AI Skills Gap in Saudi Arabia

An AI skills gap is the difference between the work an organisation needs people to perform and the competence it can actually demonstrate for that work. It is not simply a shortage of employees with AI in their job title. A team may have capable analysts yet lack someone who can evaluate an unfamiliar failure, maintain a release or explain an operational limitation to users.

Saudi employers should diagnose that gap at task level before buying training or recruiting another generalist. Separate missing competence from insufficient time, unclear authority, unavailable data and an unsuitable tool. These problems can look similar when a project is waiting, but they require different responses. The purpose of an assessment is to identify the smallest credible intervention that makes the intended work sustainable.

Define the gap without inventing a national shortage statistic

A company’s own evidence can establish a capability gap even when no reliable national estimate exists for its exact roles. Describe the task, required performance, current evidence and missing condition. Avoid using an unsourced claim about the Saudi labour market as a substitute for understanding your team. A market narrative cannot tell you which responsibility is uncovered on Monday morning.

The Ministry of Human Resources and Social Development’s Saudi Skills Taxonomy provides a shared way to relate skills and occupations. That is useful context for consistent role language. An employer still needs evidence for the actual task, tools and responsibility in its organisation; a taxonomy entry does not establish that an individual is ready for a particular deployment.

Start with work expected in the next operating period, not a speculative list of every AI specialism. Include maintaining existing systems as well as building new ones. A company can overinvest in experimentation skills while leaving evaluation, source maintenance and user support unassigned. The gap should be defined by the service the business intends to run.

Map tasks before mapping job titles

Break the proposed capability into observable activities. Examples include preparing an authorised dataset, writing a task-specific evaluation case, explaining an unexpected result, approving a source update, releasing an agreed change or guiding a user through an exception. Use verbs and outputs. “Understands AI” is too broad to support a staffing decision.

For each activity, record the source or input, accepted output, operating frequency, required independence and consequence of error. Identify who can judge the work. A person may be able to perform an activity under supervision but not review someone else’s work. Both can be useful capabilities, provided the operating plan reflects the distinction.

Do not assume one job title covers every activity. The Ministry of Communications and Information Technology (MCIT) Future Skills digital-skills resource distinguishes skills from jobs and responsibility areas. Use that distinction when describing a role: a role can require several skills, and one skill can appear in several roles. The resulting task map should be comprehensible to managers and practitioners alike.

Use evidence that shows performance in context

Ask for a work sample, a supervised demonstration or a recent example the person can explain. Match the evidence to the intended responsibility. Someone who can reproduce a tutorial may be ready for a guided exercise, but that does not establish independent judgement when the source changes or the result becomes ambiguous. Assess the relevant variation rather than demanding unrelated technical knowledge.

Use simple internal evidence labels: not yet observed, demonstrated with assistance, demonstrated independently, or able to review another person’s work. These are proposed assessment labels, not an official Saudi qualification scale. Define what each means for the task and record the conditions under which the evidence was obtained. A label without its supporting example becomes another unsupported opinion.

Keep skills assessment separate from personality judgements. A confident presentation is not proof of operational competence, and a careful employee who asks for clarification may understand the risk better than someone who proceeds quickly. Use consistent criteria, reasonable access to practice and feedback that explains the observed gap. The aim is a better capability decision, not a public ranking of employees.

Separate competence, capacity, access and authority

Ask four questions when work is blocked. Can someone perform it to the required standard? Do they have enough allocated time? Can they access the permitted sources and systems? Are they authorised to make the relevant decision? A no on any one can stop the workflow. Training addresses only some of these conditions.

A skilled reviewer with no time is a capacity problem. A person who knows how to check a source but cannot access it has an access problem. Two teams waiting for each other to approve a rule have an ownership problem. Calling all three a skills gap can lead to courses that leave the original obstruction intact.

Record both the capability owner and the operational decision owner. They may be different people. A technical specialist can assess model behaviour, while the business owner decides whether the resulting limitation is acceptable for the task. Leadership AI literacy helps managers ask the necessary questions without expecting one technical employee to make every business decision.

Work through a field-service team capability map

Consider a fictional Saudi field-service business preparing an internal workload forecast and exception summary. The output supports planning; it does not independently dispatch technicians or approve maintenance. The business already has strong analysts, but its proposed recurring service needs more than analysis. The following weekly hours are invented planning assumptions, not staffing benchmarks.

Required work and demonstrated allocated capacity

Activity	Required hours	Allocated hours with demonstrated competence	Task-specific gap
Prepare and check source data	8	10	0; two hours above this task’s need
Evaluate forecast and exception outputs	6	2	4
Prepare and verify an agreed release	4	0	4
Handle operating checks and exceptions	4	2	2
Review planning rules with the business	3	3	0
Coordinate user learning and feedback	2	1	1
Total	27	18	11 hours of uncovered task demand

The arithmetic matters. Required work totals twenty-seven hours and allocated competent capacity totals eighteen, suggesting a simple nine-hour difference. But two available hours sit in data preparation, where the requirement is already covered. Only sixteen hours match the listed task demand, leaving eleven uncovered. Spare time in one skill does not automatically substitute for a different capability.

The four-hour release gap is initially a competence gap because no internal person has demonstrated that task. The two-hour operating gap may be a time-allocation problem if the existing practitioner can perform the work but is assigned elsewhere. The evaluation gap needs investigation: is there a capable person without time, a developing practitioner needing coaching, or no available expertise?

The map therefore does not immediately justify hiring another analyst. It identifies the questions a staffing decision must answer. Reallocating two data hours could help only if the person can perform the destination task or is given a credible development path. Count the capacity after that competence is demonstrated, not simply after a manager changes a spreadsheet cell.

Choose between development, reassignment, hiring and specialist support

Develop an existing employee when the target skill is close enough to their current work, supervised practice is available and the organisation can afford the learning period. Define the task they will demonstrate at the end. A course can provide concepts, but the employer still needs an opportunity to practise the actual responsibility and receive competent feedback.

Reassign work when competence already exists but its time is committed elsewhere. Make the tradeoff explicit: which other activity will move, reduce or receive support? A nominal allocation that leaves all previous commitments unchanged is not new capacity. Managers should agree the calendar and coverage, particularly where the same specialist is shared by several projects.

Hire when a recurring responsibility is substantial enough to justify a durable role and the organisation can describe the evidence it needs. Write the role around tasks, decision boundaries and collaboration requirements. Avoid a long list of technologies that combines several different jobs without explaining the actual work. A clearer role is easier to assess and support after appointment.

Use specialist support when the gap is narrow, urgent, intermittent or requires experience the company cannot develop immediately. Specify what the specialist will do and what knowledge remains with the employer. The internal team still needs enough understanding to accept the work and operate the retained responsibilities. Capability ownership decisions help clarify which knowledge the business should preserve.

Prioritise gaps by dependency and consequence

Do not rank gaps only by the number of missing hours. A four-hour release responsibility may block the whole service even when a larger task can continue manually. Identify which missing capability prevents a bounded, acceptable operating arrangement and which mainly limits efficiency. This helps the organisation sequence interventions without pretending every skill must be developed simultaneously.

In the fictional team, the business could keep the existing planning process while a qualified specialist supports a controlled evaluation. That does not close the release gap or authorise production use. It creates a limited learning arrangement with explicit boundaries. The manager should record what may proceed, what must wait and which evidence is required to move to the next responsibility.

Examine interactions between gaps. If a new employee needs supervision from the same person already short of evaluation time, recruitment may initially increase the bottleneck. If a partner assumes the business will prepare test cases but nobody owns that task, the contract may leave an essential dependency uncovered. Map the support each intervention consumes as well as the capacity it promises to add.

Use a short action record for each priority: missing task, type of gap, interim arrangement, intervention owner, required evidence and review date. Keep proposed future capacity separate from available capacity. Revisit the record when evidence arrives and retire actions that no longer address the real constraint. This is a practical staffing decision aid, not a comprehensive transformation roadmap or a guarantee that a particular hiring sequence will work.

Design an assessment that a practitioner can trust

Use a realistic but appropriately bounded task with permitted data. Provide the same instructions and resources to people being assessed for the same responsibility. State what help is allowed and record it. An assessment should not secretly depend on knowing a particular interviewer’s preferred phrasing or having access to a private dataset that was never supplied.

For evaluation work, ask the person to identify a meaningful failure, construct a case that exposes it and explain the acceptance decision. For release work, ask for the evidence and checks required to introduce an agreed change. For user support, ask how they would distinguish an input problem from a system limitation. Keep each exercise aligned to the actual role.

Explain the result with specific observations. “Recognised an outdated source but did not identify which output depended on it” provides a useful development target. “Not sufficiently innovative” does not. Allow a fresh demonstration after targeted practice where appropriate, and distinguish a learning decision from any formal employment process the organisation separately applies.

Turn a training purchase into a capability transfer plan

Before approving training, record the starting evidence, target task, practice environment, coach and independent demonstration. Agree how much employee and supervisor time will be available. If the business cannot provide appropriate practice or feedback, the course may increase awareness without closing the operational gap. Describe that outcome honestly rather than counting every certificate as deployable capacity.

Use applied employee training for the learning design. Keep the target narrower than becoming an AI expert. An employee who learns to evaluate one defined summary task can add real value without becoming responsible for every model, source and application in the company. Narrow competence is useful when its boundary is visible.

For a partner engagement, request a teach-back or joint work session tied to the retained responsibility. The internal person should explain the method and perform a fresh example, not merely receive documents. Schedule this before the partner leaves so misunderstandings can be corrected while the relevant expertise is still available.

Protect continuity when knowledge is concentrated

A capability can be demonstrated and still be fragile if only one person holds it. Identify tasks whose absence would stop the service, then decide what backup is proportionate. A backup may need a documented manual procedure, access to a qualified specialist or sufficient practice to cover a bounded task. It does not always require duplicating every role.

Record the assumptions needed for handover: sources, known limitations, accepted examples, unresolved questions and where to find current instructions. Keep documentation connected to practice. A folder of old material is weak evidence that another person can perform the work under time pressure. Test the handover with a realistic task and observe what remains unclear.

Revisit concentration when workload grows or the service becomes more consequential. The same staffing arrangement may be adequate for an occasional internal draft and inadequate for a daily operational dependency. Use capacity planning and total cost of ownership to include the resources needed to sustain the capability, not just create it.

Measure whether the gap actually closed

Track the original task and its missing condition. Did independent performance improve? Is the promised time allocated? Can the person access the required source? Is decision ownership clear? A headcount increase or training completion is an input; it does not prove all four conditions now hold. Review the result against the same work boundary used in the diagnosis.

Keep evidence current when the task changes. A new source, language requirement or responsibility may require a focused reassessment. For bilingual tasks, Arabic quality criteria can identify the language expertise actually needed. Avoid assuming every technically skilled person can judge every register, dialect expression or business commitment in the intended output.

The adoption-barrier guide helps when the remaining obstacle is not competence. The Saudi AI authority hub connects skills to the broader programme. Open for AI consulting can support a task-based capability assessment; bring the responsibilities your team cannot yet cover and the evidence of where work is waiting.