Watching isn’t the proof. The work is.

1. How Modern AI Actually WorksExplain what a language model is doing when it answers, and predict where it will be wrong.
2. Prompting Like a ProWrite a prompt that specifies the task, the context and the shape of the answer.
3. Picking the Right Model for the JobChoose a model for a job on cost, latency and capability rather than on reputation.
4. Context Is EverythingAssemble the context a task needs, and recognise when the window is the constraint.
5. Working With Files, Images, and Your Own DataPut your own files and data in front of a model without hand-pasting them.
6. AI for CodingUse AI on real code without accepting output you cannot review.
7. Tool Use and AgentsDescribe what makes something an agent, and what tool use costs you in reliability.
8. Building Simple AI Workflows Without CodeBuild a working automation end to end without writing an application.
9. EvalsWrite an eval that catches a failure a demo would have hidden.
10. Staying Current Without DrowningKeep up with the field on a budget of an hour a week, without chasing noise.

Final exam: 20 questions, 60 minutes, closed book, 70% to pass. 30 module-quiz questions along the way.

1. The AI-Native ProfessionalSay what AI-native actually means — delegate, verify, systematize — and tell it apart from using a chatbot occasionally.
2. How to Talk to AIWrite a request with all four parts — role, context, task, format — and fix an answer that came back wrong by naming which part was missing.
3. Context Is EverythingAttach before you ask: the document, one example of good, the raw data and the constraints — and say why one example beats a paragraph describing it.
4. The Everyday ToolkitUse the five verbs — draft, summarize, analyze, brainstorm, build — with the move that separates an okay result from a good one in each.
5. Trust but VerifyTriage what needs checking, demand sources, and run the thirty-second check — because fluency is not accuracy and a wrong answer never looks wrong.
6. Your AI WorkspaceBuild a project with the three layers — standing instructions, knowledge files, the chats themselves — so you stop retyping the same setup every week.
7. Workflows, Not PromptsSpot which of your tasks qualify, and turn one into a five-field recipe — trigger, inputs, template, quality check, destination — that runs the same way twice.
8. Capstone: Your Three WorkflowsBuild three workflows in your real job, run each at least twice, and log the honest before-and-after minutes against a baseline.

Final exam: 15 questions, 50 minutes, closed book, 70% to pass. 32 module-quiz questions along the way.

1. Where Your Funnel LeaksLocate the leak as a shared-context problem rather than a team problem — the same question answered four ways because nobody holds the same version of the company.
2. The Shared GTM BrainWrite the five blocks of standing context — who you sell to and who you don't, positioning, objections and the rest — and name the owner who keeps it alive past week three.
3. Content Only Your Company Could WriteTurn a campaign goal into a brief and then a draft by assembling it from the context blocks, so the output says something a competitor could not have said.
4. The Account BriefRun a six-section account brief in minutes — including the disqualifier — and hold the rule that keeps an invented fact away from a buyer.
5. Before and After the MeetingWalk in with the one-pager that beats the CRM, and clear the three jobs that land when a call ends in a single pass instead of by Friday.
6. Retention Is RevenueStand up a help-docs assistant that never answers without a citation, and read the four signals that flag a renewal while there is still time to act.

Final exam: 15 questions, 45 minutes, closed book, 70% to pass. 24 module-quiz questions along the way.

1. LLM and generative AI fundamentalsDescribe tokens, context, sampling and cost well enough to explain a bill and a latency spike.
2. Python for AIWrite and debug the Python an AI application actually needs — async calls, typing, error handling, environment.
3. Working with model APIsCall OpenAI, Anthropic and Gemini APIs directly, handle their failure modes, and swap between them behind one interface.
4. Structured outputs and tool callingGet validated JSON out of a model reliably, and let it call your functions without letting it call them wrongly.
5. Embeddings and vector databasesChoose an embedding model, chunk a corpus sensibly, and explain why your chunking decides your retrieval quality.
6. Retrieval-augmented generationBuild a RAG pipeline that cites sources and declines to answer when retrieval comes back empty.
7. AgentsBuild an agent with a loop, tools and a stopping condition, and say what it does when a tool fails.
8. MCP and external toolsConnect a model to real systems over MCP, and scope what it is permitted to reach.
9. Evaluation and guardrailsWrite an eval set that catches a regression a demo would hide, and wire it into the path to production.
10. Deployment, monitoring and costShip it, watch it, and know your cost per request before finance asks.

Final exam: 15 questions, 50 minutes, closed book, 70% to pass.

1. AI product fundamentalsSay what changes about product management when the feature is non-deterministic.
2. Identifying AI opportunitiesFind the places in a product where a model beats a rule, and the places it doesn't.
3. User discovery for AI featuresRun discovery that surfaces what users would tolerate being wrong, not just what they'd like.
4. Writing AI product requirementsWrite a requirement that specifies behaviour, failure modes and an acceptance threshold.
5. Choosing modelsPick a model on capability, cost, latency and data terms, and know what would make you switch.
6. PrototypingBuild a working prototype yourself, far enough to learn something engineering couldn't tell you.
7. AI UXDesign for uncertainty — confidence, citation, correction and undo — instead of pretending the model is right.
8. Evaluating model qualityDefine what good looks like as a measurable thing, and run it before and after a change.
9. Cost, latency and reliability tradeoffsMake the call between a better answer and an affordable one, with the numbers in front of you.
10. Responsible AI in productIdentify the harm your feature could do at scale, and design the control that stops it.
11. Launching and iteratingShip an AI feature behind the right guardrails, and use production signal to decide what happens next.

Final exam: 15 questions, 50 minutes, closed book, 70% to pass.

1. The FDE roleSay what an FDE does that a software engineer and a solutions architect do not.
2. Customer discoveryRun stakeholder interviews that turn a vague complaint into a scoped, buildable requirement.
3. Rapid AI prototypingGet a working prototype in front of a customer inside a week, on their data.
4. Enterprise integrationsIntegrate against systems you didn't choose, with authentication you didn't design.
5. Enterprise dataWork with real customer data — messy, permissioned, unstructured — without breaking their rules.
6. ProductionisationTake a prototype to something the customer's own team can operate.
7. Client deliveryScope a pilot, run it, and produce an implementation roadmap they'll fund.
8. Value realisationMeasure the time or money saved and show it, so the pilot becomes a contract.
9. Executive communicationDemo to an executive, handle the objection you didn't prepare for, and close the next step.
10. Deployment simulationRun the whole motion end to end against a scenario you've never seen.

Final exam: 15 questions, 50 minutes, closed book, 70% to pass.

1. What AI is actually doingExplain in one paragraph what a language model does when it answers, and predict two kinds of task it will be bad at.
2. ChatGPT, Claude and GeminiChoose between the three for a given task on grounds other than which one you have open.
3. Prompting that survives contactWrite a prompt that states the task, the audience, the constraints and the output shape, and fix it when the first answer is wrong.
4. Context engineeringAssemble the context a task needs — files, examples, prior decisions — and recognise when the window is the constraint.
5. Deep researchRun a research task that produces sourced claims, and check the sources rather than the summary.
6. Working with documentsGet a real answer out of a long PDF, contract or report without reading all of it or trusting a summary blind.
7. Working with images and slidesUse image input and generation for genuine work — reading a chart, drafting a diagram, fixing a deck.
8. Data analysis without being an analystAsk a spreadsheet a question, get an answer, and verify it against the raw numbers.
9. AI productivity that lasts a monthRebuild one recurring task so that the saving is still there in four weeks, not just on the day you set it up.
10. Automating your first workflowChain two tools and a model so a task you currently start by hand starts itself.
11. Responsible use at workName what you may not put into a model at your own employer, and what you must check before sending its output onwards.

Final exam: 15 questions, 50 minutes, closed book, 70% to pass. 44 module-quiz questions along the way.

1. Workflow thinkingDecompose a real process into triggers, steps, decisions and handoffs before touching a tool.
2. n8n and MakeBuild, debug and version a multi-step automation in a visual builder without it becoming unreadable.
3. APIs and webhooksRead an API doc, authenticate, and receive a webhook — the three things every automation eventually needs.
4. Structured AI outputs in a pipelineMake a model return data the next step can consume, and validate it before the next step runs.
5. AI agents in automationDecide when a workflow needs an agent rather than a branch, and pay the reliability cost knowingly.
6. Multi-step workflowsBuild something with state, retries and branching that survives a partial failure halfway through.
7. Email, calendar and CRM integrationsAutomate against the systems a business actually runs on, including their rate limits and their quirks.
8. Human in the loopPut an approval step where a wrong answer would be expensive, and make it fast enough that people use it.
9. Reliability and error handlingMake an automation fail loudly and recoverably instead of silently and permanently.
10. Production automation designHand an automation to somebody else — documented, monitored, and safe for them to change.

Final exam: 15 questions, 50 minutes, closed book, 70% to pass.

1. Excel and Sheets with AIGet a model to build, explain and fix real spreadsheet logic without breaking the workbook.
2. SQL with AI assistanceWrite and verify SQL against a schema you didn't design, and catch the query that silently returns the wrong rows.
3. Python for analysisLoad, reshape and summarise a dataset in Python with AI writing most of it and you reading all of it.
4. Cleaning dataTurn a dirty export into something analysable, and record every decision that changed a number.
5. Exploratory analysisInterrogate a new dataset methodically instead of hunting for the chart you already wanted.
6. Data visualisationChoose a chart that makes the finding obvious, and reject the one that makes it look bigger.
7. Statistical reasoningTell a real effect from noise, and say out loud how confident you actually are.
8. ForecastingProduce a forecast with a stated method and a stated error, not a line extended to the right.
9. Building dashboardsBuild something a stakeholder can read alone, that doesn't quietly break when the data changes shape.
10. Automating recurring analysisTurn a monthly manual report into a pipeline that runs itself and flags when it looks wrong.
11. Communicating insightDeliver the finding, the confidence and the recommended decision in the first thirty seconds.

Final exam: 15 questions, 50 minutes, closed book, 70% to pass.

1. Customer research with AISynthesise calls, reviews and support tickets into what customers actually say, in their words.
2. PositioningWrite positioning that survives a competitor comparison, backed by evidence rather than adjectives.
3. Content strategyBuild a content plan tied to a funnel stage and a measurable outcome, not to a posting cadence.
4. Copywriting with AIProduce copy in a defined voice that a customer wouldn't clock as machine-written.
5. Image generationGenerate usable brand-consistent imagery, and know what you can and can't legally ship.
6. Video creationTake a script to a finished short-form video with AI doing the production work.
7. SEO in an AI search worldOptimise for both classic search and AI answers, and measure the difference.
8. Paid adsGenerate, test and kill ad creative at volume without losing the ability to say what won.
9. Social contentTurn one asset into a week of channel-appropriate posts without producing seven versions of nothing.
10. RepurposingBuild the pipeline that turns every long-form piece into its derivatives automatically.
11. Marketing automationWire research, production, scheduling and reporting into one workflow that runs weekly.
12. Analytics and attributionAttribute a result to the thing you did, and be honest about the baseline you compared against.

Final exam: 20 questions, 60 minutes, closed book, 70% to pass.

1. Finding AI opportunitiesScan a business and produce a longlist of candidate use cases with a reason attached to each.
2. Process mappingMap a real process to the level of detail where the automatable steps become obvious.
3. AI maturity assessmentAssess an organisation's data, skills, tooling and governance, and say what the binding constraint is.
4. Use-case prioritisationRank a longlist on value, feasibility and risk, and defend the ordering to someone who disagrees.
5. ROI calculationBuild an ROI model whose assumptions are visible and whose downside case is honest.
6. Building an AI roadmapSequence twelve months of work so that early wins fund and de-risk the later ones.
7. Workforce transformationSay what happens to the people whose work changes, before somebody else asks.
8. AI operating modelsChoose between centralised, federated and embedded AI capability, and staff it.
9. Vendor evaluationEvaluate an AI vendor on data terms, lock-in and evidence rather than on demo quality.
10. Change managementGet a team to actually use a tool that works, which is the part that usually fails.
11. GovernancePut the minimum viable governance in place — policy, review, logging — without stalling delivery.
12. Measuring adoptionInstrument adoption and value so the programme can be defended or stopped on evidence.

Final exam: 20 questions, 60 minutes, closed book, 70% to pass.

1. From business problem to AI systemRestate a business problem as a system with inputs, outputs, constraints and a definition of correct.
2. Build versus buyMake and defend a build-buy call with a cost model and an exit plan.
3. Model selection at system levelChoose models per component rather than per company, and design for replacing them.
4. System architectureDraw the architecture — services, queues, stores, boundaries — so an engineer could build from it.
5. RAG architectureDesign retrieval for a real corpus: ingestion, chunking, indexing, permissions and freshness.
6. Agent architectureDecide where agency belongs, bound it, and design the failure path when it goes wrong.
7. Data pipelinesDesign the pipeline that keeps the system's knowledge current without a person running it.
8. Security by designPut trust boundaries, permissions and secrets handling into the diagram, not into a later ticket.
9. ReliabilityDesign for the model being down, slow, or wrong, and for all three at once.
10. Evaluation architectureMake evaluation a component of the system rather than a spreadsheet someone maintains.
11. ScalingTake a working design to a hundred times the traffic and say what breaks first.
12. Cost architectureModel cost per request across the whole system and design the lever that reduces it.

Final exam: 20 questions, 60 minutes, closed book, 70% to pass.

1. AI governance fundamentalsDescribe what governance is for, and the difference between a policy and a control.
2. Hallucination and reliabilityExplain why models fabricate, and design the check that catches it in a specific workflow.
3. BiasFind bias in a system's outputs with a method you could show an auditor.
4. Data privacySay what may be sent to which model under which agreement, and enforce it technically.
5. Model riskAssess a model's risk in the context of the decision it influences, not in the abstract.
6. Human oversightPlace a human where their review changes the outcome, rather than where it is easiest to add.
7. Evaluation for assuranceBuild evaluation evidence that answers an auditor's question rather than an engineer's.
8. DocumentationProduce the model and system documentation a regulator or customer will actually ask for.
9. Writing AI policiesWrite a policy a real employee can follow, and name the three things it stops them doing.
10. Enterprise controlsTranslate a policy into controls — access, logging, retention, review — that operate on their own.
11. Regulatory awarenessTrack what applies to a given deployment, and what to do when it changes.
12. Responsible deploymentTake a system live with its risks documented, its controls tested and its off-switch defined.

Final exam: 20 questions, 60 minutes, closed book, 70% to pass.

1. AI security fundamentals and threat modellingProduce a threat model for an AI system that names assets, actors and paths.
2. Prompt injection and jailbreaksExecute direct and indirect injection against a system you control, and explain why filtering alone fails.
3. Data exfiltration and leakageGet sensitive data out of a naive system, then show the control that stops it.
4. RAG securityPoison a retrieval corpus, and design ingestion and permissioning that makes it not matter.
5. Agent permissions and excessive agencyScope an agent's tools and credentials to the least it needs, and prove the boundary holds.
6. Secure APIs, credentials and secretsHandle keys and tokens in an AI system so a leaked prompt doesn't leak an account.
7. Authentication and authorisationCarry a user's identity and permissions through retrieval and tool calls without widening them.
8. Adversarial testing and red teamingRun a structured red-team exercise and write up findings a team can fix.
9. Monitoring, logging and incident responseDetect abuse in production, and have a plan for the day it happens.
10. Secure AI deployment architectureDesign a deployment where the blast radius of a successful attack is small and known.

Final exam: 15 questions, 50 minutes, closed book, 70% to pass.

1. AI infrastructure fundamentalsDescribe what actually runs when a request hits an AI product, and where the time goes.
2. Cloud architectureStand up the cloud components an AI service needs, and justify each one.
3. Containers and DockerContainerise an AI application so it runs identically on your machine and in production.
4. Model serving and inference APIsServe a model behind an API with batching, timeouts and backpressure.
5. GPU fundamentalsSay when you need a GPU, which one, and what it costs you per hour and per request.
6. CI/CD for AI systemsBuild a pipeline where a prompt change goes through the same gates as a code change.
7. Evaluation pipelinesGate deploys on an eval suite that fails the build when quality regresses.
8. Observability and monitoringInstrument traces, quality and cost so a production problem is diagnosable.
9. Latency, caching and performanceCut p95 latency with caching and routing without changing the answer.
10. Cost optimisationHalve inference cost with routing, caching and model choice, and prove the quality held.
11. Versioning and rollbacksVersion models, prompts and data together, and roll back all three in one action.
12. Scaling production AITake a service to a hundred thousand users and know the number that breaks first.

Final exam: 20 questions, 60 minutes, closed book, 70% to pass.

1. The AI landscape for consultantsHold an accurate, current map of what AI can and can't do, well enough to be challenged on it.
2. Client discovery and stakeholder interviewsRun interviews that get you the truth about a process rather than the org chart's version.
3. Process mappingMap a client process to the detail where the cost sits, in a single session.
4. AI maturity assessmentAssess a client and produce a defensible current-state read.
5. Identifying use casesGenerate a longlist that includes the boring high-value ones nobody demos.
6. Prioritising use casesScore and sequence use cases with a method the client can re-run without you.
7. Build versus buy analysisAdvise on build, buy or wait, with the cost and lock-in consequences of each.
8. AI business cases and ROIBuild a business case that survives a CFO reading the assumptions tab.
9. Prototype creationBuild the prototype that makes the recommendation concrete, yourself, in days.
10. Implementation roadmapsProduce a roadmap with owners, sequencing, dependencies and decision points.
11. Change managementPlan for adoption from day one, because the technical work is not the risk.
12. Executive communicationPresent to a board: the recommendation first, the evidence behind it, and what you need from them.

Final exam: 20 questions, 60 minutes, closed book, 70% to pass.

1. AI fundamentals for financeSay where a model can and cannot be trusted in work that has to reconcile.
2. AI and ExcelBuild, audit and fix financial models with AI without losing the audit trail.
3. Financial statement analysisExtract and check the numbers that matter from a set of statements.
4. Company and industry researchProduce sourced research on a company and its market, with every claim traceable.
5. Earnings calls and document intelligenceTurn transcripts and filings into a brief that flags what changed rather than what was said.
6. Financial modelling with AIBuild a model faster with AI while keeping every formula something you can defend.
7. Python and SQL for financeQuery and reshape financial data directly instead of exporting to a spreadsheet.
8. Forecasting and scenariosProduce scenarios with explicit drivers, and stress the ones that actually move the answer.
9. Investment researchWrite an investment memo with a thesis, the disconfirming evidence, and what would change your mind.
10. Reporting and presentationTurn analysis into a board-ready pack without the numbers drifting between slides.
11. Finance workflow automationAutomate a recurring finance process with the controls the process requires.
12. Risk, privacy and model controlsApply the controls a regulated finance function needs before AI touches its data.

Final exam: 20 questions, 60 minutes, closed book, 70% to pass.

1. Generative AI for legal professionalsState plainly what AI can do in legal work and where professional review is non-negotiable.
2. Legal research workflowsRun AI-assisted research where every authority is verified before it is relied on.
3. Case and document summarisationSummarise a long document so that the omissions are deliberate and stated.
4. Contract analysisExtract obligations, rights, dates and risks from a contract, each traced to its clause.
5. Contract drafting assistanceDraft with AI from a precedent bank without importing a term you didn't intend.
6. Due diligenceRun a document-heavy diligence exercise with AI and a defensible sampling method.
7. Discovery and document reviewUse AI to prioritise review, and measure its recall before trusting it to deprioritise.
8. Comparing clauses and obligationsCompare agreements at scale and produce a reliable differences report.
9. Legal knowledge systemsBuild a retrieval system over a firm's own materials that cites its sources.
10. Legal workflow automationAutomate intake, triage or reporting without automating a judgement call.
11. Confidentiality and data handlingDecide what may be sent to which system under privilege and client obligations.
12. Verification and professional responsibilityBuild the verification step that makes AI-assisted work safe to sign.

Final exam: 20 questions, 60 minutes, closed book, 70% to pass.

1. The AI landscape in healthcareMap where AI is being used across care, operations and research, and where it isn't allowed.
2. Healthcare data fundamentalsWork with healthcare data structures and their constraints without mishandling them.
3. Clinical documentation workflowsDesign an AI-assisted documentation workflow with clinician review built in.
4. Medical research and literature analysisSynthesise literature with sources verified and evidence quality stated.
5. Patient communication workflowsDraft patient-facing communication with AI, with review and reading-level controls.
6. Healthcare operationsFind and size the operational bottlenecks AI can actually move.
7. Scheduling and administrative automationAutomate an administrative process end to end within its compliance boundary.
8. Knowledge retrieval systemsBuild a retrieval assistant over an organisation's own protocols and policies.
9. Evaluating healthcare AIEvaluate a vendor's clinical or operational AI claim against its evidence.
10. Privacy and sensitive dataApply the handling rules that patient data requires, technically and procedurally.
11. Human oversight and safetyPlace oversight where an error would reach a patient, and make it workable in practice.
12. Deploying AI responsibly in healthcareTake an implementation live with governance, training, monitoring and a rollback path.

Final exam: 20 questions, 60 minutes, closed book, 70% to pass.

1. AI for modern HRSay where AI helps in the employee lifecycle and where using it creates legal exposure.
2. Workforce planningModel skills, gaps and headcount scenarios with AI doing the analysis.
3. Job descriptions and competency frameworksProduce role definitions and competency frameworks grounded in the actual work.
4. Recruiting research and sourcingSource candidates with AI research, and audit the shortlist for who it systematically missed.
5. Candidate communicationsAutomate candidate communication that stays personal and legally clean.
6. Interview preparationBuild structured interview guides and scoring rubrics tied to the competencies.
7. Interview intelligenceUse recording and analysis to improve decisions, with consent and bias controls.
8. Employee onboardingBuild an onboarding assistant that answers from real internal policy.
9. Learning and developmentDesign a skills-based L&D programme and measure whether capability moved.
10. Internal knowledge assistantsDeploy an internal assistant with correct permissions on sensitive HR data.
11. HR analyticsAnswer a people question with data and state the confidence honestly.
12. Bias, privacy and human oversightEvaluate your own AI hiring system for adverse impact, and document the review.

Final exam: 20 questions, 60 minutes, closed book, 70% to pass.

1. AI customer experience strategyDecide what should be automated, what should be assisted, and what must stay human.
2. Support workflow mappingMap contact drivers and volumes to find where automation pays.
3. Knowledge base designRestructure a knowledge base so retrieval can actually answer from it.
4. RAG for customer supportBuild support retrieval that answers from approved content and cites it.
5. AI chat agentsDeploy a chat agent with scoped capabilities and a tested refusal behaviour.
6. Voice AI agentsBuild a voice agent that handles interruption, ambiguity and handoff.
7. Ticket classification and routingClassify and route accurately enough to change queue times, and measure it.
8. Agent-assist systemsGive human agents AI assistance that speeds them up without degrading quality.
9. PersonalisationPersonalise from customer context without crossing a privacy line.
10. Escalation and human handoffDesign handoff that carries context and triggers before the customer gives up.
11. Quality assurance and evaluationEvaluate AI-handled conversations at scale on correctness and tone.
12. Support analytics and ROIReport deflection, resolution and satisfaction against a defensible baseline.

Final exam: 20 questions, 60 minutes, closed book, 70% to pass.

1. AI-enabled business analysisSay what changes about the BA role when AI can do the analysis and the drafting.
2. Requirements gatheringGather requirements that state outcomes and constraints rather than a requested feature.
3. Stakeholder interviewsInterview stakeholders who disagree, and reconcile what you heard.
4. Business process mappingProduce an as-is map accurate enough that the people in the process recognise it.
5. Data analysis with AIInterrogate business data to test whether the stated problem is the real one.
6. Excel and SheetsDo serious spreadsheet analysis with AI assistance and keep it auditable.
7. SQL with AI assistanceGet the data yourself instead of queuing for it, and verify what comes back.
8. AI-assisted researchResearch a market, vendor or regulation with sourced, checkable output.
9. Requirements and user storiesWrite stories and acceptance criteria a builder can work from without asking you a question.
10. Process redesignDesign the to-be process, including what happens to the steps you removed.
11. Automation opportunity identificationIdentify and size automation candidates, and reject the ones that only look good.
12. Business cases and ROIBuild a business case with visible assumptions and an honest downside.
13. Dashboards and communicationBuild the dashboard and the narrative that let a decision get made.
14. From analysis to implementationStay useful through delivery — clarifying, testing and confirming the benefit landed.

Final exam: 20 questions, 60 minutes, closed book, 70% to pass.