top of page
Search

Internal Talent Marketplaces 2.0: How AI Matching Is Rethinking Internal Mobility

Writer: Marcus
Marcus
Aug 16
6 min read


Somewhere in your own workforce, there's probably someone who's exactly who you're searching for outside the company. The catch: neither they nor the department drafting that external job posting has any idea. Sounds absurd, but it's daily reality at most companies: between 30 and 40 percent of the skills already inside an organization stay unknown or unused, simply because nobody asks systematically. And more than half the workforce – 51 percent – doesn't even know which internal opportunities exist in the first place.


This is exactly where the new generation of internal talent marketplaces comes in. Not as a digital bulletin board for open roles, but as an AI-powered matching engine that connects skills, interests, and open positions across a company in ways no intranet ever could.



From Job Board to Matching Engine


The first generation of internal talent marketplaces was essentially an internal job board with a search bar: employees had to actively browse, assess themselves, and hope the posting matched their profile. If you didn't know what to search for, you found nothing.


The second generation – the one this article is about – flips that logic. Instead of people searching for roles, the system searches for people. An AI matching engine continuously compares skills profiles against open roles, projects, mentoring opportunities, and learning paths, and proactively suggests good fits – before anyone has even started looking. A passive directory becomes an active matchmaker.




What's Under the Hood


For this to work, three technical building blocks have to come together.

The first is a skills ontology – a structured database that captures abilities not just as free-text keywords but in relation to one another: which skills are related, which build on each other, which actually matter for which role. Without this structure, every AI recommendation is just guesswork.


The second building block is skill inference: the software doesn't just derive abilities from resumes or self-reported data, it also reads them from actual behavior – completed projects, learning content, internal collaboration, performance data. If you've managed a team's budget planning for three years, you don't need to note that on a CV for the system to pick it up.


The third – and currently the most exciting – building block is the jump to agentic AI. The first software generation showed employees matching offers to click through themselves. The current generation goes further: it proactively initiates career conversations, coaches people through identified skill gaps, and suggests concrete next steps instead of just presenting a list.


A recommendation feature becomes an active digital career coach that thinks along instead of just sorting.



The Business Case in Numbers


For the finance people who have to sign off on these systems, it ultimately comes down to the math – and the math is remarkably clear. External hires cost roughly 1.7 times as much as filling the same role internally once you factor in recruiting effort, onboarding, and ramp-up time. Mastercard has put a number on its internal talent marketplace: 21 million US dollars in documented savings. One international organization cut its external recruiting spend by 30 percent within 18 months.


Market momentum tells the same story: the share of companies actively using AI in their talent strategy has more than doubled, from 17.9 percent in 2025 to 42.3 percent. The talent marketplace software market is growing more than 15 percent a year. According to Gartner, more than half of all HR leaders plan to invest in an internal talent marketplace within three years – not as a nice-to-have, but as a response to skills scarcity, retention risk, and the desire to fill roles internally faster than external recruiting ever could.



Who Builds the Marketplaces


Anyone shopping for the right software runs into two camps. On one side are specialized vendors like Gloat, Eightfold, Phenom, Beamery, Fuel50, or the European player Neobrain, whose entire product is built around skills matching.



On the other side, the big HR suites are folding the function straight into their existing systems: SAP has built an AI-powered recommendation engine directly into its SuccessFactors Opportunity Marketplace – already running inside the HR suite many large companies across the DACH region (Germany, Austria, and Switzerland) use anyway – Workday is following suit with Skills Cloud, and Cornerstone OnDemand is adding the same logic to its learning system.


For HR tech leaders at larger organizations, that's often the real strategic question – not "AI matching, yes or no," but "best-of-breed point solution or a module of the suite we already have." Companies already running SAP or Workday save themselves the integration effort. Those who want deeper skills analytics and faster iteration tend to look at the specialized vendors, some of which – Fuel50, with references like KeyBank, for instance – already show solid usage numbers: 72 percent of users return to the platform regularly, and more than 9,800 skills have been captured and rated across the organization.



Where the Technology Hits Its Limits

As mature as matching engines have become, they don't automatically resolve the real blockers to internal mobility. Three of them show up in practically every rollout.


Manager Hoarding

60 percent of high performers name their own manager as the biggest obstacle to an internal move. No AI recommendation, however good, helps if managers would rather hold onto their best people than let them go.


Data Quality

A matching engine is only as good as the skills data behind it. Outdated profiles, patchy skills taxonomies, or inconsistently maintained systems produce recommendations nobody takes seriously – and once a system loses people's trust, it doesn't get a second chance.


Algorithmic Bias

When an AI learns from historical promotion or project data, it also absorbs the patterns that produced that data – for instance, if certain groups were historically put forward for leadership projects less often. Reputable vendors explicitly test their matching models for bias before going live. Anyone introducing such a system should ask about this upfront, not after the fact.



The DA(-CH) Wrinkle: Co-Determination Is Mandatory, Not Optional


Anyone rolling out an AI-powered talent marketplace in Germany, Austria, or Switzerland runs into something international vendor pitches rarely mention: the works council, Germany's elected employee representation body with statutory co-determination rights. A ruling by Germany's Federal Labor Court this year tightened co-determination requirements for AI systems even further – what matters now isn't whether a system was designed to monitor behavior or performance, but whether it's objectively capable of doing so. Under Section 87(1) No. 6 of the Works Constitution Act, that's already enough to trigger a co-determination right.


For skills-matching systems that continuously evaluate performance and project data, that threshold is essentially always crossed. Companies that only loop in the works council after signing a vendor contract – or after the system is already live in a pilot – risk a court-ordered shutdown, and forfeit exactly the trust that later determines whether people actually use the system.


The pragmatic path: bring the works council and data protection officers in from day one and lock down the details in a formal works agreement, rather than trying to retrofit one at the end.



What to Clarify Before Rolling It Out


Before rolling out an internal talent marketplace, it's worth taking a sober look at five questions:


  • How clean and complete is the existing skills data, really – and who keeps it updated on an ongoing basis?

  • Is approving internal moves explicitly built into managers' own performance goals, or does mobility stay a vague aspiration?

  • Does a specialized point solution fit the existing systems landscape better, or is a module within the current HR suite the more pragmatic route?

  • Has the matching algorithm been demonstrably tested for bias before it starts influencing career decisions?

  • And: are the works council and data protection team at the table from the start, rather than showing up right before go-live?


Realistic timelines help manage expectations too: pilots typically run three to six months, a company-wide rollout more like twelve to eighteen – and aligning with the works council belongs in that timeline, not in a footnote afterward. Communicating that openly upfront saves you from disappointed business units expecting instant results.




The Technology Is Ready – The Organization Has to Catch Up


AI matching has turned internal talent marketplaces from a digital bulletin board into a genuine hiring engine. The technology can now recognize who in an organization has which skills and proactively suggest good matches – before anyone even starts looking. The next step, agentic AI that actively coaches rather than just recommends, is already live in production at a handful of vendors.


What's left is the human side of the equation: managers who have to let go, data upkeep nobody enjoys, and algorithms that need to be built to distribute opportunity fairly instead of repeating old patterns.


The best matching engine is worthless if the organization around it doesn't keep pace. That's where the real work lies for the years ahead.

Sources


Comments

Binningen, Schweiz

Abo-Formular

Vielen Dank!

  • LinkedIn
  • Twitter
  • Pinterest
  • Facebook

©2020 Marcus Fischer

bottom of page