Choose your profession.
Under each profession you find the courses connected to it: the ones written specifically for it, and the ones that apply across every field.
01
Healthcare
A process from your work
How long does it take you to write up your visit notes?
- Step 1With the patient's consent, the tool listens to the visit and drafts the note for the record.
- Step 2You read the note before saving it and correct any line about something that did not happen in the visit.
- Step 3The diagnosis, the decision and the signature stay yours.
- Where the tool goes wrongThe note adds a lung examination and rules out chest pain, though neither happened.
- The verification stepBefore saving, delete every finding for an examination you did not do today.
What was measured
What changed, as measured
In the randomised trial, time spent on notes fell by 0.36 hours a day, about 21.9 minutes, for those using the AI scribe.
Who was measured: Doctors and other practitioners in outpatient clinics of one US university health system, randomised to use an ambient AI scribe or to work as before.
The caveat: Participants volunteered, so they may be early adopters of new technology; the study was run in a single health system.
The course covers notes, letters and messages drafted with AI, with every fact checked.
We can help you with this process: “Healthcare AI: Clinical documentation”, 8 hours.
Where work with AI goes wrong
In a pharmacy or a clinic, AI invents a drug interaction that does not exist or puts a wrong dose into the patient instructions.
The problem, the workflow and the evidence, in full
- The problem
- The error goes unseen because the text sounds as sure as a reference source, and nobody can tell until it is checked against the prescription.
- The workflow
- The tool proposes; the pharmacist, dentist or clinician compares every claim with the prescription, the examination or the official source, removes patient identifiers before any upload; the diagnosis, the dispensing decision and the signature stay in their hands. Administrative staff take no clinical decisions.
- What you leave with
- A written verification routine, applied before any AI text enters a chart or a patient instruction, and a record noting for each claim the source it was compared with (the prescription, the examination or the official source) and who decided.
02
Law and justice
A process from your work
How long does it take you to draft a legal memo or a judgment?
- Step 1The tool searches the sources and writes the first draft from the file.
- Step 2You check every decision, article and quotation at its official source before using it.
- Step 3You record every check on the verification sheet.
- Where the tool goes wrongThe draft cites a unifying High Court decision that does not exist, with a quoted sentence.
- The verification stepFind the decision at the official source and the quoted sentence word for word in it.
What was measured
What changed, as measured
With a legal research tool or a reasoning model, time fell by roughly 20 to 28% on five of six junior lawyer tasks; drafting a confidentiality agreement was the exception.
Who was measured: Law students in their final years at two US law schools, randomised to work with a legal research tool, a reasoning model or neither.
The caveat: Law students, not practising lawyers or judges; with these tools invented authorities were rare, but they did occur.
In the course you review and draft legal documents with AI, with a protocol for checking citations, acts and articles.
We can help you with this process: “Justice AI: Document review and drafting”, 8 hours.
Where work with AI goes wrong
A legal memo cites a court decision that does not exist, and a power of attorney drafted with AI quietly widens what the representative may do.
The problem, the workflow and the evidence, in full
- The problem
- In legal work a well-formed citation and tidy wording look like finished work, so nobody goes to look the decision up and nobody notices the wider scope.
- The workflow
- Every citation, article and date is checked against the official text before it is used; every clause with legal weight is marked for the lawyer or the notary; the act, the legal judgement and the classification stay with the responsible professional.
- What you leave with
- Every memo, power of attorney or register comes with its verification sheet: for each citation, article and date, the link to the official text and the date of the check, while the clauses left for the lawyer or the notary are noted in a separate column. It is filled in before the document is used.
22 specific courses in this field+ 25 courses that apply across every field
The shared course for police and justice staff is counted in both fields: here, and in Law and order.
03
Public sector
A process from your work
How much of your day goes on finding information and writing emails?
- Step 1The tool finds the relevant documents and drafts the email or the letter.
- Step 2You check every fact at its source before using it, and rewrite the text for its reader.
- Step 3The letter goes out under your name, and the responsible person takes the decision.
- Where the tool goes wrongThe facts, names, dates or references the tool finds may not match their source document.
- The verification stepFor every fact, name, date and reference, open the document it comes from and compare.
What was measured
What changed, as measured
Compared with a group of non-users, staff with the assistant saved an estimated 19 minutes a day across routine tasks; the biggest savings were in searching for information and writing emails, 26 and 25 minutes.
Who was measured: Staff of a UK government department who received an office AI assistant, compared by survey with colleagues who did not.
The caveat: Times are self-reported, and licences went to volunteers first, not at random, so early adopters may overstate the benefit.
The course covers drafting, editing and checking text with AI, with no invented facts and no personal data in public tools.
We can help you with this process: “Public administration AI: Documents and communication”, 8 hours.
Where work with AI goes wrong
A reply to a citizen promises the outcome of the decision, and a customs declaration carries the neighbouring tariff code instead of the right one.
The problem, the workflow and the evidence, in full
- The problem
- In administration the document leaves with the institution's stamp, and the stamp lends weight to every line before anyone asks where it came from.
- The workflow
- The tool prepares; the officer checks every name, number, date and legal reference against the file and the text in force, rewrites it for the reader, and the responsible person takes and signs the decision.
- What you leave with
- The decision trail is kept with the file: the tool's proposal, the officer's changes, the source each name, number, date and legal reference was checked against, and the signatory. When the auditors come, the decision can be reconstructed from it.
15 specific courses in this field+ 25 courses that apply across every field
04
Business and operations
A process from your work
How long does it take you to resolve a customer's chat request?
- Step 1During the chat, the tool suggests replies and points to the relevant knowledge base articles.
- Step 2You compare every promise the suggestion makes with the company's terms before sending it.
- Step 3The conversation stays yours: you can accept, change or ignore the suggestion.
- Where the tool goes wrongThe draft invents a freeze rule the gym does not have: longer, free, no medical report.
- The verification stepFind the matching clause in the current terms for every promise; delete any promise without one.
What was measured
What changed, as measured
Agents with the AI assistant resolved 14% more issues per hour on average, and novice and low-skilled agents 34% more; experienced agents gained little.
Who was measured: Chat support agents of a large business software company and its outsourced contact centres, as the assistant was rolled out in stages.
The caveat: A staged rollout at one company, not a randomised trial; the most skilled agents may see the quality of their conversations fall with the assistant.
The course covers correspondence, tickets and the knowledge base with AI: every answer verified against the official source.
We can help you with this process: “Customer service AI: Correspondence, support tickets and knowledge base”, 8 hours.
Where work with AI goes wrong
A VAT return carries a wrong rate, and a loan offer goes out with last month's interest rate.
The problem, the workflow and the evidence, in full
- The problem
- In both, the figure has the right format and sits in the right place in the document, so nobody reconciles it until the error surfaces with the client or the tax office.
- The workflow
- The tool proposes; the accountant, banker or manager reconciles every figure with the records or the source system, checks every legal or contractual line in its own source, and signs only what they have checked. The decision stays in human hands.
- What you leave with
- Before any signature you run a reconciliation check written step by step, so anyone in the office can repeat it; next to it, the figures record shows the source system of each figure and who approved it.
75 specific courses in this field+ 25 courses that apply across every field
Accounting and Finance
5 specific coursesBanking
11 specific coursesCustomer service
10 specific coursesE-Commerce Operations
5 specific coursesHuman resources professional
5 specific coursesLogistics and Supply Chain
5 specific coursesMarketing and content professional
5 specific coursesProject Management
5 specific coursesRetail Operations
5 specific coursesSales operations professional
5 specific coursesStatutory Audit
5 specific courses
05
Insurance
A documented failure pattern
Clause 4.2 is about floods, not a burst pipe
- What the tool doesA draft claim refusal letter, produced by an AI tool for a claims handler, to be sent to the policyholder after a water leak in a flat, under a fictional home insurance policy.
- Where the tool goes wrongThe draft applies the exclusion for flooding from outside to a burst pipe inside the building.
- The verification stepSplit the exclusion into its elements and check each against the adjuster's report and photos.
Why it is easy to miss, and the full check, are on the Insurance professional page.
What it looks like
Policy conditions (fictional), clause 4.2Damage from flooding is not covered where river water or rainwater enters the home from outside.
Reported lossDamp in the kitchen ceiling from a burst water pipe in the flat on the floor above.
The AI tool's draft letterWe inform you that your claim is not accepted because, under clause 4.2 of the policy, damage caused by water is excluded from cover.
Translated from the Albanian original.
Where work with AI goes wrong
A claim summary deducts the excess twice, a renewal rate is priced on a claims history that is not the one in the core system, a regulatory report cites a circular that has been replaced.
The problem, the workflow and the evidence, in full
- The problem
- In insurance, an amount or a rate worked out by the tool looks like any other line in the file; the mismatch with the core system only shows when the claim is paid or the supervisor reads the report.
- The workflow
- AI drafts; the specialist checks every amount against the core system, the adjuster's report and the policy terms, keeps every regulatory reference marked as unverified until it is found in the official text, and a person always decides on the claim or the risk.
- What you leave with
- The map of drafted texts lists every kind of text AI produces (claim summary, renewal rate, regulatory report) with the note it must carry, the source where it is verified and the role that signs it. It is used whenever such a text comes in, even by someone on their first day in the job.
10 specific courses in this field+ 25 courses that apply across every field
Other fields
These fields have no face of their own on the cube. Each has its example, its professions and its courses below.
06
Law and order
A process from your work
How long does it take you to write up the report after a call?
- Step 1From the body camera audio or the notes, the tool drafts the narrative of the report.
- Step 2You fill in the structured data yourself and read the draft against your notes.
- Step 3Any line your notes do not support is removed; the signature stays with the officer.
- Where the tool goes wrongThe record adds as a participant a colleague who was only on the phone.
- The verification stepWrite from your notes who was physically at the scene and compare name by name.
What was measured
What changed, as measured
In a preregistered randomised trial, AI drafting did not significantly shorten report writing; the authors conclude that no time saving should be expected from it.
Who was measured: Patrol officers of a medium-sized US police department, in a preregistered randomised trial.
The caveat: One police department; the trial measured the time to write, not the quality of the reports; other effects remain untested.
The course covers work documents drafted with AI: marking every part the tool touched, review by a colleague and strict verification.
We can help you with this process: “Policing: Work documents, sources and versions”, 8 hours.
Where work with AI goes wrong
In daily work, an AI-drafted text turns a proposal into a decision already taken or sets a deadline that was never given, and a few days later there are three files named ‘final’.
The problem, the workflow and the evidence, in full
- The problem
- Nobody knows which one was checked, because nothing in the text shows where the tool's proposal ends and the official's word begins.
- The workflow
- AI assists with drafting and documentation only. Proposals are kept apart from decisions, deadlines come only from the source, every version and change is logged, and personal data never goes into a public tool. Decisions, evidence handling and legal judgement stay with the responsible official.
- What you leave with
- Each drafted text gets a control log: which version was checked, what changed and why, which lines are proposals and which are decisions, and which source each deadline comes from. When the file passes to a colleague, the log is where they start.
6 specific courses in this field+ 25 courses that apply across every field
07
Education and specialised fields
A process from your work
How much of your week goes on preparing lessons?
- Step 1From the topic and objectives you give it, the tool proposes a lesson plan and materials.
- Step 2You choose, adapt and check all material before using it with the group.
- Step 3What is taught and how it is assessed stays the teacher's or the trainer's decision.
- Where the tool goes wrongThe facts, definitions or exercise answers the tool suggests may not be right.
- The verification stepCompare every fact and definition with the programme or approved textbook, and work the exercises yourself.
What was measured
What changed, as measured
Teachers using ChatGPT with a guide spent around 56.2 minutes a week on preparation against 81.5 in the comparison group, saving 25.3 minutes a week on average.
Who was measured: Lower secondary science teachers in schools in England, randomised to prepare with ChatGPT and a guide, or as before.
The caveat: School teachers, not vocational trainers; most used the new approach for only part of their timetable.
The course covers a teacher's or trainer's daily workflow with AI: lesson planning, materials and assessment, with a human check at every step.
We can help you with this process: “Teaching and training AI: Workflow”, 8 hours.
Where work with AI goes wrong
A technical report states a seismic zone that does not match the official zoning map, a legal term in a translated contract is not the one in the client's glossary, a training report takes the employment rate of a small group as proof of success.
The problem, the workflow and the evidence, in full
- The problem
- In technical and academic work the text goes to a reader who trusts the expert, and after the expert nobody opens the map, the glossary or the register.
- The workflow
- The tool drafts or proposes; the engineer, translator or programme manager checks every value, term and figure against the project data, the glossary or the register, while the signature and the conclusion remain theirs.
- What you leave with
- In the findings table, every value, term or figure sits next to its own source: the project data, the glossary or the register. The second reviewer opens it before signature and has to reach the same conclusion.
08
Cybersecurity
A process from your work
How long does your team take to resolve a security incident?
- Step 1The tool summarises the alerts, proposes the analysis steps and drafts the incident report.
- Step 2The analyst verifies every conclusion in the system's own data before acting.
- Step 3The notification deadline, recipient and article come from the law in force, not from the tool.
- Where the tool goes wrongThe answer cites a repealed law and does not give the reporting deadline in force.
- The verification stepOpen the law in force on qbz.gov.al and compare the article, deadline and recipient.
What was measured
What changed, as measured
Organisations that adopted a generative AI security assistant saw their mean time to resolve incidents fall by 30.13%.
Who was measured: Security operations centres of organisations using one vendor's security platform, adopters of its AI assistant compared with non-adopters over several months.
The caveat: The tool maker's own study; it shows an association in observational data, not proof of causation.
The Cybersecurity Seminar takes a team through incidents in four one-day episodes; the training prepares the person who runs it.
We can help you with this process: “Running the Cybersecurity Seminar”, 6 hours.
Where work with AI goes wrong
When a cybersecurity exercise day is adapted for an organisation, a legal line gets cited from memory or from an AI text instead of the references file, and after the first episode a manager asks for the quiz results person by person.
The problem, the workflow and the evidence, in full
- The problem
- A line cited from memory sounds familiar, so nobody compares it with the file, and the manager's request looks like an ordinary request to track progress.
- The workflow
- Changes go only through the intake form and the swap table, every legal line is compared with the references file before the day, and the report contains counts only, with no group smaller than ten people shown.
- What you leave with
- The customisation record contains every line changed through the intake form and the swap table, together with each legal reference as compared with the references file. The next instructor reads it before the exercise day.
2 specific courses in this field+ 25 courses that apply across every field
Across every field
AI tools at work
These 25 courses are the same for every field: the foundations of AI, checking its answers, agents, and what to do when you get stuck.
Where work with AI goes wrong
A complaint with a name and an order number goes straight into a free chat assistant.
The problem, the workflow and the evidence, in full
- The problem
- A note to a client contains an article number and a statistic attributed to a real institution, with the source never checked. An agent given the whole shared drive also reads the staff files. None of these steps leaves a mark on the text that comes out: the paste is a single move, the citation looks complete, and nothing shows which files the agent read.
- The workflow
- Before pasting, you classify the text and remove identifiers; after every reply from the tool, you check facts, tone and policy and verify every citation at its source; an agent gets only named folders, and a person approves before anything is sent.
- What you leave with
- Three small tools for every day: the pasting rule, which says what comes out before text goes into an assistant; the checking card for facts, tone, policy and citations after each reply from the tool; and the disclosure line added to the text before it is sent.
The figure on each field above counts only the courses written for that field, each once. These 25 courses are not in it, because they apply to every field and profession.
09
By organisation type
These are not professions. They are courses and programmes written for a type of organisation. You can take them on your own or ask for them as a workshop for your team.
Chambers of commerce
2 specific coursesCivil Society Operations
2 specific coursesCorporate Group Operations
7 specific coursesCybersecurity Seminar instructors
1 specific courseCybersecurity Seminar participants
1 specific courseLocal Public Administration
2 specific coursesRegister Operations
1 specific course
Not finding your profession?
The catalogue lists every course, with filters for profession, duration and format.