Structure records of AI model interpretability so that they answer the questions an auditor will actually ask.
Interpreting Artificial Intelligence Model Decisions
A senior-level treatment of AI model interpretability, focused on what changes outcomes.
📍 Abu Dhabi🗓️ 5 training days📚 4 modules🎓 Accredited certificate
5intensive training days
4scientific modules
8training sessions
32detailed points
Course Overview
Technology choices around AI model interpretability are easy to reverse on paper and expensive to reverse in practice. Most organisations now hold more data about this strand of digital and data-driven work than they can actually act on. Plans for this part of digital and data-driven work often fail at the handover point between functions. Cases are chosen to expose the trade-offs in AI model interpretability rather than to illustrate ideal conditions. Participants gain a clear basis for measuring what the digital and data-driven work discipline has actually achieved. Post-incident reviews keep identifying weaknesses in the digital and data-driven work capability that were visible long before the incident. It is written for people who have to make AI model interpretability work with the resources they already have. Participants leave with a method for this aspect of digital and data-driven work, not a set of opinions about it. The final session converts the week's work on the digital and data-driven work discipline into commitments with owners and dates.
Expected Learning Outcomes
Specify the data AI model interpretability depends on, where it originates and who is accountable for its quality.
Set retention, lineage and deletion rules for the data flowing through AI model interpretability.
Document the decision record for AI model interpretability so successors understand why it is built this way.
Draw practical lessons from failures in AI model interpretability without assigning blame that suppresses reporting.
Establish monitoring that detects model or service degradation in AI model interpretability before users report it.
Apply a repeatable review cycle to AI model interpretability and act on what it produces.
Who Should Attend
Public sector digital leads applying AI model interpretability under procurement and privacy rules.
Risk managers assessing the exposure created by AI model interpretability.
Technology and digital transformation managers leading AI model interpretability.
Business partners who must understand AI model interpretability well enough to challenge it.
Vendor and contract managers overseeing suppliers involved in AI model interpretability.
Data and analytics leads responsible for the pipelines behind AI model interpretability.
Course Modules
AI model interpretability: people, skills and the change that follows
2 sessions · 8 pointsSession 1Reading the current state of AI model interpretability honestly
- Assess the regulatory obligations AI model interpretability triggers in each jurisdiction.
- Build the competence framework that supports AI model interpretability.
- Decide what will be stopped to create capacity for AI model interpretability.
- Map the handovers in AI model interpretability between functions and secure them.
Session 2Where AI model interpretability touches systems nobody wants to change
- Identify single points of dependency in AI model interpretability and reduce them.
- Plan how AI model interpretability is versioned and how a bad release is rolled back.
- Define the service level AI model interpretability must meet and what happens when it is missed.
- Verify that AI model interpretability still performs when input volume doubles unexpectedly.
AI model interpretability: architecture, integration and the existing estate
2 sessions · 8 pointsSession 1The pilot that actually settles the argument about AI model interpretability
- Record the reasoning behind each architectural choice in AI model interpretability.
- Set escalation thresholds for AI model interpretability that work out of hours.
- Set the metrics that will show whether AI model interpretability is drifting from its intended behaviour.
- Decide which legacy process AI model interpretability retires, and set the date.
Session 2Comparing AI model interpretability with recognised practice
- Design the pilot for AI model interpretability so that a negative result is still useful.
- Distinguish symptoms from causes when AI model interpretability underperforms.
- Agree who is on call for AI model interpretability outside working hours.
- Name a single owner for each element of AI model interpretability.
AI model interpretability: monitoring, drift and operational ownership
2 sessions · 8 pointsSession 1The data question everyone skips at the start of AI model interpretability
- Estimate the resource AI model interpretability requires to run as designed.
- Estimate compute and licensing cost for AI model interpretability at expected and at peak load.
- Identify the skills the team lacks to operate AI model interpretability independently.
- Record what was learned when AI model interpretability did not go as planned.
Session 2The governance AI model interpretability needs and the governance it does not
- Classify the data in AI model interpretability and apply access controls that match the classification.
- Define the exit route from the supplier supporting AI model interpretability.
- Confirm the retention and deletion rules applied to data inside AI model interpretability.
- Measure the current quality of the data feeding AI model interpretability before assuming it is usable.
AI model interpretability: business case, scope and the data it depends on
2 sessions · 8 pointsSession 1Making AI model interpretability secure without making it unusable
- Record the rationale for each significant choice made about AI model interpretability.
- Establish the boundary of AI model interpretability and record what sits outside it.
- Set out the decisions in AI model interpretability that require sign-off and by whom.
- Identify where judgement in AI model interpretability is legitimate and where it is not.
Session 2The paperwork for AI model interpretability that is actually needed
- Benchmark the organisation's AI model interpretability against comparable operations.
- Test AI model interpretability against edge cases drawn from real historical records.
- Confirm that reporting on AI model interpretability reaches the people who can act.
- Verify six months later that changes to AI model interpretability have held.
Choose the package that suits you
Silver Package
At least 3 people
USD1,250
- Workshop or Program Participation
- Airport Transfers
- Customized Badge
- Expert Mentorship (Private Sessions)
- Supervision & Secretarial Services
- Accredited Certificate of Participation
- Complete Training Kit
- Coffee Break
- Closing Ceremony
Gold Package
At least 3 people
USD1,850
- 5-night stay in a 5-star hotel
- Workshop or Program Participation
- Airport Transfers
- Customized Badge
- Expert Mentorship (Private Sessions)
- Supervision & Secretarial Services
- Accredited Certificate of Participation
- Complete Training Kit
- Coffee Break
- Closing Ceremony
Complete your registration
We will contact you within one business day to confirm.