TL;DR
AI internal mobility tools match employees to internal job opportunities, projects, and career paths using skills data instead of job titles. Seven criteria separate a platform that moves people from one that quietly becomes an internal job board: the quality of the underlying skills data, two-way matching, evidence-based career paths, the breadth of the opportunity catalog, explainable recommendations, integration with the existing HR stack, and built-in support for manager adoption.
Most evaluations go wrong in the same place. Buyers compare feature lists and demo interfaces, then discover eighteen months later that the matching only works as well as the skills data nobody maintained.
Employers expect 39% of workers' core skills to change by 2030, and 63% name skills gaps as the single biggest barrier to business transformation (World Economic Forum, Future of Jobs Report 2025). Internal mobility software is one of the main instruments organizations are buying to respond. It is worth choosing carefully.
1. The skills data underneath matters more than the interface
An AI internal mobility tool is only as good as the skills data feeding it. Ask how the platform builds employee profiles: from self-declaration alone, or by inferring skills from CVs, job history, project assignments, and learning records. Then ask how that data stays current, and who owns the taxonomy when roles change.
Self-declaration alone produces thin, uneven coverage. The employees who need visibility most are usually the ones who never fill in their profile. Inference from data your organization already holds gives you a usable starting point, which employees then correct rather than create from scratch.
Deloitte's research on skills-based talent models found that organizations getting results are far more likely to maintain a library covering both sides of the equation: the critical skills the business needs, and the skills their workers actually have. A tool that only models one side will produce matches you cannot act on.
Three questions worth asking every vendor:
- What percentage of employee profiles are populated on day one, before anyone logs in?
- Does the skills framework work across all the languages our workforce uses?
- When we create a new role, does the taxonomy update automatically or does someone maintain it manually?
2. Employee skills matching should run in both directions
Matching works both ways. Employees need to see internal job opportunities that fit their skills and their goals. Managers and recruiters need to surface qualified internal candidates they would never have found by filtering on job titles. A tool that only does one direction leaves half the value unbuilt.
One-directional tools tend to be dressed-up internal job boards. The employee searches, the employee applies, and the hiring manager still reviews a shortlist assembled the old way. Nothing changes on the demand side, which is where most internal mobility programs stall.
Two-way matching also changes what recruiters do. Instead of posting a role and waiting, they open a search across the internal skills graph and see who is 80% of the way there, including people in adjacent functions with no obvious title match.
3. AI career pathing is only useful when the paths are evidence-based
AI career pathing should be built from real signals: internal moves people have actually made in your organization, measured skills adjacency between roles, and current business demand. A tool that renders a static ladder with an AI badge on it will not tell an employee anything their manager could not already say.
Ask a vendor to explain where a specific suggested path comes from. A good answer references the skills overlap between the current role and the target, the gap that remains, and the learning or experience that closes it. A vague answer about proprietary algorithms is a signal.
Also ask what happens with cold starts. Most organizations have plenty of historical data for common roles and almost none for emerging ones. The roles you most need to fill in three years are exactly the ones with no move history, so the model needs to reason from skills rather than precedent.
4. A talent marketplace is more than a list of open roles
Full-time vacancies are the rarest unit of internal mobility. Good talent marketplace software also circulates projects, short missions, part-time assignments, mentoring, and job shadowing. These are how employees test a direction before committing to it, and how organizations move capacity without moving headcount.
When Deloitte interviewed 13 organizations investing in these platforms, 58% described the purpose in deployment terms such as skills matching, while around 50% described it as talent mobility (Deloitte Insights, 2020). Both framings matter, and the platforms worth shortlisting handle both in the same system rather than treating projects as a separate module.
There is a timing argument too. Mobility that only triggers when a vacancy opens is reactive by design. A marketplace that surfaces a two-month project can redeploy someone within weeks, and it generates fresh skills data every time an employee completes an assignment.
5. Explainability and bias controls are not optional
An internal mobility tool decides who sees which opportunity, so it shapes careers whether or not anyone calls it a decision system. Ask the vendor to explain any single recommendation in plain language, to show which data influenced it, and to document how the system is tested for bias across gender, age, and background.
For European organizations, this is also a regulatory question. AI systems used in employment and worker management sit in the high-risk category under Annex III of the EU AI Act (Regulation (EU) 2024/1689), which brings documentation, transparency, human oversight, and record-keeping obligations. Regulation (EU) 2026/1744, the Digital Omnibus on AI, in force since 27 July 2026, moved the application date for stand-alone Annex III systems from 2 August 2026 to 2 December 2027. The obligations themselves did not change. The runway did.
Practically, that means procurement should be asking now for the technical documentation, the human oversight design, and the data governance evidence that will be required anyway. Vendors who cannot produce it during evaluation are unlikely to produce it after signature.
Two more things to verify: whether an employee can see and correct what the system holds about their skills, and whether HR can audit why a given population is or is not being matched to a given family of roles.
6. Integration with the HR stack determines whether it gets used
An internal mobility tool that does not talk to your HRIS, your ATS, and your learning platform becomes a parallel system that someone has to keep in sync by hand. Check that data flows in both directions, that a new requisition appears without manual re-entry, and that completed learning updates the skills profile automatically.
The learning connection is the one buyers most often underrate. Career pathing that identifies a skills gap and then stops is a diagnosis without a treatment. When the platform can route an employee straight into the training that closes the gap, and then reflect the completion back into their profile, workforce development stops being a separate initiative.
Ask for the integration to be demonstrated with your own systems during the pilot, not described in a slide. Ask what happens when your HRIS changes its data model, and who is responsible for the connector.
7. The tool does not fix manager behavior, so check what supports it
Software cannot solve talent hoarding on its own. If managers are measured only on their own team's output, they will keep blocking moves regardless of what the platform recommends. Look for the features that make mobility visible and low-friction for managers, and be clear about the policy work that sits alongside the tool.
The size of that problem is documented. Deloitte reports that 46% of managers resist internal mobility, a behavior reinforced by misaligned management incentives. In Deloitte's 2019 Global Human Capital Trends survey, 76% of respondents rated internal talent mobility important, yet more than half said it was easier for an employee to find a job outside the organization than inside it.
Useful platform features on this front include manager-side visibility into who on their team is exploring, notice periods and backfill workflows built into the move process, and reporting that credits managers who export talent. Alongside that, set an internal-first hiring policy with teeth, and agree on the metrics before launch.
Track at minimum: internal fill rate, profile completeness, time to staff a project, and regrettable turnover in populations using the platform. LinkedIn's 2022 Workplace Learning Report found that companies excelling at internal mobility retained employees for an average of 5.4 years, against 2.9 years for those that struggled with it. That gap is the business case, and it is measurable in your own data within a year.
What to do next
Build the evaluation around evidence rather than demos. Ask each vendor to run a proof of concept on a real slice of your workforce data, then judge four things: how complete the generated profiles are before employees touch them, how sensible the matches look to the managers who know those people, how clearly the system explains itself, and how much manual work the integration leaves behind.
