AI Training for Saudi Managers: Use Cases Before Tools

AI Training for Saudi Managers: Choosing Use Cases Before Tools

Picture a management team at a Riyadh company returning from two days of AI training. They have watched a dozen tools generate slides, summarize reports and draft emails. On Sunday morning, each manager has a new account and the same question: which of my team’s tasks should we actually use this for? By Tuesday, one has pasted a confidential supplier contract into a public chatbot, and another has stopped logging in.

The AI training was not wrong about what the tools can do. It started in the wrong place. Managers do not need to master every tool; they need to decide which work in their area is worth trying, what data may be used and who checks the result. Those decisions outlast any single product.

This article is for HR teams and functional training owners in Saudi organizations who are designing or choosing AI training for managers. It sets out a use-case-first approach in three steps: sorting tasks that are suitable for a trial, defining data constraints, and building human review into every output. It names no AI products; tool choices belong to your IT and information security teams.

Why AI Training for Managers Should Start with Use Cases

AI technologies change monthly, but the judgment of which tasks suit them lasts. Research supports that emphasis. In a field experiment with 758 knowledge workers, run with Boston Consulting Group and published as a Harvard Business School working paper by Dell’Acqua and colleagues, AI assistance improved performance on some tasks and worsened it on others, a pattern the authors call a jagged technological frontier. On realistic tasks inside that frontier, participants using AI completed 12.2% more tasks, 25.1% more quickly on average. On a task outside it, they were 19% less likely to produce correct solutions than participants without AI.

For managers, the lesson is practical: tasks that look similar can fall on opposite sides of that frontier, and the only reliable way to know is to test them carefully. AI training that teaches managers to select, test and review use cases prepares them for whatever tools their organization approves next. AI training that only demonstrates features prepares them for the tools on the trainer’s screen.

Step 1: Start AI Training by Sorting Tasks for a Trial

The first exercise uses each manager’s own work. Ask managers to list 10 to 15 recurring tasks from their team, then sort each into one of three groups:

• Try now: the output is a draft a person reviews anyway, errors are easy to spot and cheap to fix, and the data involved is approved for use. Examples include structuring meeting notes, drafting a routine letter or summarizing a long internal report.

• Try with conditions: the task is promising but needs redacted data, a specialist reviewer or a narrower scope first, such as categorizing customer complaints or drafting a tender clarification.

• Not now: the task involves decisions with legal, financial, safety or personal consequences for individuals, such as screening CVs, rating performance or approving credit, or it depends on facts the tool cannot check.

For each task in the first two groups, managers complete a one-page use-case card: the task, how often it occurs, the current effort, the expected benefit, the data involved, the reviewer and an observable success criterion, for example ‘the first draft needs only minor edits’. The card turns enthusiasm into a test that can pass or fail. Good AI training ends this exercise with three or four cards per manager, not a list of tools to try.

Step 2: Set Data Constraints Early in AI Training

In Saudi Arabia, data questions belong at the start of AI training, not on a closing slide. The Saudi Data and Artificial Intelligence Authority’s AI Ethics Principles apply to all AI stakeholders designing, developing, deploying, implementing, using or being affected by AI systems in the Kingdom. Under the privacy and security principle, the use of personal information should be limited to what is necessary for the proper functioning of the system.

For managers, that becomes a simple habit: before using any tool, ask what you are about to share.

AI training can build the habit with a short data classification step on every use-case card:

• Public: information already published, which can usually be used with approved tools.

• Internal: routine business information, to be used only with tools your organization has approved for it.

• Confidential or personal: contracts, pricing, customer records, employee files and health data, which stay out of any tool unless your data governance and information security teams have explicitly cleared that use.

Then practice redaction during the AI training: give managers a sample report and ask them to remove names, figures and identifiers before using it. This article is not legal advice; the rules for your data come from your organization’s data governance policies and the regulations it follows.

Step 3: Practice Reviewing Outputs in AI Training

Every AI output needs a named human reviewer. The U.S. National Institute of Standards and Technology’s Generative AI Profile (NIST AI 600-1) lists confabulation among the risks of generative AI: the production of confidently stated but erroneous or false content by which users may be misled or deceived. SDAIA’s principles point the same way, stressing human-controlled AI systems in which authorized users can intervene.

In AI training, human review is a skill to practice, not a rule to announce. Give managers outputs with planted errors, such as a wrong figure, an invented reference or a misstated policy, and ask them to find each one. Then agree, for each use case:

• Who reviews the output, and whether that person has the subject knowledge to spot errors.

• What the reviewer checks: facts, figures, names, dates, tone and alignment with policy.

• When an output is discarded rather than corrected.

• How AI assistance is recorded or disclosed, in line with your organization’s policy.

If your learning team also uses AI to draft training content, the same discipline applies; our article on reviewing AI-assisted course creation before approval covers that case.

What Use-Case-First AI Training Looks Like

A practical AI training design for managers runs over a few weeks rather than two intensive days:

• Session 1: task sorting and use-case cards, built on each manager’s own team.

• Session 2: data classification and redaction practice with sanitized samples.

• Session 3: a hands-on trial of one or two ‘try now’ cards with a tool your IT team has approved, plus a review exercise with planted errors.

• Follow-up: a four-week trial after the AI training in which each manager runs one use case, keeps a short review log and reports what worked and what did not.

Notice that in this AI training the tool appears in the third session, not the first. Demonstrations still matter, but they make sense only once managers know what they want to test. To check whether the AI training moved managers from watching demonstrations to practicing, the approach in our guide to testing whether training moves employees from viewing to practice can be adapted.

A Hypothetical Example: Project Managers at an Engineering Firm in Riyadh

This example is hypothetical and is included only to show how the method works.

An engineering and facilities company in Riyadh and Dammam plans AI training for 18 project and operations managers. Instead of a tool showcase, the training owner asks each manager to bring a list of recurring team tasks, and together they sort 140 of them.

Three use cases make the ‘try now’ group: summarizing weekly site reports into a one-page update, drafting first replies to routine tender clarification requests, and turning meeting notes into action lists. Grouping customer complaints by theme goes into ‘try with conditions’ until complaint records can be anonymized. Screening CVs is placed in ‘not now’, because it involves personal data and decisions about individuals.

Each use case gets a reviewer: project managers check site summaries against the original reports, and the commercial lead approves every tender reply. In the month after the AI training, managers log how much editing each draft needed and every error they caught. The training owner then reviews the logs with IT to decide which use cases to extend, narrow or stop, without claiming results the logs do not show.

Questions to Ask Before Choosing AI Training for Managers

Whether you build the program internally or buy it, ask:

• Does the AI training start from participants’ own tasks or from tool features?

• Is it tool-neutral, or tied to one vendor’s product?

• Does it cover data classification in line with Saudi principles and your own policies?

• Do managers practice reviewing flawed outputs, not just generating new ones?

• Are Arabic-language tasks and documents included where teams work in Arabic?

• Is there a follow-up trial, and how will results be reported?

Be cautious about AI training proposals that promise fixed productivity gains; as the research above shows, results depend on the task. If you are still defining what managers need, our article on building a corporate training brief from a performance problem can help you frame the requirement first.

Summary

AI training for Saudi managers is most useful when it starts with the work, not the tools. Sorting tasks into ‘try now’, ‘try with conditions’ and ‘not now’ gives managers a clear starting point, and use-case cards turn interest into tests that can succeed or fail.

Data constraints and human review make those tests safe. When AI training builds both habits before the first demonstration, managers can apply them to whatever tools their organization approves, now and later.

Click here. If you are planning AI training for managers in a specific function, the SkillUp MENA team can discuss the requirement with you.