Skills Readiness for AI

Skills readiness for AI means that people can perform the particular tasks a proposed workflow requires, recognise when its output is unsupported and use the correct route for clarification or escalation. Attendance at a course can be useful evidence of participation. It does not establish that a person can review a consequential output, protect restricted information or operate the agreed workflow under realistic conditions.

This guide is for Saudi team leaders, learning managers and AI sponsors deciding what capability must exist before a trial. The illustrative case is a publishing-rights catalogue team using an assistant to prepare internal comparisons of recorded permissions. The assistant does not grant licences, interpret disputed contracts or make a legal determination. Synthetic records are used to assess skills before anyone considers access to sensitive commercial material.

Define the work before defining the training course

“Learn AI” is too broad to support a readiness decision. Describe what the employee will actually do: select an eligible record, prepare a bounded request, inspect the source, compare the answer with the recorded conditions and route an unresolved issue. These tasks can require different skills from those needed to build or administer the system. Do not require every user to become a model developer.

For the catalogue analyst, the critical task is preserving distinctions among title, format, language, territory and permitted period. A fluent summary that merges an ebook permission with an audio restriction is not acceptable simply because it sounds confident. The analyst needs to recognise the mismatch and know who can resolve it. The readiness exercise tests that behaviour without asking the analyst to practise law.

The UK AI skills tools package offers a role-sensitive reference spanning technical, responsible-use and non-technical skills. It applies to England, not as a Saudi requirement. Its relevance here is the principle of matching capability to work; the task demonstrations below are an original assessment design.

Separate user, reviewer and owner capabilities

A user may know how to request a comparison but lack the knowledge to accept it. A reviewer may understand the catalogue rules but not know how to identify an unavailable source. An owner may approve the trial without understanding what a performance result actually establishes. Skills readiness should reflect those differences instead of giving everyone the same generic questionnaire.

Illustrative task capabilities for a publishing-rights catalogue team

Role	Task to demonstrate	Evidence of capability
Catalogue user	Select the correct record and request a bounded comparison	Correct identity, permitted input and clearly stated task
Output reviewer	Check material conditions against the source	Located evidence, identified mismatch and reasoned disposition
Workflow owner	Decide whether a case belongs inside the trial	Applied scope rule and a clear escalation decision
Support coordinator	Distinguish a task error from an access or tool problem	Useful issue record without unnecessary sensitive content
Trial sponsor	Interpret the findings and fund a proportionate next step	Decision that respects unresolved conditions and test limits

Several roles may sit with one person, but the evidence should still cover each decision they will make. Use ownership, evaluation and adoption readiness to check that capability is accompanied by authority and time. A skilled employee who is not authorised to make a decision cannot close that gap through training alone.

Build exercises from the mistakes that matter

Start with the errors that could undermine the proposed task. In the illustrative catalogue, these include confusing two editions, overlooking a format restriction, treating a missing permission as approval and accepting a generated statement that cannot be located in the source. The exercise should reveal whether a person notices the relevant condition, not whether they memorised the vocabulary of AI.

Prepare ten synthetic cases: four with clear recorded conditions, three with missing information, two with conflicting records and one outside the permitted input scope. This is a designed learning set, not a representative sample of publishing contracts. Label it as synthetic and keep the expected handling rationale available to assessors. Avoid using identifiable customer or author material merely to make a training exercise feel realistic.

For example, one record may permit an ebook in a specified language while separately excluding audio. The expected action is to preserve that distinction in an internal comparison and refer any disputed interpretation to the authorised owner. A request for a broad “all digital rights” conclusion should not cause the learner to erase the limitation. The exercise tests source discipline, not legal advice.

Observe the process as well as the final answer

A correct answer can be produced for the wrong reason. Someone may guess, copy a previous answer or rely on the assistant's tone. Ask the learner to show the source passage, identify the condition that matters and explain what they would do if the evidence were missing. The demonstration should make the reasoning inspectable without demanding a lengthy essay for every routine task.

Use a short observation record: task attempted, source selected, material condition recognised, output checked, uncertainty handled and next action. Include the tools and reference materials the learner was allowed to use. This helps distinguish a task that can be performed with a clear job aid from a task that still requires specialist support. Both can be legitimate operating designs.

Do not confuse fast completion with competence. An analyst who pauses to identify a source conflict may be performing better than one who produces an immediate unsupported answer. Measure time only alongside accepted quality and appropriate handling. If the work requires careful judgement, the readiness question is whether the organisation can support that judgement, not whether everyone can imitate a rapid demonstration.

Assess information handling in the actual task context

Generic reminders to avoid sensitive information are insufficient if the user cannot recognise which fields matter. In the exercise, include an obviously synthetic record marked as outside the permitted trial material. Ask the learner to explain why it should not be submitted and where they can obtain a suitable substitute. This assesses a concrete decision without exposing real confidential information.

Teach the boundary between an approved account and a convenient personal tool. The employee should know which environment is authorised for the proposed work and how to report an access problem. Do not assume that a subscription, familiar brand or successful login establishes permission to process every category of business information. The applicable arrangement must be confirmed by the organisation.

Keep skills assessment separate from legal or security certification. A learner's good handling of synthetic cases does not establish that the production data flow satisfies every obligation. It provides evidence about a person's behaviour within the exercise. Permissions and provenance readiness addresses the corresponding organisational access conditions that must also be established.

Test Arabic and English source comprehension separately

A bilingual interface does not make every operator a bilingual evaluator. Identify the language of the source and the language of the output for each assigned task. An analyst may be comfortable reading a generated Arabic summary while unable to verify the decisive qualification in an English source. That is a support requirement to address, not a reason to assume the output is correct.

Use paired cases that preserve the same material conditions while allowing natural wording in each language. Ask reviewers to locate a restriction, identify the relevant format and explain an uncertainty. Do not penalise harmless stylistic differences. Focus on whether the meaning that affects the internal decision has been preserved.

If the underlying Arabic and English records conflict, the correct skill may be escalation rather than translation. An employee should not silently choose the broader permission because it makes the task easier. Use bilingual source-conflict readiness to establish who can settle the discrepancy. A language course cannot repair the absence of an authoritative business decision.

Distinguish a knowledge gap from a broken work design

When a learner struggles, inspect the conditions before prescribing more training. Were the source records identifiable? Was the acceptance rule clear? Did the account provide the required evidence? Was the task within the role's authority? A capable person cannot reliably verify an answer if the workflow withholds the source or presents two conflicting definitions without a resolution route.

Diagnosing a weak task demonstration

Observed problem	Question to resolve	Likely next action
Cannot locate supporting condition	Is source navigation available and understood?	Improve access or practise evidence retrieval
Misses a clear restriction	Does the learner understand the task-specific distinction?	Targeted practice with feedback
Two assessors expect different answers	Is the reference judgement defensible?	Resolve the rule before retesting staff
Knows the issue but cannot escalate	Is an authorised route defined?	Repair the operating model
Performs well only with personal coaching	Will that support exist during the trial?	Provide a job aid, supervision or a narrower role

The diagnosis matters commercially. Training budget spent on an access problem will not produce the missing permission. Conversely, buying another platform may not help an employee who needs practice distinguishing a supported statement from an unsupported inference. Record the evidence for the chosen remedy so that the next assessment tests the right change.

Set readiness criteria without inventing a universal pass mark

A single percentage can conceal an important capability gap. A person might perform nine routine tasks well and still disclose restricted material or treat a missing permission as approval in the tenth. The owner should define which behaviours are prerequisites for the assigned role and which can be supported during a supervised trial. There is no universal eighty-percent rule that fits every AI task.

Use explicit dispositions such as ready for the defined supervised task, ready after a specific practice requirement, needs a different support arrangement or not yet assessed for this role. Keep the condition and evidence with the conclusion. A result should not become a permanent label about an employee's general ability or suitability for every future technology.

Give learners access to the expectations and useful feedback. Explain why a response was accepted, what evidence was missing and how improvement will be demonstrated. A skills-readiness process should help people perform the work, not surprise them with hidden criteria. Where employment decisions are involved, the organisation needs its normal appropriate review and procedures beyond this workflow exercise.

Design learning around a demonstrated gap

After the baseline exercise, choose a specific learning activity. An analyst who overlooks format restrictions may need paired examples contrasting ebook and audio conditions. Someone who cannot locate evidence may need practice navigating the catalogue. A sponsor who overinterprets test results may need an exercise distinguishing coverage, accuracy and business benefit. Different gaps require different interventions.

Include practice, feedback and a new demonstration. Showing the same solved example again can test memory rather than transfer. Use a different synthetic case with the same underlying skill requirement and ask the learner to explain the decision. Retain enough of the assessment record to show what changed without collecting unnecessary personal commentary.

Job aids can be part of readiness. A concise record-selection checklist or escalation guide may support reliable work more effectively than expecting every detail to be memorised. Test the task with the aid that will actually be available. If success depends on an expert sitting beside the learner, record that supervision requirement honestly in the proposed trial arrangement.

Plan the time needed for people and assessors

Skills preparation consumes operating capacity. Suppose eight analysts each complete a forty-five-minute exercise. That is three hundred and sixty minutes, or six participant-hours in total. If an assessor then spends twenty minutes reviewing each person's evidence, that adds one hundred and sixty minutes, or two hours and forty minutes, of assessor time. These are illustrative planning assumptions, not training-market benchmarks.

The two totals describe different workloads and should not be substituted for each other. They also exclude preparation, feedback, repeat practice and any group calibration. Schedule those activities alongside normal work. A course purchased without time for practice may create an attendance record while leaving the workflow's actual skill needs unresolved.

Use the smallest exercise that can reveal the relevant gap. A team assigned only internal comparisons may not need the same preparation as one authorised to change records or contact counterparties. Link the learning scope to the proposed authority. If the workflow later becomes agentic, revisit the skills needed to understand approvals, actions and uncertain outcomes through agentic-workflow readiness.

Check transfer from a classroom exercise to a controlled trial

A training environment may simplify the very conditions that make the work difficult. Before the trial, confirm that users can access the approved environment, recognise the real source structure and use the actual support route. Keep the first operational scope bounded while observing whether the learned behaviour transfers. Do not assume that a certificate or successful synthetic exercise establishes full production competence.

Compare the conditions explicitly: source formats, language, task limits, available job aids, review expectations and time pressure. If a synthetic exercise used perfectly labelled records while the real catalogue contains unresolved duplicates, address that difference. The learner may be ready for the taught task while the intended workflow remains unready.

Use readiness versus proof of concept to separate permission to attempt a controlled trial from evidence that the trial achieved its objectives. Skills readiness supports the decision to test within specified conditions. The trial still has to establish whether the combined people, process and technology perform acceptably in the agreed setting.

Record a release condition that the team can actually apply

A useful capability record says more than “training complete”. For the illustrative analyst, it could state that the person demonstrated comparison of clearly identified catalogue records using the approved job aid, while conflicting or missing conditions must go to the rights owner. Record the exercise date, tasks observed, permitted support and the assessor's reason for the conclusion. This describes a bounded work arrangement rather than granting new legal or commercial authority.

Suppose the analyst initially overlooks an audio restriction, completes targeted practice and then handles two different synthetic format cases correctly. Retain the earlier gap and the later evidence. The follow-up supports a judgement about that specific skill under those conditions; it does not erase the need for supervision or prove that every future record will be handled correctly. The team can use the evidence to choose a proportionate first trial assignment.

Make the record usable by the operational manager. They should be able to see which cases the employee may handle, which require support and who provides that support. If the manager cannot translate the conclusion into an assignment, the assessment has remained too abstract. A narrowly defined permitted task is more helpful than an unexplained score that appears precise but says little about the work.

Allow the employee to ask for help when a new case falls outside the demonstrated conditions. Appropriate escalation is a capability to preserve, not a failure to hide. This keeps the release decision connected to real evidence while giving the organisation a clear route for extending the role later.

Keep capability evidence current without endless retraining

Reassessment should follow meaningful change. New permissions, a different source language, altered catalogue rules or a shift from drafting to action may introduce new skills. A cosmetic interface change may require a brief orientation rather than a full programme. Record which task assumptions changed and choose a proportionate follow-up demonstration.

Also plan for staff turnover and absence. A workflow dependent on one experienced analyst may be ready for a narrow supervised trial but not for wider operation. Document the job aid, assessment rationale and support route so that another person can prepare for the role. Capability should not depend entirely on informal knowledge that disappears when one employee is unavailable.

Feed recurring gaps into a focused remediation backlog and connect role preparation with Enterprise Data & AI Readiness. For a skills-readiness enquiry, bring the proposed task, the roles involved and examples of acceptable and unacceptable decisions. These inputs allow the assessment to focus on demonstrated work rather than generic confidence about AI.