What problem does it solve?
In many large organizations employees find internal opportunities through their network and their manager, if at all. Open roles and short projects are advertised unevenly, skills are recorded in job titles rather than in anything searchable, and managers who need help quickly hire a contractor because they cannot see who inside the company has the skill and the time. People who want to grow leave to do it elsewhere: Schneider Electric's internal surveys, as reported by its platform vendor Gloat, found that nearly half of departing employees cited a lack of internal growth opportunities as their main reason.
At the same time, skills needs change faster than job architectures. HR teams want to know which skills they have, where the gaps are and how to redeploy people, but a skills inventory built by hand soon falls behind.
How does it work?
- Build a skills profile. The employee imports a CV or professional profile; AI extracts and infers skills from it and from HR data, and the employee confirms, adds aspirations and availability.
- Post opportunities. Managers post projects, gigs, open roles and mentoring offers with the skills needed and the time involved.
- Match both ways. The platform recommends opportunities, mentors and learning to each employee, and suggests candidates to the manager, ranked by skills fit and stated interest.
- Apply and agree. Employees apply, managers choose, and the employee's own manager agrees the time commitment; the platform records the assignment.
- Learn from the data. Completed assignments update the skills profile, and aggregated skills data shows HR where the gaps are for learning and hiring plans, as Mastercard describes.
- Audience
- Employee facing
- Autonomy
- Assist
- Adoption
- Early adopters
- Channels
- Internal tools
What is it worth?
Benchmarks are computed from the public deployments below: one data point per organization per KPI, with who made each claim.
| KPI | Median | Reported range | Data points | Claimed by |
|---|---|---|---|---|
| Cost savings | Not pooled | at least USD 15 million | 1 | 1 vendor |
| Users served | Not pooled | at least 2300 | 1 | 1 vendor |
Value drivers: Employee productivity, Lower cost to serve, Speed and cycle time, Inclusion and access.
Indicative value
A company with 20,000 employees
USD 96,000 to USD 960,000
External contractor and hiring cost avoided per year
How this is calculated
Formula: employees * gigShare * hoursPerGig * externalShare * contractorRate. The low scenario uses every low input, the high scenario every high input.
| Input | Low | High | Basis |
|---|---|---|---|
| Employees with access to the marketplace employees, employees | 20,000 | 20,000 | The reference company. |
| Share of employees who complete an internal project or gig in a year gigShare, fraction of employees | 0.01 | 0.02 | Editorial assumption, replace with your own participation data. None of the evidence records gives an annual participation rate. |
| Hours of work per project or gig hoursPerGig, hours per gig | 40 | 60 | Editorial assumption, replace with your own data. With these values the company logs 0.4 to 1.2 hours of internal gig work per employee a year. The evidence gives no annual rate and does not support one: Schneider Electric's hours are cumulative since its April 2020 launch with no end date, and Gloat's story is inconsistent (360,000 hours in its text, 550,000 in its header), which is 2.3 to 3.5 hours per employee in total for a workforce of about 155,000, so fewer per year. |
| Share of those hours that would otherwise have been bought from contractors or new hires externalShare, fraction of hours | 0.2 | 0.4 | Editorial assumption; much internal gig work would otherwise not be done at all. |
| Cost of an external contractor hour contractorRate, USD per hour | 60 | 100 | Editorial assumption. |
What it leaves out: Counts only avoided external spend on gig work. It leaves out the effect on retention and hiring for open roles, the value of work that would otherwise not be done, the time employees spend away from their main job, and the platform and change management costs.
Who already uses it?
4 public deployments, strongest evidence first. Grades: A regulator or audit, B the organization itself, C vendor case study, D anonymous or estimate.
Mastercard
United States · Payments and cards · 2024
Mastercard uses AI in Unlocked, its internal talent marketplace, to match employees to short term projects, volunteering, open roles, mentors and learning pathways, based on the skills they have and the skills they want to build. The company says 90% of its workforce is on the platform, with 500,000 project hours delivered, and that the skills data shows where it has gaps so it can plan learning paths or hiring. The same article describes an AI interview scheduling tool.
No outcome disclosed.
Federal Bureau of Prisons
United States · Government and public sector · 2023
The Federal Bureau of Prisons, part of the US Department of Justice, runs Pathfinder to help its employees with career pathways. The system generates assessments from what the user enters, scores them and offers the employee options for career pathways. The inventory lists it as deployed since September 2023 on Azure. It is narrower than a full talent marketplace (no project or mentor matching is described) and no outcome figures are published.
No outcome disclosed.
Unilever
United Kingdom · Manufacturing · 2020
Unilever's Flex Experiences platform uses AI to match employees with project opportunities across the business, so that people can work on projects for part of their time and build new skills while project leaders find the expertise they need. By 2020 it had been rolled out to more than 60,000 employees in more than 100 countries, and during the pandemic lockdowns of 2020 teams used it to staff urgent work quickly, such as an information and analytics (I&A) squad. Unilever publishes no outcome figures on the page cited.
No outcome disclosed.
Schneider Electric
France · Manufacturing · 2020
Schneider Electric's internal surveys, as reported by Gloat, showed that nearly half of departing employees cited a lack of internal growth opportunities as their main reason for leaving. The company launched Open Talent Market, an AI talent marketplace, with a pilot in HR and a global launch in April 2020. Employees build a profile with their skills and aspirations and receive recommendations for part time projects, internal positions and mentors; managers posting projects see employees whose skills fit. Gloat reports that more than 2,300 employees began to explore new roles within the first two months, alongside an adoption rate above 60% whose base it does not state, and cumulative figures of more than 360,000 unlocked hours and over USD 15 million in productivity gains and reduced recruitment costs.
- Users served: at least 2300, first two months after launch
"Within the first two months of launch, the platform achieved an adoption rate that surpassed 60%, enabling more than 2,300 employees to begin to explore new roles within the business."
Claimed by: vendor - Cost savings: at least USD 15 million, cumulative since the April 2020 launch, end date not stated
"To date, Schneider Electric’s talent marketplace has unlocked more than 360,000 hours and created a savings of over $15,000,000 in productivity gains and reduced recruitment costs."
Claimed by: vendor
How do you implement it?
A model agnostic playbook: what to prepare, the order to build in, and what goes wrong.
Data you need
- Employee profiles with skills, experience and aspirations, confirmed by the employee
- A skills taxonomy or ontology mapped to roles
- Open roles, projects and mentoring offers with required skills and time
- Policy on eligibility, time allowance and manager approval
Systems to integrate
- HR information system (employee records, org structure, job architecture)
- Applicant tracking system for internal roles
- Learning management system
- Collaboration tools for notifications, such as Microsoft Teams
Complexity: Medium
The matching technology is available off the shelf. The hard parts are a usable skills taxonomy, manager behaviour (posting work and releasing people), works council and data protection agreements, and making the platform fair and explainable enough to be trusted.
- 1
Agree the rules before the platform
Decide who can take part, how much time employees may spend on gigs, how managers approve and how internal applications are treated. Involve works councils early, because the system processes employee data and influences career decisions.
- 2
Pilot in one function
Schneider Electric piloted in HR before a global launch. Choose a function with enough project work, seed it with real opportunities and measure participation and manager feedback.
- 3
Let employees own their profile
Show employees the skills the system inferred and let them correct them. Profiles people trust produce matches people accept.
- 4
Test the matching for fairness
Compare recommendation and selection rates across groups, check that protected characteristics and proxies are not used, and document the results.
- 5
Connect it to workforce planning
Use aggregated skills data to plan learning and hiring, and report outcomes such as roles filled internally and contractor spend avoided, not only registrations.
Guardrails
- Recommendations only; managers and employees make every selection decision
- No protected characteristics or obvious proxies in matching features
- Employees can see, correct and delete inferred skills
- Regular fairness testing of recommendations and outcomes, with results retained
- Transparent explanation of why an opportunity or candidate was suggested
KPIs to instrument
- Active users per month as a share of employees, not only registrations
- Roles and projects filled internally, and time to fill
- Recommendation acceptance rate and selection rate by group
- Contractor spend and external hires avoided for filled gigs
- Retention of participants versus comparable non participants
Human in the loop
Employees decide what to apply for, managers decide whom to select, and the employee's own manager agrees the time. HR owns the taxonomy, reviews fairness reports and handles complaints about recommendations.
Common failure modes
- Registrations without activity
- High sign up figures hide low regular use. Report monthly active users and filled opportunities.
- Managers do not release people
- Employees apply but their managers block the time. Set a time allowance policy and make releasing talent part of manager goals.
- Biased matching
- Skills inferred from past roles reproduce past inequalities in who got which job. Test outcomes by group and let people correct their profiles.
- Stale skills data
- Profiles are filled once and never updated. Refresh them from completed assignments and learning, and prompt employees periodically.
What are the risks and rules?
EU AI Act
Depends on design
Annex III point 4 lists AI used for the recruitment or selection of natural persons (4(a)) and AI used to make decisions affecting promotion, or to allocate tasks based on individual behaviour, personal traits or characteristics (4(b)). A marketplace that ranks employees for internal roles or allocates projects on the basis of inferred traits is therefore high risk. Recommending learning content or mentors to an employee who chooses freely is usually not. Deployers of the high risk part must inform workers' representatives and the affected employees before use (Article 26).
Rules that apply
Guidance
- Annex III, high risk AI systems referred to in Article 6(2) (European Union, Europe). Point 4 covers employment, workers' management and access to self employment, including selection, promotion and task allocation.
- Article 26, obligations of deployers of high risk AI systems (European Union, Europe). Human oversight, logs, and informing workers' representatives and affected workers before a high risk system is used at the workplace.
- Automated employment decision tools (Local Law 144) (New York City Department of Consumer and Worker Protection, North America). Employers may use an automated employment decision tool only after a bias audit, with the results published and notices given to employees or job candidates. Because the notices also cover employees, check whether a tool that ranks employees for internal roles is in scope under the law's definitions.
Controls to put in place
- AI Act classification per feature, with the high risk features managed as such
- Data protection impact assessment and works council agreement where required
- Fairness testing before launch and at least yearly, with results retained
- Named HR owner for the matching logic and for complaints
- Supplier due diligence on model documentation, data use and hosting
Frequently asked questions
- What results do companies report from AI talent marketplaces?
- Mastercard says 90% of its workforce is on its Unlocked marketplace, with 500,000 project hours delivered. Gloat reports that more than 2,300 Schneider Electric employees began to explore new roles in the first two months, and over USD 15 million in productivity gains and reduced recruitment costs since the 2020 launch. Registration is not the same as regular use, so track active users and filled opportunities.
- Is an internal talent marketplace high risk under the EU AI Act?
- Parts of it can be. Ranking employees for internal roles or allocating tasks based on inferred traits falls under Annex III point 4. Recommending mentors or courses that the employee chooses freely usually does not. Classify each feature and inform workers' representatives before using a high risk one.
- Does it replace internal recruiters?
- No. It makes opportunities visible and suggests matches; managers and recruiters still select. The gain is in speed and reach, especially for short projects that would otherwise go to a contractor.
How to cite this page
Blits.ai AI Use Case Library, "AI internal talent marketplace for matching employees to projects, roles and mentors", last verified 27 September 2026, https://www.blits.ai/ai-use-cases/internal-talent-marketplace-matching. Licensed under CC BY 4.0. Method: how we verify use cases.
Changelog
- 27 September 2026: First published