Mickai / Trust Agent
Choose a published course around the decisions, information and responsibilities in your work. Learn in the browser, use the workbook and practise with invented examples before applying the method to a real situation.
4 published courses
Start with a measurable business problem and a non-AI baseline. Define the information boundary, the person with decision authority, the time available for review and the evidence required before a pilot can expand.
What you will practise: Build a readiness decision, map an information-security scope, or test when an agent must stop for fresh human approval. Bring an invented workflow to the exercises, not customer records or confidential business files.
These courses teach decision methods and control design. Completing them does not certify an organisation or demonstrate that a deployed system is safe.
Recommended for you
A practical method for leaders investing in AI: define the problem, compare the alternatives, build an honest business case, check readiness, then pilot it inside real gates.
Recommended for you
Design AI oversight that acts before impact: distinguish presence from approval, bind decisions to actions, test safe stops and build a proportionate evidence dossier.
Recommended for you
Build a working ISMS for an AI-enabled organisation: scope, assets, risk, controls, suppliers, incidents and audit, ending in a scoped implementation plan.
Recommended for you
Take an AI agent from demo to production: identity, least-privilege tools, network and data boundaries, budgets, protected logging, release gates and a rehearsed incident response.
1 published course
Yes. A summary, scheduling message or information handoff can alter what people understand or do. Assess the actual purpose, affected people, possible consequences and human escalation route before choosing controls.
What you will practise: Work through a fictional appointment-summary workflow. Develop an intended-purpose statement, hazard log, data boundary, release decision and response to change. The synthetic exercise uses no patient information.
This is organisational governance education, not diagnosis, treatment or a medical-device classification service. Select the relevant reference pathway and obtain appropriate clinical, legal and regulatory review for a real deployment.
Recommended for you
Map healthcare AI purpose, indirect harm, privacy and human review. Build a tested appointment-workflow assurance dossier with clearly scoped regulatory references.
1 published course
AI can help organise approved intake information, but an uncertain identity match, confidentiality concern or possible conflict needs an authorised review. A draft acknowledgement must not imply that an engagement has been accepted.
What you will practise: Use fictional enquiries to distinguish a person from a similarly named organisation, preserve unresolved matches and design a clear handoff. Test the limits of a synthetic intake gate and document who may decide.
The course does not clear conflicts, establish a lawyer-client relationship or give jurisdiction-specific legal advice. Its selected professional references have explicit scope limits.
Recommended for you
Design a bounded legal-intake workflow that preserves uncertainty, protects confidential information and leaves conflict and engagement decisions with qualified people.
1 published course
Preserve the received source, work on an authorised copy and record every transformation. Label a transcript, enhancement or summary as a derivative, retain its relationship to the source and make uncertainty available to the reviewer.
What you will practise: Trace synthetic exhibits through custody and transformation records. Check what byte-level integrity establishes, what it cannot establish, and what information a reviewer needs to challenge or reproduce a result.
This is evidence-handling and governance education using invented records. It does not teach operational surveillance or establish lawful acquisition, authenticity or admissibility. Applicable rules require the relevant authorised professional review.
Recommended for you
Preserve source material, trace AI derivatives, challenge integrity claims and prepare a jurisdiction-scoped evidence handover with accountable human decisions.
1 published course
Begin with a defined task, approved information and a named reviewer. A draft of public reception information is different from a decision about housing, benefits or a person's rights. Set a clear stop and escalation route before a trial.
What you will practise: Plan a four-week induction for a fictional council enquiries team. Compare AI assistance with a simpler method, check a public-information draft against its source and practise refusing tasks outside the agreed purpose.
The course teaches public-service task selection and staff learning. It does not authorise automated eligibility or adverse decisions. Its selected jurisdiction references are starting points for appropriate local review.
Recommended for you
Choose suitable public-service AI tasks, protect information, check drafts and build a practical staff induction with human authority and accessible alternatives.
1 published course
Keep the purpose narrow, use approved public or synthetic material and retain an authorised human decision. A plausible draft, a public policy document or access to a tool does not grant permission to process information or act for an organisation.
What you will practise: Use invented staff-learning and administrative situations to check purpose, information boundaries, permissions, expiry and escalation. Record why a request should proceed to review, pause for clarification or stop.
This is routine noncombat workplace administration education. It excludes operational planning, intelligence, targeting, weapons, mission logistics, military equipment, readiness and military technical capability. It is not departmental approval or professional accreditation.
Recommended for you
Learn to assess routine administrative AI use, protect information, verify policy drafts and preserve accountable human authority through synthetic staff-learning cases.
1 published course
Identify the learning or administrative purpose, check the information being shared, understand the tool limitations and assign human responsibility for the result. Keep safeguarding escalation and school approval procedures visible.
What you will practise: Use fictional school situations to evaluate a proposed AI use, verify generated material and record a reasoned approve, hold or stop decision. Start with staff understanding before introducing a tool into work involving pupils.
This course is for adults working in schools. It does not provide an unsupervised pupil tutor, a complete curriculum or a substitute for local safeguarding and professional requirements.
Recommended for you
Choose appropriate school AI tasks, verify resources, protect information and test a synthetic teacher-approval gate with clear safeguarding and local policy boundaries.
New to AI? Begin with What is AI?, then return to your field. Each course lists its audience, prerequisites, expected study time and learning outcomes. Use the complete course catalogue for technical and everyday topics.
A course with a scored exam can issue a named, dated Mickai educational certificate for a score strictly above 80 percent. It records the assessment result. Written exercises are ungraded unless the course states otherwise. It is not accreditation, a professional licence or identity assurance.
Teams can work through the public browser lessons and downloaded workbooks. Check the course scope, use fictional practice data and ask the appropriate responsible person to review any proposal for your organisation. This page lists available learning material, not a promise of complete coverage of every profession.
If you need an organisation-specific implementation, explore Mickai's bespoke AI work. Start by agreeing the data boundary, deployment location, decision rights and acceptance evidence. A course exercise does not establish that a production system meets those requirements.
