AI & Assessment Transparency Statement

1. About This Statement

Talent Cloud Global Ltd ("Talent Cloud", "we", "our" or "us") provides technology that helps organisations design, deliver, govern and measure workforce programmes.

Talent Cloud supports assessments, readiness evaluation, matching, interviews, programme reporting and other activities that may involve human judgement, calculations, algorithms, automated processing or artificial intelligence.

These processes do not all work in the same way and should not all be described as AI.

This Statement explains:

• where human assessment is used;
• how readiness indicators are calculated;
• how technology-assisted matching works;
• where third-party AI-assisted services may be used;
• how programme reporting and analytics operate;
• what these technologies are intended to do;
• what they are not intended to do;
• where human responsibility remains important;
• how protected and sensitive information should be treated;
• how material technology changes are governed; and
• how an individual can question or challenge an output.

The purpose of this Statement is to provide meaningful transparency rather than applying the label "AI" to every automated or calculated platform feature.

This Statement should be read alongside our Privacy Notice, which explains how personal information is processed and the respective roles that Talent Cloud, customers and programme organisations may have.

2. Our Approach

Talent Cloud is designed as a Workforce Programme Operating Platform.

The platform helps authorised organisations coordinate activities such as:

• learner and participant administration;
• programme delivery;
• assessments;
• evidence;
• work-readiness evaluation;
• employer engagement;
• opportunity matching;
• placement activity;
• programme assurance;
• reporting; and
• programme outcomes.

Technology can help authorised people organise evidence, calculate indicators, compare relevant information, identify potentially relevant opportunities and support decision-making.

It should not remove appropriate human accountability for material decisions affecting individuals.

Our approach is based on the following principles.

Transparency

People should receive meaningful information about significant assessment and technology-assisted processes affecting them.

Human Accountability

Material programme, assessment, placement and employment decisions should remain attributable to the people or organisations responsible for making them.

Purpose Limitation

Assessment and technology-assisted outputs should be used for legitimate and defined programme or service purposes.

Proportionality

The significance placed on an output should reflect what the output actually measures, how it was produced and the limitations of the underlying process.

Challenge and Correction

Individuals should have an appropriate route to raise concerns about materially inaccurate information or inappropriate use of an output.

Evidence Over Assumptions

A calculated score, assessment result, match or AI-assisted analysis should not be treated as inherently correct merely because technology produced, calculated or assisted it.

3. Assessment Decisions and Automated Decisions Are Different

An assessment decision and an automated decision are not necessarily the same thing.

An authorised assessor may make a substantive assessment judgement about a learner.

For example, an assessor may determine that a learner is competent or not yet competent based on the applicable assessment process and evidence.

The fact that this result is then recorded, stored, calculated, displayed or used by software does not make the underlying assessment judgement an automated decision.

Similarly, a system may calculate an indicator from information entered by a human without independently making the underlying judgement.

Understanding who or what actually made a decision is important when interpreting an assessment or technology-assisted output.

4. Different Types of Processing

Talent Cloud currently supports several materially different forms of assessment and technology-assisted processing.

For transparency, we distinguish between:

1. human assessment;
2. calculated readiness indicators;
3. technology-assisted matching;
4. third-party AI-assisted interview or assessment services; and
5. deterministic programme reporting and analytics.

The technology used may vary by product, customer and programme.

Not every Talent Cloud user will encounter every type of processing described in this Statement.

5. Human Assessment

Within Talent Cloud academy and learnership workflows, assessment outcomes may be determined by authorised human assessors.

An assessor may record information such as:

• whether a learner is competent or not yet competent;
• a numeric score where applicable;
• assessment feedback;
• assessment dates;
• supporting evidence; and
• other information relevant to the applicable assessment process.

The platform stores and presents those outcomes and may apply programme rules based on the information entered.

For example, a recorded assessment outcome may contribute to a learner's module, tracker or programme progress state.

Where the underlying assessment is human-assessed, the platform should not describe the assessment judgement as an AI decision merely because software stores, calculates, presents or subsequently uses the result.

The assessment judgement remains attributable to the authorised assessor.

6. Readiness Indicators

Talent Cloud may support structured work-readiness evaluation.

In the current readiness model, authorised assessors may evaluate defined criteria such as:

• communication;
• problem solving;
• teamwork;
• motivation;
• reliability;
• learning agility;
• digital literacy; and
• role-related skill.

Criteria may be scored using a defined scale.

The platform may then calculate an overall readiness indicator from the scores entered by the assessor.

What the Calculation Does

The calculation provides a consistent way of summarising assessor-entered readiness information.

It can help programme teams understand a participant's assessed readiness across defined criteria and identify potential areas for development or further support.

What the Calculation Does Not Do

The calculation does not independently:

• determine whether a person should be employed;
• predict workplace success with certainty;
• decide whether a learner should remain in a programme;
• assign a learner to an employer;
• make a final placement decision; or
• determine a person's overall worth or potential.

A readiness indicator is one source of information that may support programme and employer decision-making.

It should be interpreted in context rather than treated as a definitive prediction of future performance.

7. Technology-Assisted Matching

Where matching functionality is enabled, Talent Cloud may use technology to identify and rank potentially relevant relationships between individuals and opportunities.

Matching may use a combination of:

• programme and cohort rules;
• eligibility filters;
• skills;
• qualifications;
• location;
• availability;
• employment preferences;
• employer or opportunity requirements;
• semantic comparison of relevant profile and opportunity information; and
• configurable weighting across relevant matching dimensions.

The current matching architecture combines rules and filters with semantic comparison using locally operated embedding technology.

Semantic Comparison

Semantic comparison allows the system to compare relationships between relevant profile and opportunity information without relying only on exact keyword matches.

This can help identify potentially relevant relationships where similar skills, requirements or concepts are expressed using different language.

Match Ranking

The system may calculate and store a match result together with information explaining relevant components of the match.

A higher-ranked match indicates that the system identified a stronger relationship according to the factors, rules and configuration being used.

A match score measures configured relevance. It does not measure human worth, overall employability or overall candidate quality.

A higher-ranked match therefore does not mean that one individual is objectively "better" than another person or that an employer should select that individual.

Human Responsibility

Matching outputs are intended to support authorised people.

A match does not itself:

• make an employment decision;
• hire a person;
• reject a person;
• guarantee an interview;
• guarantee a placement;
• allocate a person to an employer; or
• determine that another person is unsuitable.

Material employment and placement decisions remain with authorised people and organisations.

8. Protected, Demographic and Sensitive Information

Talent Cloud programmes may process information such as:

• race or ethnicity;
• B-BBEE-related information;
• gender;
• disability information; and
• other demographic or protected information.

Some of this information may receive additional protection under applicable law.

The availability of race, B-BBEE, gender, disability or other protected or sensitive information within Talent Cloud does not mean that the information is automatically used to rank employment suitability.

Similarly, the presence of information in a learner or programme record does not mean that every field is automatically supplied to an algorithm, assessment or matching process.

Where protected or sensitive information is used in an eligibility rule, programme process, reporting requirement, matching rule or other technology-assisted process:

• there should be an identified purpose for that use;
• there should be an appropriate legal basis or justification;
• the use should be relevant to the applicable programme or requirement;
• appropriate fairness and discrimination considerations should be taken into account; and
• access should be limited according to authorised roles and legitimate requirements.

The use of a protected characteristic in one legitimate programme context does not mean that it is appropriate to use that characteristic for an unrelated purpose.

9. Third-Party AI-Assisted Interviews and Assessments

Some Talent Cloud workforce workflows may use approved third-party technology providers for interviews, assessments or related analysis.

The provider used depends on the product, programme and customer configuration.

JobMojito

Where JobMojito functionality is enabled, a learner or participant may undertake a video interview through a JobMojito-supported workflow.

JobMojito may process relevant interview information and return outputs such as:

• interview status;
• analysis;
• feedback;
• recruiter-facing analysis;
• interview-related recommendations or observations; and
• other configured interview outputs.

Those outputs may then be stored or presented within the relevant Talent Cloud workflow.

Important Product Distinction

Third-party AI interview analysis should not be assumed to operate throughout every Talent Cloud product.

At the date of this Statement, JobMojito interview AI is used within relevant workforce workflows and is not the mechanism used to determine academy learnership assessment outcomes.

The existence of a technical field, schema capability or potential integration does not mean that an AI process is currently operating in that workflow.

Third-Party Governance

Where a third-party AI or assessment provider is used, that provider's processing may also be subject to its own technology, privacy and information-governance arrangements.

Talent Cloud evaluates the role of the provider within the relevant service and maintains appropriate contractual and governance arrangements where required.

The use of a third-party provider does not mean that Talent Cloud claims to have independently validated every aspect of that provider's model, methodology, accuracy, fairness or performance.

Use of Third-Party Outputs

Third-party AI-assisted outputs are decision-support information.

They should be considered alongside appropriate human judgement and other relevant evidence.

They should not be treated as an unquestionable or complete representation of an individual's ability, potential, suitability or future performance.

10. Other Assessment and Learning Services

Talent Cloud may integrate with approved external assessment, learning, certification or simulation services.

Depending on the programme, these may include services such as:

• Skillable;
• Strengthscope;
• learning platforms using LTI or other approved integration methods; and
• other authorised assessment or certification providers.

An external provider may determine its own assessment methodology and produce an assessment result that is returned to or recorded within Talent Cloud.

Where this occurs, Talent Cloud should not represent an externally determined assessment methodology as though Talent Cloud itself generated the underlying assessment result.

The respective roles of Talent Cloud, the customer and the assessment provider depend on the particular service and processing involved.

11. Programme Reporting and Analytics

Talent Cloud provides programme reporting and analytics.

These may include information concerning:

• enrolment;
• attendance;
• progress;
• assessments;
• readiness;
• evidence;
• employer engagement;
• placements;
• programme activity;
• interventions; and
• outcomes.

Many reports and indicators are generated through database queries, defined business rules and deterministic calculations.

They should not be described as generative AI merely because software calculates, aggregates or presents them.

Where a report incorporates an AI-assisted output from another process, the nature and source of that output should remain appropriately understood.

12. Technology Used for Core Matching

At the date of this Statement, Talent Cloud's current matching architecture uses locally operated semantic embedding technology rather than sending matching profile text to a general-purpose cloud large language model for core semantic matching.

Our current in-scope architecture does not rely on OpenAI, Amazon Bedrock or Anthropic to perform the core Talent Cloud matching calculation.

This describes the current platform architecture. It is not a commitment that the technology stack can never change.

A material change to the technology used for assessment, matching or significant decision-support processing should be subject to appropriate product, privacy and governance review before production use.

Where a material change affects information that individuals should reasonably receive, this Statement or other relevant transparency information should be updated.

13. Personal Information Used in Technology-Assisted Processing

The information used depends on the particular workflow.

Depending on the service, relevant information may include:

• skills;
• qualifications;
• education;
• work experience;
• programme and cohort information;
• availability;
• location;
• employment preferences;
• assessment information;
• readiness information;
• opportunity requirements;
• interview responses; and
• other information relevant to the defined purpose.

Technology-assisted processing should use information relevant to its intended purpose.

The presence of personal information elsewhere in the Talent Cloud platform does not mean that every field is automatically used in every algorithm, score or recommendation.

Further information about the processing of personal information is available in our Privacy Notice.

14. Automated Decision-Making

South Africa's Protection of Personal Information Act regulates certain decisions that produce legal consequences or substantially affect an individual where the decision is based solely on automated processing of personal information intended to provide a profile of that person.

Talent Cloud's current operating model is designed to preserve meaningful human responsibility for material programme, placement and employment decisions.

Technology-assisted assessment, readiness, interview and matching outputs are intended to support human decision-making rather than independently make material decisions solely through automated processing.

The current platform design does not use these processes to automatically:

• reject a learner from a programme;
• remove a learner from a programme;
• change a learner's programme status solely through AI;
• assign a learner to an employer;
• make a final hiring decision;
• make a final employment decision; or
• make a final placement decision.

Whether POPIA section 71 or another legal requirement applies to a particular processing activity depends on the actual workflow, the role of the technology and how an organisation uses its output.

Where an organisation uses Talent Cloud outputs, that organisation remains responsible for the decisions it makes and for using the information lawfully and appropriately.

Nothing in the platform should be interpreted as transferring an employer's, assessor's or programme operator's decision-making responsibility to an algorithm.

15. Human Oversight

The type of human involvement depends on the workflow.

It may include:

• an assessor making an assessment judgement;
• an authorised programme professional reviewing readiness information;
• a programme team reviewing participant progress;
• an employer reviewing a potential match;
• a recruiter reviewing interview information;
• a moderator or quality-assurance professional reviewing evidence;
• an authorised person reviewing an exception or intervention; or
• another appropriately authorised person making the final material decision.

Human oversight should be meaningful where it is required.

It should not consist merely of automatically approving a technology-generated recommendation without appropriate consideration where the resulting decision materially affects an individual.

A person responsible for a material decision should be able to consider relevant context and should not be required to accept an advisory technology output merely because the system produced it.

16. Advisory Outputs Must Not Become Undisclosed Automatic Decisions

Talent Cloud scores, rankings, readiness indicators, matches and AI-assisted outputs may be designed as advisory or decision-support information.

An organisation should not convert such an output into an undisclosed automatic rejection, exclusion or other adverse decision rule without determining that the proposed use is:

• lawful;
• appropriate for the intended purpose;
• sufficiently understood;
• appropriately validated where required;
• consistent with relevant programme and employment requirements; and
• subject to any required human-review, transparency and other safeguards.

For example, the existence of a readiness score should not automatically be interpreted as authority to reject everyone below a particular threshold if that threshold was not designed, validated or lawfully established for that purpose.

The way an organisation uses an output can materially change the legal, fairness and governance implications of the underlying technology.

17. Limitations of Assessments and Technology-Assisted Outputs

Assessments, scores, rankings and AI-assisted outputs have limitations.

They may be affected by matters such as:

• the quality and completeness of information supplied;
• the scope of the assessment;
• the criteria being measured;
• programme configuration;
• the context in which a response was given;
• the assessment provider's methodology;
• weighting choices;
• technical limitations;
• changes in an individual's circumstances;
• errors or incomplete information; and
• the purpose for which an output was originally designed.

For these reasons:

• a readiness indicator is not a guarantee of workplace performance;
• a match score is not a guarantee of suitability;
• an assessment result should be interpreted within its defined assessment context;
• AI-assisted interview analysis should not be treated as infallible; and
• a technology-assisted recommendation should not substitute for appropriate human judgement where human judgement is required.

18. What We Do Not Claim

Talent Cloud does not claim that:

• an assessment, score or AI-assisted output is infallible;
• a match score proves that one individual is objectively better than another;
• a readiness indicator reliably predicts employment success in every context;
• technology can eliminate human bias;
• automation necessarily makes a decision fairer;
• every third-party assessment or AI model has been independently validated by Talent Cloud;
• every technology-assisted workflow is appropriate for every programme, population or purpose;
• an assessment designed for one purpose is automatically suitable for another;
• a technology-assisted output should be relied upon without appropriate context; or
• the use of a Talent Cloud output makes a customer's resulting decision lawful.

These limitations are part of responsible use of the platform.

19. Fairness and Responsible Use

Talent Cloud aims to support fair and responsible use of assessment and technology-assisted processing.

Relevant risks may include:

• inaccurate inputs;
• incomplete information;
• inappropriate weighting;
• unfair or irrelevant criteria;
• historical or contextual bias;
• over-reliance on automated outputs;
• inappropriate use of demographic or sensitive information;
• misunderstanding what a score actually measures;
• using an output outside its intended context; and
• attaching consequences to an output that were not contemplated when it was designed.

Our governance approach is intended to support appropriate review of material assessment, matching and AI changes.

Organisations using Talent Cloud outputs also have responsibilities.

Customers, programme operators, assessors and employers should not use Talent Cloud outputs in a discriminatory, unlawful or materially misleading way.

Technology does not make an otherwise unlawful or unfair decision lawful or fair.

20. Challenging or Correcting an Output

Individuals should have an appropriate route to raise concerns about information or technology-assisted outputs relating to them.

You may contact us if you believe that:

• personal information used in an assessment or matching process is materially inaccurate;
• an assessment outcome has been recorded incorrectly;
• a readiness indicator contains or relies on inaccurate information;
• a matching output appears to rely on incorrect information;
• an AI-assisted interview output is materially inaccurate;
• an output has been attributed to the wrong person;
• a technology-assisted output has been used inappropriately; or
• you require information about an applicable automated process affecting you.

Contact:

[support@talentcloudglobal.com](mailto:support@talentcloudglobal.com)

Depending on the processing involved, Talent Cloud may:

• investigate the platform record;
• correct information we are authorised to correct;
• route the matter to an authorised assessor;
• refer the matter to the organisation responsible for the programme;
• refer an assessment-related question to the relevant assessment provider;
• support an authorised human review; or
• provide appropriate information concerning the relevant technology-assisted process.

Talent Cloud cannot necessarily overturn an assessment, programme, employment or placement decision made by another organisation.

Where another organisation is responsible for that decision, the matter may need to be reviewed or resolved by that organisation.

21. Information About Automated Processing

Where applicable law provides a right to information concerning qualifying automated processing, Talent Cloud will support the exercise of that right according to our legal role.

Where Talent Cloud is the responsible party or controller, we will respond in accordance with applicable requirements.

Where Talent Cloud acts as operator or processor for another organisation, we may refer the request to that organisation or assist it in responding.

Where POPIA section 71 applies to a qualifying decision, applicable safeguards may include an opportunity for the individual to make representations and sufficient information about the underlying logic of the automated processing to enable those representations.

22. Children and Younger Participants

Talent Cloud may support programmes involving participants under the age of 18.

Assessment and technology-assisted processing involving children requires particular care.

Where a programme involving children uses assessments, readiness indicators, matching or another technology-assisted process, the responsible organisation must ensure that:

• the processing has appropriate legal authority;
• the technology is appropriate for the programme, purpose and participant;
• any required safeguards are applied;
• relevant information is presented in an appropriate manner;
• the limitations of the process are understood; and
• material decisions remain subject to appropriate human responsibility.

Talent Cloud does not assume that an assessment or technology-assisted process suitable for an adult programme is automatically appropriate for children.

Additional privacy requirements relating to children's information are explained in our Privacy Notice.

23. Customer, Programme Operator and Employer Responsibilities

Talent Cloud provides technology and decision-support capabilities.

Organisations using those capabilities remain responsible for their own decisions and for using the platform appropriately.

Depending on their role, customers, employers, programme operators and other authorised organisations should:

• ensure that they have an appropriate basis for processing participant information;
• use assessment and matching outputs only for appropriate purposes;
• understand what relevant outputs measure and do not measure;
• provide appropriate human review;
• consider relevant limitations;
• avoid unlawful discrimination;
• maintain appropriate user permissions;
• respond appropriately when an individual challenges information or an outcome;
• avoid converting advisory outputs into inappropriate automatic decision rules; and
• comply with applicable employment, equality, privacy, programme and regulatory requirements.

Talent Cloud does not guarantee that an organisation's use of a platform output will be lawful merely because the output was generated, calculated or displayed by Talent Cloud.

24. Employment and Placement Decisions

Talent Cloud does not employ participants on behalf of employers merely by operating the platform.

Employers remain responsible for their employment decisions.

Programme organisations remain responsible for programme decisions within their authority.

Assessors remain responsible for assessment judgements within their authorised role.

Where matching, readiness, assessment or interview information is provided to an employer or programme organisation, it should be considered as part of an appropriate decision-making process.

Talent Cloud does not guarantee:

• employment;
• placement;
• progression;
• programme completion;
• assessment success;
• employer selection; or
• any other particular outcome.

25. Changes to Technology and AI

Talent Cloud's technology will continue to develop.

New technologies may be introduced and existing technologies may change.

A material technology-assisted processing change may include a significant change to:

• the purpose of a score, model or recommendation;
• categories of information used as inputs;
• weighting methodology;
• decision thresholds;
• eligibility rules;
• an AI or assessment provider;
• a model or significant model version;
• local processing versus external provider processing;
• the use of protected or sensitive characteristics;
• the degree of automation;
• the people or organisations receiving an output;
• the consequences attached to an output;
• the role of human review; or
• the way an output materially affects an individual.

A material technology-assisted processing change should not be treated merely as a software release.

It should trigger appropriate consideration of matters such as:

• privacy;
• transparency;
• fairness;
• security;
• contractual requirements;
• provider governance;
• human oversight; and
• applicable legal or regulatory requirements

before production use.

Where a material change affects information that individuals should reasonably receive, this Statement or other relevant transparency information should be updated.

26. Auditability and Governance

Appropriate auditability helps us understand and investigate significant technology-assisted outputs.

Depending on the workflow and its significance, relevant records may include information concerning:

• the type of processing that produced an output;
• the applicable configuration or version;
• relevant input information;
• the provider or technology involved where applicable;
• the date or time of the processing;
• the relevant programme or context; and
• associated human review or decision information where appropriate.

The precise records maintained depend on the system, purpose and processing involved.

This information is not necessarily all displayed directly to a learner or participant, but it may support governance, investigation, challenge handling and appropriate review.

27. Relationship With the Privacy Notice

This Statement explains Talent Cloud's approach to assessments, readiness, matching and technology-assisted processing.

Our Privacy Notice explains the wider processing of personal information, including:

• categories of information processed;
• purposes of processing;
• responsible party and operator roles;
• special personal information;
• international processing;
• retention;
• security;
• children's information; and
• privacy rights.

The documents should be read together.

28. Changes to This Statement

We may update this Statement when:

• platform functionality changes;
• assessment methodologies change;
• technology providers change;
• matching processes materially change;
• new AI-assisted functionality is introduced;
• the consequences attached to technology-assisted outputs change;
• legal or regulatory requirements change; or
• our governance arrangements develop.

The current version and effective date will be published on the relevant Talent Cloud website or platform.

Where a material change affects individuals, additional transparency or other appropriate measures may also be required.

29. Contact Us

For questions about assessments, readiness, matching, AI-assisted processing or the use of technology within Talent Cloud, contact:

Talent Cloud Global Ltd

Suite A 82 James Carter Road Mildenhall Suffolk IP28 7DE United Kingdom

Email: support@talentcloudglobal.com

Where your question concerns a specific programme, assessment, matching output or employment decision, please identify the relevant programme and, where known, the organisation responsible for it so that we can route the matter appropriately.