AI Risk and Impact Assessment in Practice

A practitioner course that takes you through one AI impact assessment from the first screening question to an approved, published and monitored result, under the federal Directive, Quebec law, the EU AI Act and the GDPR, with one fictional company and a template for every step.

59 USDPractitionerENFR
1h 43m of learning
Read the free first lesson

You can ask for a full refund of a course within 14 days of buying it, if you have not passed its final exam.

Educational content, not legal advice.

What you will learn

  • When an AI system needs an impact assessment
  • Describing the system: purpose, decision, data, people
  • Scoring the impact: the four levels
  • Measures that match the level
  • The people affected: notice, explanation, recourse
  • Approve, publish, keep it current

Curriculum

  1. 1When an AI system needs an impact assessmentFree lesson
  2. 2Describing the system: purpose, decision, data, peopleIncluded with purchase
  3. 3Scoring the impact: the four levelsIncluded with purchase
  4. 4Measures that match the levelIncluded with purchase
  5. 5The people affected: notice, explanation, recourseIncluded with purchase
  6. 6Approve, publish, keep it currentIncluded with purchase

About this course

This is a practitioner course for the person who has to assess one AI system and get that assessment approved. It starts with the first screening question, which rules ask for an assessment of this use and who carries the duty, and ends with an assessment that is approved, published where the rules call for it, and kept current.

The whole course follows one fictional company, Quillmere Decision Systems, and its claims triage model, which a federal department, a Quebec insurer and a European life insurer each want to use. Each lesson takes one step of the assessment under the federal Directive on Automated Decision-Making, Quebec law, the EU AI Act and the GDPR, and gives you a template for that step, so you finish with a kit you can adapt for your own system.

Who this course is for

The course is for the people who run or feed an AI impact assessment:

  • AI governance and privacy leads;
  • product owners whose system makes or supports decisions about people;
  • vendors whose system will be used by a federal department, a Quebec enterprise or a European deployer;
  • managers and counsel who review assessments before they are signed.

It assumes you know what an AI governance program is. If you do not, start with the free AI Governance foundation course, which teaches the program; this course runs one assessment end to end.

What you will be able to do

By the end of the course you will be able to:

  • Tell which rules ask for an impact assessment of a given use of an AI system, and who carries each duty.
  • Write the system profile an impact assessment starts from.
  • Reach an impact level with the Algorithmic Impact Assessment and explain it.
  • Choose the quality assurance, peer review and human involvement each impact level calls for.
  • Write the notice and the explanation each regime asks for, and set up a real route to a person who can review the decision.
  • Get an assessment approved and published before production where the Directive applies, and keep it current.

What the law and the frameworks say

The federal Directive on Automated Decision-Making applies to institutions subject to the Policy on Service and Digital, and to any automated decision system in production used to make an administrative decision, or a related assessment, about a client. TBS Directive on Automated Decision-Making ss.5.1, 5.2, 8.1, 8.2; Appendix A (administrative decision); Policy on Service and Digital s.6.1 The duty sits with the department that uses the system. TBS Directive on Automated Decision-Making ss.5.1, 5.2, 8.1, 8.2; Appendix A (administrative decision); Policy on Service and Digital s.6.1 The vendor supplies much of what the assessment needs.

The Algorithmic Impact Assessment places the system at an impact level from Level I to Level IV, and the level sets the requirements of Appendix C of the Directive. TBS, Algorithmic Impact Assessment tool, ss.1, 2, 2.1, 2.2 (Tables 1.1 to 5); TBS Directive on Automated Decision-Making Appendix B The final results are approved and published on the Open Government Portal before the system is in production, then reviewed and updated on a scheduled basis and when the system's functionality or scope changes. TBS Directive on Automated Decision-Making s.6 (opening words), ss.6.1.1 to 6.1.3

In Quebec, for a decision based exclusively on automated processing of personal information, the enterprise informs the person, on request explains the information used and the principal factors and parameters, and gives the person an opportunity to submit observations to a staff member who can review the decision. P-39.1 s.12.1

In the European Union, some deployers of high risk AI systems perform a fundamental rights impact assessment before first use, among them deployers of systems used for life and health insurance risk assessment and pricing, and notify the market surveillance authority of the results. AI Act Art. 27(1) to (4); Art. 27(4) as replaced by Regulation (EU) 2026/1744, Art. 1(13)(a) These high risk duties apply from December 2, 2027 for Annex III systems, and to a system already placed on the market or put into service only after a significant change in its design, or by August 2, 2030 for systems intended to be used by public authorities. Regulation (EU) 2024/1689 Art. 2(2), Art. 111(4), Art. 113, as amended by Regulation (EU) 2026/1744 AI Act Art. 111(1), (2) and (3), Art. 111(2) as replaced by Regulation (EU) 2026/1744; Regulation (EU) 2026/1744, recital 39 The NIST AI Risk Management Framework is intended for voluntary use, and the course uses it as a method, not as a duty. NIST AI RMF 1.0 (NIST AI 100-1)

How to describe the system, check the answers that weigh most and keep a monitoring log are left to the organization. The course teaches one sound method for each and says plainly where it is practice, not law.

Lesson by lesson

Each lesson combines reading, a short video, a worked example from Quillmere, an exercise with a model answer, a short knowledge check and a template you can adapt. Here is what each one covers.

When an AI system needs an impact assessment

In this lesson you learn to:

  • Tell which rules ask for an impact assessment of a given use of an AI system.
  • Explain who carries each duty: the department, the Quebec enterprise or the EU deployer.
  • Use the NIST AI Risk Management Framework as a voluntary method under those duties.

Template: AI impact screening table. One row per use of the system: the rule that applies, who carries the duty, which assessment it calls for, and what the vendor supplies, with the exclusions to check first.

Describing the system: purpose, decision, data, people

In this lesson you learn to:

  • Write the system profile an impact assessment starts from.
  • Name the decision, say whether the system assists or decides, and list who it affects.
  • Describe the training, input and output data, and the model's known limits.

Template: AI system profile. Purpose, the decision and whether the system assists or decides, the people and groups affected, training, input and output data, the model and its known limits, and the owners, in one document kept current.

Scoring the impact: the four levels

In this lesson you learn to:

  • Reach an impact level with the Algorithmic Impact Assessment and explain it.
  • Check the answers that weigh most, including the mitigation deduction.
  • Compare the federal levels with the EU risk management and fundamental rights tests, and Quebec's proportionate approach.

Template: Impact level worksheet. The raw impact score, the mitigation score and the deduction check, the resulting level, the evidence behind the answers that weigh most, and a side by side with the EU and Quebec approaches.

Measures that match the level

In this lesson you learn to:

  • Choose the quality assurance, peer review and human involvement each impact level calls for.
  • Plan bias testing, monitoring and approval to operate for a Level III system.
  • Write down why each measure is enough, for the federal, EU and Quebec reviews.

Template: Measures plan by level. The federal requirements for quality assurance, peer review, human involvement and approval at each level, a row per risk with its measure, owner, test and the reason it is enough, and the EU and Quebec equivalents.

The people affected: notice, explanation, recourse

In this lesson you learn to:

  • Write the notice and the explanation each regime asks for, in plain language.
  • Set up a real route to a person who can review and change the decision.
  • Answer a request about the factors behind an automated decision, with a sound time target.

Template: Decision letter and explanation checklist. A three part letter (what was decided, the main factors, how to ask for a review), a short published explanation of how the system works, and a checklist of what the Directive, Quebec law and the GDPR ask each to contain.

Approve, publish, keep it current

In this lesson you learn to:

  • Get an assessment approved and published before production where the Directive applies.
  • Spot the changes that reopen an assessment, and update it on a schedule.
  • Keep one approval and monitoring log that covers reviews, outcomes and incidents.

Template: Approval and monitoring log. The approvals and publications to finish before launch, a review schedule, and one log for changes, scheduled outcome checks and incidents, with what each meant for the assessment, under the Directive and the EU AI Act.

Frequently asked questions

We sell an AI system to a federal department. Does the Directive apply to us?

The Directive applies to the institution that uses the system for administrative decisions about clients, so the duty is the department's. TBS Directive on Automated Decision-Making ss.5.1, 5.2, 8.1, 8.2; Appendix A (administrative decision); Policy on Service and Digital s.6.1 In practice the department will ask you for most of what its assessment needs, and the course shows what to prepare.

Do we have to follow the NIST framework?

NIST states that its AI Risk Management Framework is intended for voluntary use. NIST AI RMF 1.0 (NIST AI 100-1) The course uses it as a shared method that fits under the federal, Quebec and EU duties.

Will I finish with something I can use?

Yes. Each lesson ends with a template: a screening table, a system profile, an impact level worksheet, a measures plan, a decision letter and an approval and monitoring log.

Is this course legal advice?

No. It is educational content that explains the rules and good practice in plain words. Each legal point in the lessons links to its official source so you can read it yourself, and for a decision that matters, ask a lawyer.

Can I take the course in French?

Yes. The lessons, videos, templates and exam are available in English and in French, and you can switch language from the top of any page.

How is my learning checked?

Each lesson ends with a short knowledge check, and the course ends with an exam built from scenarios.

Get the full course

One payment of 59 USD. Every lesson, the knowledge checks, the final exam and the certificate open on your account. The first lesson is free to read.

Prices are in US dollars. Terms of sale 路 Privacy notice

You can ask for a full refund of a course within 14 days of buying it, if you have not passed its final exam.

Final exam

When you have read the lessons, take the final exam. Pass it and we email you a certificate you can share and anyone can verify.

Take the final exam