Learning, Reskilling and Career Development

Introduction

The future of work is not determined only by automation or artificial intelligence. It is also determined by whether organisations can help people acquire new capabilities, apply them in changing contexts and build credible career pathways. Reskilling refers to learning capabilities for a substantially different role, while upskilling refers to extending capability within an existing role. Both are strategic responses to technological change, demographic shifts, new business models and changing employee expectations.

For human resource professionals, learning is no longer an isolated training activity. It is an integrated system linking business strategy, workforce planning, competency architecture, employee experience, internal mobility, performance and organisational capability. The central managerial question is not simply, “How many employees completed a course?” It is, “Which capabilities must change, for whom, through which learning experiences, and with what measurable effect on work and business outcomes?”

This chapter examines how HR leaders can design reskilling strategies using competency models, personalised learning, learning experience platforms, generative AI, coaching technologies, credentialing and measures of learning transfer and business impact. It also evaluates the ethical and governance issues that arise when AI becomes part of the learning and career system.

Learning objectives

By the end of this chapter, learners should be able to:

1. Distinguish learning, upskilling, reskilling, redeployment and career development.
2. Construct a strategically relevant competency model and identify capability gaps.
3. Design personalised learning journeys supported by learning experience platforms and AI.
4. Evaluate coaching technologies, credentials and internal talent marketplaces.
5. Measure learning transfer, learner equity and business impact.
6. Critically assess ethical, privacy, validity and governance risks in AI-enabled development.
7. Recommend an integrated reskilling strategy for a contemporary organisation.

1. From training activity to capability strategy

Traditional training systems often begin with a catalogue: courses are offered, employees enrol and completion is reported. A capability strategy begins elsewhere—with the organisation’s future value proposition and the work required to deliver it. It asks four linked questions:

• What strategic outcomes must the organisation achieve?
• What work processes and roles will produce those outcomes?
• Which competencies will be required in the future?
• How can current employees move from present capability to required capability?

Learning is a process through which knowledge, skills, attitudes and judgement change. Development is broader: it includes experiences, relationships, reflection and identity formation over time. Career development concerns the individual’s evolving direction, employability and access to opportunities. Reskilling connects these concepts to organisational change by preparing people for new or substantially redesigned work.

A useful distinction is between job-based and skill-based approaches. A job-based system assumes relatively stable occupations and uses fixed job descriptions. A skill-based system identifies capabilities that can be combined across projects, roles and career paths. Skill-based approaches can improve mobility, but they may also oversimplify expertise if skills are treated as isolated commodities rather than socially and contextually embedded practices.

Reskilling should therefore not be interpreted as an obligation placed solely on employees. Employers control access to learning time, technology, work assignments and advancement opportunities. Effective reskilling is a shared employment relationship supported by managers, HR, learning specialists and employees themselves.

2. Theoretical foundations

2.1 Adult learning and experiential learning

Adult learners bring prior experience, prefer relevant problems and are influenced by autonomy, identity and perceived usefulness. Learning is strengthened when employees can connect new ideas with real work rather than merely consume information. Experiential learning perspectives emphasise a cycle involving concrete experience, reflection, conceptualisation and experimentation. In practice, this means that a digital course should be complemented by a work-based challenge, feedback and an opportunity to retry.

2.2 The 70–20–10 idea

The frequently cited 70–20–10 model proposes that development is associated broadly with challenging assignments, relationships and formal learning. It is best treated as a design heuristic, not a universal empirical law. The percentages should not be used as a rigid budget or as evidence that formal education is unimportant. A regulated technical skill, for example, may require substantial formal instruction and assessment before workplace practice is permitted.

2.3 Deliberate practice and feedback

Capability improves when learners perform meaningful tasks at an appropriate level of challenge, receive timely feedback and repeatedly refine performance. AI simulations, role plays, coding environments and coached projects can support this process. However, practice must represent the real work; high scores in a simplified simulation do not automatically establish job competence.

2.4 Career construction and employability

Career development is shaped by both personal agency and structural opportunity. Employees construct career meaning through experiences and transitions, but organisational systems determine who receives sponsorship, developmental assignments and visibility. A responsible career architecture therefore combines self-directed exploration with transparent criteria and equitable access to opportunities.

3. Competency models for reskilling

A competency is an observable and assessable combination of knowledge, skill, behaviour and judgement that contributes to effective performance in a context. Competencies may be:

• Core: expected across the organisation, such as ethical judgement or collaboration.
• Functional: associated with a discipline, such as workforce analytics or procurement.
• Technical: linked to tools and methods, such as data modelling or cloud administration.
• Leadership: concerned with direction, decision-making and enabling others.
• Emerging: capabilities whose importance is increasing, such as AI literacy or responsible automation design.

A competency model should not become an inflated list of desirable traits. It should explain what effective performance looks like, at what level and in which context. Behavioural indicators are more useful than vague labels. For example, “data literacy” may be defined through behaviours such as selecting appropriate measures, identifying data limitations, interpreting uncertainty and communicating findings to non-specialists.

Building a future-oriented competency model

A practical sequence is:

1. Scan strategy and environment. Review business priorities, technology road maps, customer expectations, regulation and labour-market evidence.
2. Decompose future work. Identify tasks, decisions, interfaces and outputs rather than relying only on existing job titles.
3. Cluster capabilities. Group related knowledge, skills and behaviours into coherent competency families.
4. Define proficiency levels. Describe observable performance from foundational to advanced or expert.
5. Validate with stakeholders. Include subject-matter experts, managers, employees, professional bodies and, where relevant, worker representatives.
6. Map roles and adjacent pathways. Identify which existing capabilities transfer to target roles.
7. Specify evidence. Determine how proficiency can be demonstrated through work samples, assessments, credentials or observed performance.

Competencies should be reviewed periodically. A model that is too stable becomes obsolete; one that changes continuously becomes unusable and undermines trust.

Gap analysis

A simple capability-gap representation is:

where G_i is the gap for competency i, R_i is the required proficiency and C_i is current demonstrated proficiency. The expression assumes that both values use the same valid scale and that a higher score means greater proficiency. It is a diagnostic device, not a precise measurement of human potential. A positive gap indicates development need; a zero or negative value does not prove that performance is effective in every context.

For prioritisation, organisations may consider strategic importance, gap size, scarcity and feasibility. A simple weighted priority score can be used only when its scales are explicitly defined:

Here, P_i is the priority score, S_i strategic importance, G_i gap magnitude, C_i external scarcity and F_i feasibility; the w terms are non-negative weights that sum to 1. Scores are ordinal or scaled indices, not objective truths. The model supports discussion and transparent assumptions rather than replacing managerial judgement.

4. Personalised learning architecture

Personalisation means adapting learning to an employee’s role, prior capability, goals, pace, preferences and work context. It does not mean allowing an algorithm to determine a person’s career without human involvement.

A robust learning journey generally includes:

1. Diagnosis: self-assessment, manager assessment, work samples, validated tests and evidence from performance.
2. Goal setting: a target capability and a meaningful work outcome.
3. Sequencing: foundational concepts before advanced application, with branching for prior knowledge.
4. Practice: realistic tasks, simulations, projects and deliberate repetition.
5. Social learning: peers, communities of practice, mentors and subject experts.
6. Feedback and reflection: formative feedback during learning and structured reflection afterward.
7. Demonstration: evidence that capability can be applied in work.
8. Mobility: access to assignments, vacancies or projects where the new capability is used.

Personalisation can be rule-based, such as assigning introductory content after a diagnostic test, or adaptive, where the system continuously adjusts recommendations. It can also be human-led, with a manager or coach interpreting evidence and negotiating a development plan. The strongest approach is usually blended.

Learning experience platforms

A learning management system primarily administers assigned learning, records completion and supports compliance. A learning experience platform, or LXP, generally aggregates learning resources from multiple sources and supports discovery, recommendation, social interaction, user-generated content and pathways. The distinction is not absolute because products increasingly overlap.

An LXP can improve access by connecting employees to internal experts, external courses, communities and project opportunities. Its value depends on content quality, search relevance, accessibility, integration with talent systems and the credibility of recommendations. Excessive choice can create decision fatigue. Recommendation engines may also reproduce popularity bias, favouring highly visible content or users who already possess digital confidence.

5. Generative AI in learning and development

Generative AI can support learning by:

• producing explanations at different levels of difficulty;
• generating practice questions, scenarios and role plays;
• simulating customer, employee or managerial conversations;
• providing first-draft feedback on writing or code;
• translating or transforming content for accessibility;
• summarising resources and creating study plans;
• assisting learning designers with outlines and alternative examples.

These uses require safeguards. Models can generate inaccurate, stereotyped or fabricated content. They may expose confidential information if prompts are sent to an unapproved service. They can create an illusion of personalisation while relying on weak inferences. AI feedback may also be inappropriate for sensitive issues such as mental health, disciplinary matters or protected disclosures.

A responsible design process includes approved tools, data minimisation, human review, source verification, user disclosure, appeal routes and monitoring for differential effects. Employees should know when they are interacting with an AI coach and what data are retained. AI should augment expert judgement, not silently determine access to training, promotion or redundancy.

6. Coaching technologies and career support

Digital coaching ranges from scheduling and goal tracking to conversational systems that prompt reflection. Human coaching offers contextual judgement, confidentiality and relational trust; AI-enabled coaching offers scale, availability and consistent prompts. Neither is universally superior.

An appropriate division of labour is to use technology for low-risk reflection, reminders, resource discovery and practice, while escalating complex, sensitive or high-consequence matters to a qualified human. Coaching systems should not be presented as therapists or as substitutes for employee assistance services.

Career technologies can display adjacent roles, capability requirements, learning pathways, mentors and internal vacancies. Their design should avoid turning career movement into a purely algorithmic matching problem. A person’s aspirations, caring responsibilities, health, location, identity and informal experience may not be represented accurately in organisational data. Employees should be able to correct records and explain circumstances that an algorithm cannot observe.

7. Credentialing and evidence of capability

Credentials signal that a learner has met defined requirements. They include degrees, professional certifications, licences, certificates of completion, digital badges and microcredentials. Their value depends on issuer credibility, assessment rigour, transparency, currency and labour-market recognition.

A completion certificate shows participation; it does not necessarily show competence. Stronger credentials specify learning outcomes, assessment methods, issuer, date, evidence and renewal requirements. A digital badge should contain verifiable metadata rather than merely a decorative image.

Organisations should distinguish between:

• Knowledge evidence: tests, explanations or written work.
• Skill evidence: demonstrations, simulations or work samples.
• Performance evidence: sustained application in real work.
• Social evidence: validated feedback from clients, peers or supervisors.

Credential inflation is a risk. If every short activity creates a badge, employees and managers may struggle to identify meaningful signals. Credential pathways should therefore be connected to role requirements and internal mobility, not only to platform engagement.

8. Measuring learning transfer and business impact

Evaluation should begin during design, not after delivery. A useful logic chain is:

Inputs → learning experience → learning outcomes → transfer to work → operational results → strategic value.

Completion, attendance and satisfaction are process indicators. They can reveal reach and usability but do not establish learning or impact. More informative measures include assessment performance, work-sample quality, observed behaviour, time to proficiency, internal movement, error rates, productivity, customer outcomes and retention.

Learning transfer

Transfer is the sustained application of learning in the job. A transfer evaluation should ask whether employees had opportunity, resources, managerial support and psychological safety to apply the new capability. A capable employee may fail to transfer learning because systems, incentives or workflow design prevent application.

A comparison design strengthens inference. For example, an organisation may compare change in an outcome for participants with change for a similar non-participant group:

Here, Y_1,T and Y_0,T are post- and pre-intervention outcomes for the treatment group, while Y_1,C and Y_0,C are corresponding outcomes for the comparison group. This difference-in-differences estimate assumes that, absent the intervention, both groups would have followed reasonably similar trends. That assumption should be examined rather than assumed. Other changes, selection effects and measurement differences can still confound the estimate.

Return on investment

When credible financial estimates exist, training ROI may be represented as:

where B is monetised benefit and C is total relevant cost. Costs may include design, delivery, technology, learner time, travel, coaching and evaluation. Benefits may include reduced errors, reduced time to proficiency or increased revenue, but attribution must be defensible. If benefits cannot be monetised reliably, a balanced outcome dashboard is preferable to false precision.

A mature dashboard combines leading and lagging indicators:

Evaluation should protect privacy and avoid using learning data as a hidden surveillance system. Data collected for development should not automatically be repurposed for disciplinary decisions.

9. Integrated reskilling strategy

An organisation-wide strategy can be organised into seven stages:

1. Strategic diagnosis: identify future business scenarios and workforce implications.
2. Capability architecture: define competency families, proficiency levels and evidence.
3. Portfolio prioritisation: select critical capabilities and target populations.
4. Learning ecosystem design: combine formal learning, practice, coaching, projects and credentials.
5. Opportunity connection: link learning to internal gigs, roles, apprenticeships and succession pathways.
6. Manager enablement: train managers to create practice opportunities and give feedback.
7. Evaluation and renewal: measure transfer, equity and business outcomes; revise the portfolio.

A reskilling programme fails when learning is disconnected from available work. Employees will reasonably question a programme that awards credentials but offers no opportunity to use them. Consequently, workforce planning, job design, recruitment and learning must operate as a connected system.

Applied case: Northstar Services’ AI-enabled reskilling programme

This is an illustrative case, created for educational purposes.

Background

Northstar Services, a 4,000-person customer operations company, planned to introduce conversational AI for routine enquiries. Leaders expected fewer repetitive transactions but more complex escalation, quality assurance and AI-supervision work. The initial proposal was to train all service representatives in a generic “AI fundamentals” course.

Problem

Employee representatives argued that the proposal treated automation as a technology project rather than a career transition. Managers also lacked agreement about the capabilities required in redesigned roles. Completion data from previous programmes were high, but customer outcomes had not improved consistently.

Analysis

A cross-functional team mapped future tasks rather than simply deleting existing jobs. It identified four target capability families: complex problem diagnosis, digital workflow management, AI-output checking and empathetic communication. Current proficiency was assessed through work samples, structured observation and employee self-assessment. The team found that many employees already had strong customer judgement but lacked confidence with data and workflow tools.

The company created three pathways: advanced service specialist, AI quality reviewer and workflow analyst. Each pathway included foundational digital learning, realistic simulations, peer learning and a supervised project. Employees received an individual diagnostic report but could challenge the result. Managers were required to allocate practice time and document evidence of application.

An LXP aggregated internal guides, vendor resources and employee-created examples. A generative AI tutor provided low-risk practice scenarios, but answers were checked against an approved knowledge base. Sensitive employee questions were routed to human coaches. Digital badges were issued only after an assessment and work-based demonstration.

Evaluation

Northstar measured assessment performance, time to independent handling, quality scores, internal movement and employee perceptions of fairness. It compared participating teams with delayed-rollout teams while checking whether their pre-programme trends were similar. The evaluation also examined participation and outcomes across shifts, locations and demographic groups.

Outcome

The programme produced stronger evidence of capability than the original completion-based plan. Some employees moved into redesigned roles, while others chose advanced customer work. The organisation also discovered that workflow bottlenecks—not lack of skill alone—limited transfer. It therefore changed approval rules and redesigned escalation procedures.

Learning points

The case illustrates five principles: begin with future work, recognise transferable capability, connect learning with real opportunities, retain human oversight of AI and evaluate system conditions as well as individual learning. It also shows why reskilling is an organisational redesign issue, not merely a course-delivery issue.

10. Challenges, ethics and governance

Equity and accessibility

Digital learning may exclude employees with limited connectivity, disabilities, language barriers, shift constraints or lower digital confidence. Accessible design, paid learning time, multiple formats and assistive technology are core requirements rather than optional benefits.

Algorithmic bias

Recommendation, assessment and career-matching systems can reproduce historical inequality. Bias may arise from training data, proxy variables, incomplete skill records or feedback loops. Organisations should test outcomes across relevant groups, document model purposes and provide meaningful human review.

Privacy and data protection

Learning platforms may collect sensitive information about performance, aspirations, weaknesses and behaviour. Data collection should be necessary, proportionate, secure and transparent. Retention periods, access controls, vendor responsibilities and employee rights should be specified. Legal requirements vary by jurisdiction; organisations must consult applicable employment, privacy, accessibility and AI regulation.

Validity and trust

A recommendation is not valid merely because it is generated by an advanced model. Assessment validity requires evidence that the instrument measures the intended capability and relates meaningfully to performance. Employees should understand how decisions are made and have a route to correction or appeal.

The learning burden

Reskilling can become an additional workload imposed on already pressured employees. Organisations should provide protected time, realistic sequencing and recognition of learning effort. If participation is formally voluntary but informally required for career survival, that contradiction should be addressed openly.

Best-practice principles

• Start with strategic work and capability evidence, not fashionable tools.
• Use competency models as living architectures, not static dictionaries.
• Combine diagnosis, practice, feedback, social learning and real assignments.
• Link credentials to validated evidence and actual mobility.
• Treat AI as an assistive technology subject to human accountability.
• Measure transfer and business outcomes, while testing assumptions about causality.
• Provide accessibility, transparency, privacy and appeal mechanisms.
• Evaluate who benefits, who is left behind and what organisational barriers prevent application.
• Give managers responsibility for creating conditions in which learning can be used.

Chapter Summary

Learning, reskilling and career development form an integrated capability system. Effective reskilling begins with future work and translates strategy into competency families, proficiency levels and credible evidence. Gap analysis can prioritise capability needs, but scores must be interpreted cautiously and validated with stakeholders.

Personalised learning is strongest when diagnosis, sequencing, practice, feedback, social learning, demonstration and mobility are connected. LXPs can broaden discovery and access, while generative AI can support explanation, simulation and feedback under conditions of human review, data protection and content verification. Coaching technologies should extend—not replace—human judgement in complex or sensitive situations. Credentials are valuable when they represent transparent, assessed and current evidence of capability.

Evaluation must move beyond completion and satisfaction to learning, transfer, workforce outcomes and business impact. Comparison designs and ROI calculations can support inference when assumptions and costs are explicit. Finally, AI-enabled development requires governance for fairness, accessibility, privacy, validity, transparency and employee agency. The strategic test of a reskilling programme is not whether people consumed learning, but whether they gained equitable access to meaningful work and whether the organisation improved its capacity to adapt.

Key Terms

Competency: An observable combination of knowledge, skill, behaviour and judgement that supports effective performance.

Competency model: A structured description of capabilities and proficiency levels required for roles or strategic work.

Upskilling: Developing additional capability within an existing role or occupational area.

Reskilling: Developing capability for a substantially different role or form of work.

Learning transfer: Application and maintenance of learning in the workplace.

Career development: The ongoing process through which people build capabilities, direction and employability.

Learning experience platform: A system that aggregates learning resources and supports discovery, recommendation and social learning.

Personalised learning: Learning adapted to an individual’s prior capability, goals, context and pace.

Generative AI: AI that produces new text, images, code, audio or other content from learned patterns.

AI literacy: The ability to understand, use, evaluate and govern AI appropriately.

Coaching technology: Digital tools that support reflection, goals, practice, feedback or coaching interactions.

Microcredential: A focused credential representing assessed learning in a defined area.

Digital badge: A digitally verifiable representation of an achievement, ideally containing evidence metadata.

Internal mobility: Movement of employees across roles, projects, teams or career pathways within an organisation.

Capability gap: The difference between required and currently demonstrated proficiency.

Work sample: A task or artefact used to demonstrate practical capability.

Learning transfer climate: The extent to which workplace conditions support application of learning.

Difference-in-differences: An evaluation approach comparing changes over time between an intervention and comparison group.

Return on investment: Monetised benefit relative to relevant cost, expressed as a percentage when multiplied by 100.

Algorithmic bias: Systematic and unfair differences in outputs or outcomes associated with data, design or implementation.

Review Questions

Short Answer Questions

1. Distinguish between upskilling and reskilling.
2. What are the main stages in constructing a future-oriented competency model?
3. Why is course completion an insufficient measure of learning success?
4. Identify two appropriate uses and two risks of generative AI in learning.
5. What does learning transfer mean?

Long Answer Questions

1. Explain how competency models can connect business strategy, reskilling and internal mobility.
2. Critically evaluate the role of learning experience platforms in personalised learning.
3. Discuss the benefits and limitations of digital credentials and microcredentials.
4. Explain how an organisation can evaluate learning transfer and business impact.
5. Develop a governance framework for AI-enabled coaching and career development.

Analytical / Application Questions

1. An organisation identifies a required proficiency of 4 and a current demonstrated proficiency of 2 on a five-point competency scale. Interpret the gap and explain why the result should not be treated as a complete reskilling decision.
2. Design a reskilling pathway for employees whose routine administrative work is being automated. Specify competencies, learning experiences, evidence of capability and mobility opportunities.
3. A learning programme shows high completion but no improvement in customer outcomes. Diagnose at least four possible explanations and propose an evaluation design to distinguish among them.

Multiple Choice Questions with Answers

Question 1

Which statement best distinguishes reskilling from upskilling?

A. Reskilling develops capability for substantially different work, while upskilling extends capability in an existing role.

B. Reskilling is informal learning, while upskilling is always formal learning.

C. Reskilling applies only to managers, while upskilling applies only to specialists.

D. Reskilling measures performance, while upskilling measures satisfaction.

Correct Answer: A. Reskilling develops capability for substantially different work, while upskilling extends capability in an existing role.

Question 2

What is the primary purpose of a future-oriented competency model?

A. To replace all job descriptions with personality tests.

B. To connect strategic future work with observable capabilities and proficiency levels.

C. To rank employees permanently from highest to lowest potential.

D. To measure course attendance across departments.

Correct Answer: B. To connect strategic future work with observable capabilities and proficiency levels.

Question 3

Which activity provides the strongest evidence of learning transfer?

A. Downloading a course workbook.

B. Receiving a completion certificate.

C. Applying the capability successfully in a relevant work task.

D. Reporting that the course was enjoyable.

Correct Answer: C. Applying the capability successfully in a relevant work task.

Question 4

Which is an appropriate safeguard for generative AI used as a learning tutor?

A. Allowing the tool to make promotion decisions automatically.

B. Using confidential employee records in unrestricted prompts.

C. Removing all human review to increase efficiency.

D. Verifying content and escalating sensitive matters to qualified humans.

Correct Answer: D. Verifying content and escalating sensitive matters to qualified humans.

Question 5

What does an LXP generally add beyond basic learning administration?

A. Discovery, aggregation, recommendation and social learning features.

B. Guaranteed validity of every external course.

C. Automatic replacement of all human instructors.

D. A legal determination of employee competence.

Correct Answer: A. Discovery, aggregation, recommendation and social learning features.

Question 6

A competency requires proficiency 5 and current proficiency is 3 on the same scale. Using G_i = R_i − C_i, what is the gap?

A. −2

B. 0

C. 2

D. 8

Correct Answer: C. 2

Question 7

Which credential provides the strongest evidence of practical capability?

A. A record showing that an email announcement was opened.

B. A badge issued after a work-based assessment and demonstration.

C. An unverified self-rating with no supporting evidence.

D. A certificate issued solely for registering.

Correct Answer: B. A badge issued after a work-based assessment and demonstration.

Question 8

What assumption is especially important in a difference-in-differences evaluation?

A. Participants must all receive identical salaries.

B. The intervention must have no measurable outcome.

C. Treatment and comparison groups should have reasonably similar pre-intervention trends.

D. All employees must complete the programme simultaneously.

Correct Answer: C. Treatment and comparison groups should have reasonably similar pre-intervention trends.

Question 9

Which measure is primarily a business-impact indicator?

A. Number of learning-page views.

B. Learner satisfaction with video quality.

C. Course enrolment rate.

D. Reduction in verified service errors after capability application.

Correct Answer: D. Reduction in verified service errors after capability application.

Question 10

Which governance principle is most appropriate for AI-based career recommendations?

A. Employees should have transparency, opportunities to correct data and human review.

B. Recommendations should be secret so users cannot influence them.

C. Historical promotion patterns should always be reproduced.

D. The algorithm should decide career outcomes without appeal.

Correct Answer: A. Employees should have transparency, opportunities to correct data and human review.