflict, and gives practical checks for ordinary teams.
- AI at workTransparency and explainabilityRead nowTransparency and explainability
People deserve to know when AI is involved in something that affects them, and to get an explanation they can understand. This lesson covers disclosure, plain-language explanations, documentation such as model cards, and the honest limits of explaining complex models.
- AI at workPrivacy and consentRead nowPrivacy and consent
AI raises privacy questions at every stage: the personal data used to train systems, the data people paste into prompts, monitoring at work, and the sensitive facts a system can infer. This lesson explains consent and its limits and shows privacy by default in practice. Rules differ by country.
- AI at workAccountability and human oversightRead nowAccountability and human oversight
When an AI system gets something wrong, a person or organisation is still responsible. This lesson explains human in the loop, on the loop and in command, the difference between meaningful oversight and rubber-stamping, how to escalate, and how to build a responsibility table for one process.
- AI at workCreators, consent and intellectu…Read nowCreators, consent and intellectual property
Generative AI is trained on large amounts of creative work, which raises questions about consent, licensing, attribution and payment that are still being publicly debated. This lesson explains the main arguments, the risks of imitating a style or a likeness, and respectful practice for users. Education, not legal advice.
- AI at workEnvironmental and social impactsRead nowEnvironmental and social impacts
AI runs on data centres that use energy and water, and on human work that is often unseen. This lesson explains these costs in general terms, shows how ordinary choices change them with an illustrative calculation, and looks at labour, access and inclusion.
- AI at workGovernance in organisationsRead nowGovernance in organisations
Good intentions need structure. This lesson shows how organisations govern AI: a short policy, a register of uses, risk tiers, impact assessments, approvals and review, with frameworks and standards as publicly described and a worked three-tier example.
- AI at workMaking ethical decisions about AIRead nowMaking ethical decisions about AI
This lesson brings the course together in a step-by-step framework: purpose, people affected, risks, alternatives, safeguards, decision and review. It applies the framework to two workplace cases and explains how to speak up when something feels wrong.
- AI at worklanguage models work, at a usefu…Read nowHow language models work, at a useful level
A language model writes by predicting one likely piece of text after another. Knowing about tokens, the context window and temperature explains most of the behaviour you will meet, and why a fluent answer still needs checking.
- AI at workThe anatomy of a good promptRead nowThe anatomy of a good prompt
A good prompt has six parts: the task, the context, examples, the format, the constraints and a test of success. This lesson rewrites a weak prompt step by step and gives you a checklist to reuse with any assistant.
- AI at workGiving context and clear instruc…Read nowGiving context and clear instructions
Assistants do their best work when they know the audience, the purpose and the background, and when pasted material is clearly marked off from your instructions. This lesson covers specificity, delimiters, standing instructions and why saying what to do beats a list of what not to do.
- AI at workExamples: zero-shot, one-shot an…Read nowExamples: zero-shot, one-shot and few-shot prompting
Sometimes showing beats telling. This lesson explains zero-shot, one-shot and few-shot prompting, how to choose varied examples, the risk that the model copies surface features, and how to check accuracy on new items rather than trusting the examples.
- AI at workRoles, audience and toneRead nowRoles, audience and tone
Telling an assistant to answer as an editor or a tutor changes its vocabulary and perspective, but adds no expertise or facts. This lesson shows how to use roles honestly, write for a named audience, set tone and reading level, and put a short style sheet in the prompt.
- AI at workConstraints and output formatsRead nowConstraints and output formats
Constraints set the length, scope, structure and exclusions of an answer, and output formats make it usable: a table, a checklist or named fields that other tools can read. This lesson turns free text into a fixed table and shows how to check every field.
- AI at workAsking for step-by-step reasoningRead nowAsking for step-by-step reasoning
Complex questions go better when they are broken into steps. This lesson covers task decomposition, plan-then-answer prompts, asking for working you can check, why the written explanation may not be how the answer was produced, and why calculations belong in a tool.
- AI at workIteration and refinementRead nowIteration and refinement
The first answer is a draft. This lesson shows how to improve it with follow-up prompts that target one fault at a time, how to use critique-and-revise, when to start a fresh conversation, and how prompt chaining breaks a job into a sequence of smaller prompts.
- AI at workEvaluating prompts and outputsRead nowEvaluating prompts and outputs
One good answer proves nothing. This lesson shows how to judge prompts properly: a written rubric, a fixed test set with expected results, scoring two versions on the same cases, and spot-checks by a person.
- AI at workPrompt libraries and templatesRead nowPrompt libraries and templates
A prompt that works is worth keeping. This lesson turns one into a template with square-bracket fields, then shows how a team library names, versions, owns, tests, shares and retires prompts.
- AI at workHallucination and groundingRead nowHallucination and grounding
Models produce confident errors and invented citations because they predict plausible text. This lesson shows how to ground answers in supplied documents, ask for quotes and locations, give permission to say not in the source, and outlines retrieval-augmented generation.
- AI at workPrompt injection and other failu…Read nowPrompt injection and other failure modes
AI systems fail in recognisable ways. This lesson explains, defensively, how direct and indirect prompt injection work, and covers sycophancy, inconsistency between runs, over-refusal and bias, with the design response to each.
- AI at workWhere to startRead nowWhere to start
Before choosing any AI tool, find out where your week actually goes. A short time audit shows which tasks take the most hours, repeat the most and carry the least risk, and those are the best places for a small first trial.
- AI at workCustomer serviceRead nowCustomer service
AI can help a small business answer common questions faster, draft replies in a friendly tone and run a simple chat assistant. It must only repeat policies you have actually written, and it needs a clear route to hand a customer to a person.
- AI at workMarketing on a budgetRead nowMarketing on a budget
An AI assistant can draft product descriptions, social posts, local listing text and short emails, which may save a small business time. The owner keeps it true, in their own voice and on a realistic monthly plan.
- AI at workAdmin and finance helpRead nowAdmin and finance help
AI tools can help with invoices, payment reminders, sorting receipts and drafting a simple cash-flow forecast. Every figure still needs checking against your bank and your records, and tax and statutory accounts belong with a qualified accountant.
- AI at workDocuments, proposals and policiesRead nowDocuments, proposals and policies
AI assistants are useful for first drafts of proposals, quotes, simple internal policies and letters, especially from a good template. Contracts, terms and conditions and other legal documents need a qualified professional, because the wording carries legal weight.
- AI at workChoosing tools and working out t…Read nowChoosing tools and working out the costs
Choosing an AI tool means comparing free and paid tiers, reading the data terms, checking how easy it is to leave and counting every cost, not just the price. A simple cost-benefit sum shows whether a tool is likely to pay its way.
- AI at workA simple AI policy for a small t…Read nowA simple AI policy for a small team
A one-page AI policy tells a small team which tools are approved, what must never be pasted in, how outputs are checked, what customers are told and who to ask. It brings hidden use into the open and makes training simple.
- AI at workRisks and trustRead nowRisks and trust
AI makes some scams more convincing, from fake invoices to cloned voices that sound like the owner. A simple callback rule for payments, care with customer data and honest marketing protect a small business and the trust it depends on.
- AI at workWhat AI is good and bad at in ev…Read nowWhat AI is good and bad at in everyday work
AI assistants are strong at drafting, summarising, transforming and brainstorming, and weak at facts, arithmetic, recent events and judgement. This lesson gives you a simple grid of stakes against checkability so you can decide, task by task, whether an assistant is a good fit.
- AI at workWriting and editing with AIRead nowWriting and editing with AI
An assistant can turn rough notes into a first draft, rewrite for a different reader and cut a long paragraph down. This lesson shows real prompts and better versions, how to keep your own voice, and the edit pass you always do before anything goes out.
- AI at workResearch and summarisingRead nowResearch and summarising
Assistants can shrink a long document into a short brief, but a summary is only as good as its purpose and its checks. This lesson covers extractive and abstractive summaries, asking for page locations, lateral reading to test claims, and a worked one-page brief with a checking plan.
- AI at workEmail and meetingsRead nowEmail and meetings
Assistants can sort a full inbox, draft replies, build agendas and turn a meeting transcript into notes and action items. This lesson shows the prompts, the errors transcripts make, why you tell people before recording, and a worked example of six actions with owners and dates.
- AI at workSpreadsheets and analysis helpRead nowSpreadsheets and analysis help
Assistants can write and explain spreadsheet formulas if you describe the sheet clearly. This lesson shows how to describe your layout in a prompt, how to test a formula on rows you can check by hand, and a worked sales sheet where the totals and percentages add up.
- AI at workCoding help for non-developers a…Read nowCoding help for non-developers and developers
AI coding assistants can write small scripts and macros, explain unfamiliar code and suggest tests. This lesson shows how to ask well, why you read before you run, why secrets never go into a prompt, and the review steps every piece of AI-written code goes through.
- AI at workBuilding simple AI workflowsRead nowBuilding simple AI workflows
A recurring task becomes a workflow when you write down each step, decide where an assistant helps and where a person checks, and measure the result honestly. This lesson maps a weekly report, places human checkpoints and counts the checking time in the saving.
- AI at workVerifying AI outputsRead nowVerifying AI outputs
Every AI output needs checking, and the depth of the check should match the stakes. This lesson covers what to check first (names, numbers, dates, quotes and links), recomputing figures, the three-source habit, a reusable checklist, and a worked example that finds four errors in a short answer.
- AI at workPrivacy, confidentiality and com…Read nowPrivacy, confidentiality and company policy
What you paste into an AI tool may be stored, reviewed or used in ways you did not expect. This lesson covers what never goes in, how data classification works in practice, consumer and business settings in general terms, retention, acceptable use policies and when to ask. Rules differ by country and employer.
- AI at worknot to use AIRead nowWhen not to use AI
Some work should stay with people: high-stakes and irreversible decisions, decisions about people, regulated advice, grief and sensitive conversations, and skills you need to learn. This lesson explains why, covers deskilling and accountability, and applies a short decision checklist to five cases.
Intermediate
- AI at workAI across the accounting cycleRead nowAI across the accounting cycle
AI now touches every stage of the accounting cycle, from capturing invoices to closing the books, reporting and audit. This lesson maps where rules-based automation, machine learning and generative assistants each fit, and which decisions stay a professional judgement.
- AI at workBookkeeping automationRead nowBookkeeping automation
Automated bookkeeping reads invoices and receipts, suggests account codes and routes uncertain items to a person. This lesson follows a worked month through capture, confidence thresholds and the exception queue, and shows how to sample the items the tool coded on its own.
- AI at workReconciliations with AI matchingRead nowReconciliations with AI matching
Matching tools pair bank lines with ledger entries using exact rules, flexible rules and fuzzy matching. This lesson works through a bank reconciliation where every figure agrees, explains match rate, and shows why unmatched and fuzzy-matched items need a person.
- AI at workAnomaly detection in ledgersRead nowAnomaly detection in ledgers
Anomaly detection finds ledger records that differ from the usual pattern. This lesson compares rules, statistical tests and machine learning, works through a first-digit analysis with illustrative counts, and shows how to triage flags without treating them as findings.
- AI at workAudit analytics and full-populat…Read nowAudit analytics and full-population testing
Analytics let auditors test every transaction instead of a sample. This lesson explains journal entry testing criteria, follows a worked population from thousands of journals to a short list for follow-up, and shows how analytics results become audit evidence.
- AI at workDocumentation, memos and working…Read nowDocumentation, memos and working papers
Assistants can draft accounting memos, policies, close checklists and working paper narratives. This lesson shows how to prompt for a useful structure, which parts must remain the preparer's own judgement, and how to record AI use in the file.
- AI at workControls over AI outputsRead nowControls over AI outputs
When AI becomes part of a finance process, it needs controls like any other part. This lesson covers review and approval, thresholds, segregation of duties, logging, change management and periodic testing, and builds a control matrix with four rows.
- AI at workProfessional scepticism and judg…Read nowProfessional scepticism and judgement with AI
Professional scepticism means a questioning mind, and it applies to tool outputs as much as to client explanations. This lesson covers automation bias, corroborating evidence, documenting why a suggestion was accepted or rejected, and cases where a tool was confidently wrong.
- AI at workConfidentiality and client dataRead nowConfidentiality and client data
Accountants and auditors hold sensitive client information under widely published duties of confidentiality. This lesson covers data minimisation and redaction before using a tool, what to look for in the contracts and settings of business tools, client consent, and a worked redaction.
- AI at workWhere AI helps in content workRead nowWhere AI helps in content work
Content passes through a lifecycle from idea to review, and an AI assistant helps far more at some stages than others. This lesson maps where it fits, what it tends to do badly, and what the human editor must still own.
- AI at workDrafting with a clear voiceRead nowDrafting with a clear voice
Good AI-assisted drafts start with an agreed outline and a clear description of voice. This lesson shows how to draft section by section, give voice notes and samples in the prompt, and remove generic phrasing.
- AI at workEditing and styleRead nowEditing and style
AI can help with structural, line and copy editing, and with consistency checks against a style sheet. This lesson shows what to ask for, how to keep control of each change, and what an editor must still do.
- AI at workVideo and audioRead nowVideo and audio
AI can help with scripts, shot lists, rough cuts, captions and translation. This lesson covers those uses, why synthetic voices need consent, why music and footage rights still apply, and how to plan a short video whose timings add up.
- AI at workRepurposing contentRead nowRepurposing content
One substantial piece can become many smaller ones for different formats and channels. This lesson shows how to adapt rather than copy, fit each platform, and plan a worked example that turns one webinar into nine pieces with checks.
- AI at workSearch-aware draftingRead nowSearch-aware drafting
Search-aware drafting means understanding what a searcher wants and answering it clearly. This lesson covers search intent, people-first content, structure and headings, and why mass-produced low-value pages are a risk under published search-quality guidance.
- AI at workFact-checking AI-assisted contentRead nowFact-checking AI-assisted content
AI-assisted drafts can contain invented quotes, statistics, names and dates. This lesson shows how to list every claim, check primary sources first, keep a claim log, and work through a check of twelve claims.
- AI at workDisclosure and platform policiesRead nowDisclosure and platform policies
Many platforms and publishers ask creators to label AI-generated or substantially altered content, especially realistic images, video and audio. This lesson describes those rules in general published terms, along with sponsored content and endorsements. It is education, not legal advice.
- AI at workVoice and originalityRead nowVoice and originality
AI drafts tend to converge on the average of what already exists. This lesson explains why, shows how first-hand experience, data and reporting add originality, and covers plagiarism, near-copying and keeping a portfolio honest.
- AI at workA content production system with…Read nowA content production system with AI
A content team using AI needs a system: briefs, a calendar, roles, review stages, a sign-off rule and metrics that matter. This lesson builds a weekly production plan whose hours add up, and wraps up the course.
- AI at workThe analysis workflow with AIRead nowThe analysis workflow with AI
Good analysis follows six steps: question, data, cleaning, analysis, checking and communication. This lesson shows where an AI assistant helps at each step, how to describe a data set precisely without sharing it, and how to run a first exploratory pass.
- AI at workCleaning data with AI helpRead nowCleaning data with AI help
Messy data produces confident wrong answers. This lesson covers duplicates, inconsistent labels, types and dates and missing values, how to ask an assistant for cleaning steps you can repeat, and how to log every change, with a worked cleaning of a sheet of orders.
- AI at workSpreadsheet formulas with AIRead nowSpreadsheet formulas with AI
An assistant can draft lookups, conditional sums and counts, and text and date formulas in seconds, but only a precise description of your columns gets a formula that fits. This lesson shows how to prompt for formulas, how to test them on rows you can verify by hand and how to explain formulas you inherit.
- AI at workPivot tables and summariesRead nowPivot tables and summaries
Pivot tables summarise data by category and period, and an assistant can help you design them. This lesson shows how to check every summary against a control total from the source, so regional totals add up to the grand total and percentages of total add up to a hundred.
- AI at workNatural-language questions and SQLRead nowNatural-language questions and SQL
Many data tools now let you ask questions in plain words and turn them into SQL. This lesson shows how to describe tables precisely, read a generated query before trusting it, spot joins that double-count and check results with row counts and totals.
- AI at workCharts and BI dashboardsRead nowCharts and BI dashboards
A good chart answers one question honestly. This lesson covers choosing the right chart, honest axes, dashboards built around decisions and how to check a chart an assistant suggests against the data behind it.
- AI at workStatistical sanity checksRead nowStatistical sanity checks
AI assistants can produce insights faster than anyone can check them. This lesson gives you the sanity checks to apply: mean against median, outliers, correlation against causation, small samples and Simpson's paradox, with a worked example that adds up.
- AI at workData storytellingRead nowData storytelling
A finding only matters if someone understands and acts on it. This lesson shows how to turn a result into a message with a clear so-what, write a one-page insight summary with the evidence and its limits, and use an assistant to draft without overclaiming.
- AI at workPrivacy and reproducible analysisRead nowPrivacy and reproducible analysis
Using an assistant on real data raises two questions: is it safe to share, and could someone repeat what you did? This lesson covers describing, aggregating and synthesising data instead of sharing it, keeping prompts, code and steps, and writing an analysis log, then wraps up the course.
- AI at workimage models workRead nowHow image models work
Image generators start from random noise and remove it step by step, guided by your words. Knowing how that works explains seeds, image-to-image, aspect ratio, and why lettering, hands and exact counts so often go wrong.
- AI at workPrompting for imagesRead nowPrompting for images
A strong image prompt names the subject, setting, composition, lighting, style, medium and frame. This lesson improves a weak prompt over four rounds, shows how to describe a style by its features rather than an artist's name, and when to use negative prompts and reference images.
- AI at workIdeation and mood boardsRead nowIdeation and mood boards
AI makes it cheap to see many directions quickly, but choosing well still takes a brief and judgement. This lesson pairs divergent and convergent thinking, uses mood boards as conversation tools, and narrows a worked round of thumbnails to three concept directions.
- AI at workImage editing with AIRead nowImage editing with AI
Inpainting, outpainting, background removal, upscaling and retouching can save time on routine edits. This lesson shows how to use them non-destructively, why originals must be kept, and where honesty limits apply to news and product images.
- AI at workLayout and interface ideasRead nowLayout and interface ideas
An assistant can turn a written brief into layout options, draft interface copy and suggest components that fit an existing design system. This lesson turns a brief into three wireframe options and explains why every AI layout still needs testing with real users.
- AI at workAccessibility with AI helpRead nowAccessibility with AI help
AI can draft alt text, captions and plain-language versions, but the published guidelines are the test. This lesson works through the contrast formula used in accessibility guidelines with real numbers and shows how to review AI drafts.
- AI at workKeeping a brand consistentRead nowKeeping a brand consistent
Generated work drifts from a brand unless guidelines go into the prompt and every output is checked. This lesson turns brand guidelines into prompt inputs, uses style references and consistent characters, and applies a brand review checklist.
- AI at workCopyright, licensing and disclos…Read nowCopyright, licensing and disclosure
Who owns AI-assisted work, what the tool's licence allows, and when to disclose AI use are open questions that differ by country and change over time. This lesson describes the main issues as publicly described and gives a pre-delivery checklist. Education, not legal advice.
- AI at workA professional AI design workflowRead nowA professional AI design workflow
A professional workflow puts AI inside a process the designer controls: brief, research, ideation, production, checks and handover. This lesson walks through a worked job, shows time and cost honestly, and covers how to talk to clients about AI.
- AI at workWhere AI fits in financial workRead nowWhere AI fits in financial work
AI already sits in many corners of financial work, from research notes to fraud alerts. This lesson maps the main uses across research, operations, client service, risk and compliance, separates generative tools from predictive models, and shows how a use case inventory keeps every use owned, understood and tiered by risk.
- AI at workSummarising filings, reports and…Read nowSummarising filings, reports and earnings calls
AI assistants can turn long annual reports, results statements and call transcripts into short summaries and tables. This lesson shows how to ask for figures with page references, how to reconcile every extracted number with the source, and how to keep the tone of management's words separate from what the numbers say.
- AI at workAnalysing financial data with AI…Read nowAnalysing financial data with AI, and checking it
Language models predict words, so arithmetic written out in text is unreliable. This lesson shows where calculations belong, how to compute and recompute ratios and growth rates, how to catch errors in units and periods, and works through three ratios step by step with an independent recalculation check.
- AI at workForecasting: what AI can and can…Read nowForecasting: what AI can and cannot do
Language models do not predict markets, and any tool that offers exact future prices should raise a warning. This lesson explains why, covers base rates, changing conditions and spurious precision, and shows how assistants can help build scenarios and ranges that are honest about uncertainty.
- AI at workNews and sentiment monitoringRead nowNews and sentiment monitoring
AI tools can sort news flow, score sentiment and send alerts, which helps teams read what matters first. This lesson covers news triage, sentiment scoring and its pitfalls, stale data and knowledge cut-offs, false signals, and why the original release is always the final check.
- AI at workRisk and compliance usesRead nowRisk and compliance uses
Risk and compliance teams use AI to support customer checks, screen names, monitor transactions, review communications and answer policy questions. This lesson explains false positives and false negatives with numbers that add up, shows how precision and alert fatigue are linked, and why the decisions themselves stay with people.
- AI at workModel risk for generative AIRead nowModel risk for generative AI
Banks and other financial firms have long managed model risk for credit and pricing models. This lesson shows how the same ideas apply to generative tools: an inventory, validation before use, ongoing monitoring, change control when a vendor updates a model, and clear limits on what each tool may be used for.
- AI at workExplainability and audit trailsRead nowExplainability and audit trails
When AI helps with a decision, someone may later ask how it was made: a client, a manager, an auditor or a supervisor. This lesson shows how to explain an AI-assisted decision in plain words and how to keep a decision log that records the prompts, sources, outputs and the human judgement behind it.
- AI at workRegulators' expectations, as pub…Read nowRegulators' expectations, as publicly described
Financial supervisors in many countries have published statements on AI. This lesson summarises the common themes as publicly described: firms stay accountable, customers are treated fairly, data is protected, outside suppliers are managed, and use is governed and tested. It outlines the risk-based approach of the EU AI Act. It is education, not legal advice.
- AI at workA responsible AI workflow for fi…Read nowA responsible AI workflow for finance professionals
This lesson puts the course together in one worked example: a research brief on an invented company, from scope and sources to extraction, checks, human judgement, disclosure of AI use and conflicts. It shows the time each step took and where every check sits, and keeps the work education, not advice.
- AI at workWhere AI fits in marketingRead nowWhere AI fits in marketing
AI appears at every stage of marketing, from customer insight to measurement, but in different forms. This lesson separates rules-based automation, predictive analytics and generative AI, shows where each fits and what customers notice when it goes right or wrong.
- AI at workAudience research with AIRead nowAudience research with AI
Assistants can read thousands of reviews, survey answers and support tickets quickly and suggest themes. This lesson shows how to code comments with AI help, check the labels with people, report counts that add up, and why synthetic personas are never real customers.
- AI at workCampaign ideas and copyRead nowCampaign ideas and copy
AI can turn a brief into many creative routes and copy variants in minutes. This lesson shows how to prompt for distinct routes, narrow variants down, substantiate every claim and replace misleading wording with claims you can support.
- AI at workPersonalisationRead nowPersonalisation
Personalisation tailors content to people or groups using data about them. This lesson covers segments, recommendations and dynamic content, puts consent and preferences first, works through a segment plan that adds up and draws the line between helpful and intrusive.
- AI at workTesting and optimisationRead nowTesting and optimisation
Controlled tests show what a change really does. This lesson works through a test with real arithmetic, explains why testing many AI variants multiplies false winners, outlines how multi-armed bandits shift traffic and shows how to use assistants for hypotheses rather than verdicts.
- AI at workEmail and customer dataRead nowEmail and customer data
Email tools now suggest subject lines, pick send times, score leads and predict churn. This lesson explains each, works through a lead-scoring table and a churn model's hits and misses, and shows why data quality and consent decide whether any of it is worth using.
- AI at workAd creative and automated media …Read nowAd creative and automated media buying
Generated creative and automated bidding let small teams run many ads at once. This lesson explains what you still control, how brand safety and brand suitability differ, how to split a budget that adds up and leaves room to learn, and what to check before launch.
- AI at workMeasuring AI-driven marketingRead nowMeasuring AI-driven marketing
Dashboards count what they can track, not what the marketing caused. This lesson explains the limits of attribution, works through a holdout test to find incremental buyers and their true cost, and shows how to report attributed and incremental results honestly.
- AI at workRules, consent and trustRead nowRules, consent and trust
Marketing rules differ by country and change over time, but the principles are widely shared. This lesson covers consent for data and messages, disclosure of AI and paid endorsements, fake reviews and synthetic testimonials, and extra care for children and vulnerable people. Education, not legal advice.
- AI at workThe AI threat landscape at workRead nowThe AI threat landscape at work
AI assistants, connectors and agents add new entry points to every organisation. This lesson sorts AI threats into threats to people, data and systems, and shows how a four-question threat model and published frameworks help teams defend proportionately.
- AI at workDefending against prompt injectionRead nowDefending against prompt injection
Indirect prompt injection hides instructions in documents, emails and web pages that an AI tool reads for someone else. Filters alone cannot stop it, so organisations layer defences: separate data from instructions, limit tools, require approvals and monitor.
- AI at workData leakageRead nowData leakage
Data leaks through AI tools by prompts, uploads, connectors, outputs and logs, usually during ordinary work. Classification labels, approved tools and data loss prevention reduce the risk, and a worked month of incidents shows where to act first.
- AI at workDeepfakes and AI-powered social …Read nowDeepfakes and AI-powered social engineering
AI makes it cheap to imitate a director's voice, a supplier's emails or a whole video call. Spotting fakes is getting harder, so workplaces rely on process: call back on a known number, use code words and require two people for sensitive actions.
- AI at workSupply-chain and model risksRead nowSupply-chain and model risks
Every AI tool depends on providers, models, data sets, plug-ins and services. This lesson explains data poisoning in concept, how to judge add-ons and downloaded models, the questions to ask a provider, and how to test for changes after updates.
- AI at workAccess and identity for AI tools…Read nowAccess and identity for AI tools and agents
Least privilege decides how much damage a misled agent can do. This lesson covers separate identities for agents, safe handling of keys, approval steps people actually read, logging, revoking access and a worked access review.
- AI at workHandling AI outputs safelyRead nowHandling AI outputs safely
AI output can be wrong or shaped by planted content, so it should be treated as untrusted input. This lesson covers links, attachments and formulas, AI-written code and invented software packages, and how review, dry runs and sandboxes keep scripts safe.
- AI at workAI goes wrong: incident responseRead nowWhen AI goes wrong: incident response
AI incidents are often quiet: wrong advice, a misdirected attachment, a leaked figure. This lesson defines an AI incident, walks the cycle of report, contain, investigate, fix and learn, explains the first hour and who to tell, and follows a worked incident.
- AI at workBuilding a safe AI cultureRead nowBuilding a safe AI culture
Rules work only when people follow them under pressure. This lesson shows how short rules, role-based training, easy blame-free reporting, authorised red teaming and a few simple numbers build a safe AI culture, and brings the course together into one plan.
Advanced
- AI at workWhat AI agents areRead nowWhat AI agents are
An AI agent is given a goal rather than a fixed path: it plans, uses tools, observes the results and decides what to do next. This lesson separates agents from assistants and workflows, explains levels of autonomy and shows why agent mistakes need different handling.
- AI at workTools and function callingRead nowTools and function calling
Models produce text; tools let them look up records, calculate exactly and request actions. This lesson explains how a tool call works, why tool descriptions are part of the prompt, which checks a request should pass and how to read a multi-tool trace.
- AI at workRules-based automation and where…Read nowRules-based automation and where AI fits
Most automation still runs on triggers, conditions and actions. This lesson covers rule-based workflows, robotic process automation and no-code tools, explains when exact rules beat a model, and shows hybrid designs where models read and rules decide.
- AI at workDesigning agent workflowsRead nowDesigning agent workflows
Good agent workflows break work into steps with one job each, route items to the right path and hand work over cleanly. This lesson covers chaining, routing, parallel steps, handoffs and orchestration, where to put human approval, and a worked invoice workflow with step counts.
- AI at workMemory, retrieval and knowledgeRead nowMemory, retrieval and knowledge
Agents work from what is placed in front of them for each task, plus anything the system retrieves from stored knowledge. This lesson explains context and long-term memory, embeddings and vector search in plain words, retrieval-augmented generation, stale documents and access rights.
- AI at workGuardrails and permissionsRead nowGuardrails and permissions
Guardrails limit what an AI system may say or do, and for agents the action limits matter most. This lesson covers least privilege, excessive agency, approvals, spending and action limits, sandboxes, allow-lists, defences against misleading content and a working stop switch.
- AI at workTesting and monitoringRead nowTesting and monitoring
You cannot judge an agent by trying it a few times. This lesson shows how to build a test set of tasks, combine step-level and end-to-end checks, keep logs and traces, track error rates with alert thresholds, and re-run tests after every change.
- AI at workCosts and reliabilityRead nowCosts and reliability
Agents charge by use and fail by multiplication. This lesson works through running costs with illustrative unit prices, latency and retries, how small step error rates compound across a task, and designs that fail safely, visibly and recoverably.
- AI at workDeciding what to automateRead nowDeciding what to automate
Almost anything can be automated badly. This lesson gives a scoring sheet for volume, rules clarity, cost of error, checkability and data access, explains process mining, works through a decision on three candidate processes and shows how to pilot, measure, scale and keep people informed.