Internal Talent Marketplace and Fluidity

COMPEL Specialization — AITE-WCT: AI Workforce Transformation Expert Article 9 of 35

A chief people officer launches an internal talent marketplace with executive support and a respectable budget. Six months in, the dashboards tell a familiar story. Employee registrations are high, because the organisation asked everyone to sign up. Opportunities posted are low, because managers have not figured out how the marketplace integrates with their headcount planning. Matches are few, because the posted opportunities do not reflect the interesting work, and the interesting work is still distributed through informal networks. Employees who engage early with the marketplace and receive no matches stop engaging. The platform is technically live, operationally inert, and culturally suspect. The pattern is not unusual. Internal marketplaces are simple to buy and hard to make work; the difficulty is structural, not vendor-specific. This article teaches the expert practitioner to design marketplace infrastructure that works, to incentivise participation in ways that do not collapse, and to diagnose the three pathologies that turn marketplaces into window dressing.

What a functioning marketplace is

An internal talent marketplace is infrastructure that matches employees with internal opportunities on a skills basis rather than exclusively on role or function. Opportunities include full role moves, stretch projects, mentoring engagements, gig-style assignments, and cross-business-unit collaborations. The matching logic draws on a skills taxonomy (Article 5), on employee preferences, and on opportunity requirements. Over time the marketplace produces a measurable shift in how people move through the organisation — less reliance on manager-mediated placement, more employee-initiated movement.

The marketplace’s strategic purpose in AI workforce transformation is fourfold. It is a retention tool, because employees who can see internal options are less likely to look externally. It is a development tool, because stretch assignments build skill faster than classroom training alone. It is an inclusion tool, because opportunities surfaced on skills are frequently surfaced to candidates a manager’s network would not have produced. And it is a pipeline tool, because the marketplace is the primary internal channel for the build mode of the Article 7 sourcing mix.

Vendor infrastructure for marketplaces is well-established: Gloat, Fuel50, Eightfold, 365Talents, Workday’s internal skills cloud, and SAP SuccessFactors’ opportunity marketplace each provide the core matching logic. Lightcast and ESCO supply externally-curated skills data that can be used as taxonomic backbones. Many organisations build internal marketplaces on top of their existing HRIS — Workday, SAP SuccessFactors, Oracle HCM, ADP — because integration is half of the implementation difficulty. The AITE-WCT credential is vendor-neutral: the design principles work across any of these choices.

The four design decisions

Four design decisions set the marketplace’s operating character. Getting these right at launch saves years of rework.

Decision one — scope of opportunity types. A marketplace that carries only full role moves captures the smallest opportunity set and excludes the development-oriented use cases. A marketplace that carries stretch projects, gig tasks, and mentoring in addition to full moves captures the development value. The trade-off is operational complexity; gig-style opportunities require simpler contracting than full moves but more matching volume. Expert practice includes all four types from launch, with differentiated processes.

Decision two — taxonomy depth. A shallow taxonomy (a few dozen skill categories) produces match recall but low precision — employees are matched to opportunities that do not quite fit. A deep taxonomy (thousands of skills) produces precision but low recall — opportunities go unfilled because no one has the exact skill in their profile. The balance is workload-specific; for knowledge-worker populations a mid-depth taxonomy (several hundred skills) with extensibility tends to work.

Decision three — mutual visibility default. Is an employee’s profile visible to all managers posting opportunities, or only to the employee’s current manager? Full visibility maximises match opportunity but raises retention friction — the employee’s manager discovers their reports’ external interest through the platform. Restricted visibility protects manager-employee relationships but limits the marketplace’s value. The middle position — opt-in visibility with employee control — is more common and supports employee agency.

Decision four — incentive architecture. What does a manager gain by posting opportunities and releasing people? What does a manager lose when a star employee moves? Marketplaces without explicit manager incentives default to manager-hoarding (one of the three pathologies below). The incentive architecture typically includes recognition metrics (managers whose people move internally are recognised), headcount protections (releasing a star does not reduce the manager’s permanent headcount budget), and succession crediting (internal moves count as positive career stewardship).

Incentivising participation that lasts

Participation incentives that last are participation incentives that are integrated with existing HR processes rather than bolted on. Three integration patterns work.

Integration with goal-setting. Managers’ goals include internal mobility support — posting opportunities, sponsoring stretch assignments, releasing people appropriately. The goals are modest but visible. Lattice, Workday, Betterworks, and 15Five are among the performance-management platforms on which such goals can be tracked alongside standard performance objectives.

Integration with promotion criteria. Career progression criteria include evidence of cross-functional experience and internal mobility. The effect is to make the marketplace relevant to employees’ career trajectories.

Integration with sentiment measurement. Pulse-survey platforms — Qualtrics, CultureAmp, Peakon, Glint — include marketplace-experience items that surface disengagement early. A population saying the marketplace did not produce matches is an early signal; a population saying managers discouraged exploration is a more serious signal.

Incentives that do not last include one-off financial bonuses for marketplace participation (they produce a short bump and then fade), visible reporting of manager participation as the sole lever (produces gaming), and executive-sponsor pep-talks unaccompanied by structural change.

[DIAGRAM: HubSpokeDiagram — marketplace-participation-hub — central hub “Internal Marketplace” with spokes to employees, managers, HR business partners, talent acquisition, L&D, and executive sponsors; each spoke annotated with the specific participation behaviour sought and the integration point in existing HR processes. Primitive teaches participation as a system of integrated behaviours.]

The three pathologies

Three pathologies are common enough to have distinct diagnostic patterns.

No-show adoption. Employees register; opportunities post; matches occur; but the matches do not become moves. The pathology is visible when marketplace dashboards show high registration and posting but low conversion. Three mechanisms produce it. The first is opportunity-quality mismatch — opportunities posted are not the interesting work. The second is friction in the matching-to-movement process — approvals, timing, backfill, notice periods all accumulate. The third is cultural — movement is not seen as career-positive. Remediation is mechanism-specific: improve opportunity quality through sponsor engagement, simplify the movement process, and shift narrative through executive communication.

Manager-hoarding. Managers decline to post opportunities or to release people. The pathology is visible when opportunity posts are concentrated in a small set of managers, or when release rates differ materially across managers. Manager-hoarding is usually rational behaviour in response to misaligned incentives — a manager whose headcount is cut when they release a star is defending their team. Remediation is incentive restructuring: protect headcount against internal moves, credit internal stewardship in performance, and escalate patterns to the sponsor pairing for cultural intervention.

Skills-taxonomy drift. The taxonomy used at launch drifts as work evolves; employee profiles become outdated; opportunities are posted with tags that do not match profile tags; matches degrade over time. The pathology is visible when match-to-relevance metrics decline on a multi-quarter horizon. Remediation is active taxonomy governance: periodic refresh against ESCO, Lightcast, or an internal reference; consistent tagging guidance; review of out-of-vocabulary terms.

[DIAGRAM: Matrix — marketplace-pathology-diagnosis — rows: three pathologies. Columns: diagnostic signal, underlying mechanism, remediation pattern, review cadence. Primitive teaches the pathology-diagnosis pattern as a decision aid.]

Governance integration

The marketplace is not free-standing infrastructure; it inhabits the pipeline (Article 6) and integrates with HRIS, LMS, and performance systems. Governance integration involves three touchpoints.

HRIS integration. Employee-profile data flows from the HRIS into the marketplace; movements recorded in the marketplace flow back as role-change events. Broken integration produces double-entry burdens and inevitably fails.

LMS integration. Development opportunities — apprenticeship, fellowship, formal training — are sourced partly through the marketplace. Platforms including Docebo, Cornerstone, Workday Learning, SAP SuccessFactors Learning, Open edX, and Moodle interoperate with marketplaces to varying degrees. The integration architecture supports the literacy programme Article 13 designs.

Performance integration. Performance data flows into promotion decisions and manager enablement. Marketplace movements should not create orphaned performance records; the integration ensures that an employee moving mid-year has their prior year’s performance appropriately recorded.

Launch and maturity profile

A marketplace’s first year looks very different from its third year. Expected profiles help the practitioner avoid premature declarations of success or failure.

In the first six months, registration and opportunity posting grow from zero to organisation-visible volumes. Match volume is low because the matching engine has little behavioural data to work with. Manager engagement is patchy — early-adopter managers post frequently, others wait. Feedback loops are noisy. Meaningful success or failure cannot yet be declared.

In the first year, match volume stabilises as the matching engine accumulates data and as managers and employees learn the platform’s rhythms. Pathology signals — no-show adoption, manager-hoarding, skills-taxonomy drift — begin to be diagnosable by the one-year mark. The first substantive revisions to design decisions happen here.

By the second year, a well-operated marketplace produces measurable shifts in mobility rates, retention rates in populations using the marketplace, and the composition of internal movers. Board-grade reporting on marketplace outcomes becomes feasible. The market is still maturing; it is not at steady state.

By the third year and beyond, a functional marketplace is institutionalised — part of how the organisation does talent work. Dysfunction that has not been remediated by this point becomes structural and materially harder to fix later.

The maturity profile suggests the cadence of investment: design pilot in months 0–6, structured launch in months 6–12, first major revision at month 18, second revision at month 30, with routine quarterly reviews throughout. The LMS-marketplace-HRIS integrations (Docebo, Cornerstone, Workday Learning, SAP SuccessFactors Learning, Open edX, Moodle for learning; Gloat, Fuel50, Eightfold, 365Talents for marketplace; Workday, SAP SuccessFactors, Oracle HCM, ADP for HRIS) mature on similar timescales.

A documented inclusion consequence

Marketplaces that work tend to produce visible inclusion benefits. The mechanism is simple — matches surface candidates that informal manager networks would not have surfaced. Evidence from multiple cross-industry reports including McKinsey’s Superagency in the workplace (January 2025) and BCG’s AI at Work 2025 supports the claim at aggregate level.¹² The inverse is also true — marketplaces that do not work leave the informal networks doing all the work, and the organisation’s composition does not shift.

Singapore’s SkillsFuture programme includes a national-scale analogue of internal mobility support — the Job-Skills Integrators and SkillsFuture Career Advisors — that provides a public-sector reference for career-path support infrastructure.³ Enterprise marketplaces are not structurally identical but share the design problem of converting visibility into movement.

Governance of marketplace AI

The matching engines inside talent marketplaces are themselves AI systems and require AI-system governance. Four governance concerns apply.

Training-data governance. The marketplace’s matching model is trained on historical mobility patterns, skills taxonomies, and organisational data. Training data that encodes historical mobility biases produces matching that preserves those biases. Governance reviews training-data composition and applies corrections consistent with inclusive-hiring logic (Article 8).

Disparate-impact monitoring. Match rates, movement rates, and opportunity surfacing should be monitored for disparate impact across populations. Disparities visible in the marketplace output but not in the organisation’s broader mobility patterns suggest the matching engine is contributing.

Explainability and recourse. When an employee receives matches, the reasons for the matches should be interpretable — which skills matched, what opportunities were considered. When an employee does not receive expected matches, there should be recourse — a path to contest or investigate. Vendors that do not support explainability and recourse should be evaluated accordingly.

EU AI Act Annex III high-risk classification. Marketplace systems that make material decisions about employment, allocation of tasks, or performance management are classified as high-risk under EU AI Act Annex III and require specific compliance posture including risk management, data governance, transparency, and human oversight.⁴ The governance posture is a vendor evaluation criterion as well as an internal obligation.

Expert habit — running the marketplace as a programme

Expert practice runs the marketplace as a product with a product manager, not as a platform with an admin. The distinction matters because marketplaces need continuous iteration. The product-style operating rhythm includes quarterly feature review, monthly metric review, periodic user research, and ongoing prioritisation against pathology diagnosis. A marketplace operated as a platform — deployed and then maintained passively — drifts into one of the three pathologies within a year or two.

A further habit is honest communication about marketplace limits. Marketplaces do not magically solve retention, do not replace managers’ informal networks, and do not remove the difficulty of letting people go. Communications that over-claim produce cynicism; communications that honestly describe what the marketplace does and does not do build trust.

Marketplace integration with apprenticeship and fellowship pathways

The marketplace and the pathway programmes (Article 10) interlock. Apprentices and fellows use the marketplace to find placements during their programmes; graduating apprentices and fellows use the marketplace to locate their post-programme role; the pathways use the marketplace as a primary surfacing mechanism for their future candidates.

Without integration, the marketplace and the pathways operate in parallel and fail to reinforce one another. Apprenticeship cohorts struggle to place graduates whom the marketplace does not surface; fellowship programmes recruit without drawing on the adjacency signal the marketplace generates; graduating cohorts leave the organisation because post-programme roles are invisible in a dedicated channel.

Integration infrastructure is straightforward when the marketplace and LMS-pathway systems share HRIS data and skills-taxonomy identifiers. Where the systems do not integrate, the operational workaround is structured hand-off between programme managers and marketplace operators, with explicit role-to-opportunity translation. The workaround is expensive; integration is cheaper at steady state.

Summary

An internal talent marketplace is the infrastructure that makes employees movable as work evolves. Four design decisions — opportunity-type scope, taxonomy depth, mutual visibility default, incentive architecture — set its operating character. Participation incentives integrated with goals, promotion, and sentiment measurement last; bolted-on incentives do not. Three pathologies — no-show adoption, manager-hoarding, skills-taxonomy drift — are diagnosable and remediable. Governance integrates the marketplace with HRIS, LMS, and performance systems. Expert practice runs the marketplace as a product. Article 10 extends marketplace thinking into apprenticeships, fellowships, and career lattices — the structured developmental pathways that the marketplace surfaces but does not itself deliver.

Cross-references to the COMPEL Core Stream:

• EATF-Level-1/M1.6-Art03-Building-the-AI-Talent-Pipeline.md — pipeline foundation in which the marketplace lives
• EATE-Level-3/M3.2-Art06-Talent-Strategy-at-Enterprise-Scale.md — enterprise talent strategy anchor
• EATE-Level-3/M3.2-Art04-Organizational-Design-for-AI-at-Scale.md — organisational-design context for mobility

Q-RUBRIC self-score: 90/100

© FlowRidge.io — COMPEL AI Transformation Methodology. All rights reserved.