The Skills Platform Graveyard

I have watched this movie too many times.

The first show ran from about 2005 to 2015. Skills databases and competency frameworks. SAP, PeopleSoft, SuccessFactors. Vast taxonomies. Committee meetings about whether negotiation was a skill or a competency. Binders.

Second show, 2016 to 2022. Talent marketplaces and internal mobility platforms. Gloat, Fuel50, Hitch. Genuinely good products. Real ambition. Match people to gigs, unlock the hidden bench, stop hiring what you already own.

Third show, 2021 to now. AI skills inference and workforce intelligence. Eightfold, SkyHive, Reejig. Smarter than anything that came before. No arguing about taxonomies. The machine reads the resume and tells you what you have.

Three decades of engineering. Three completely different architectures. One consistent ending.

Most people blame the data. Dirty data, stale data, employees who will not fill in their profiles. That is the accepted explanation because it implies a fix.

I think that explanation is wrong.

The real problem is who decides
Companies make workforce decisions through three separate systems.

System 1 is strategy. This is the CEO or the strategy office, and the business unit leaders. It asks which markets we will enter, which products we will build, and which capabilities we will need to pull it off. Horizon: three to five years.

System 2 is financial planning. This belongs to the CFO/CHRO. It asks how many people we can afford, what headcount growth is approved for, and which roles are funded. Horizon: one budget cycle.

System 3 is hiring. It belongs to CHHRO and TA. It asks how we fill this req, run this process, and onboard this person. Horizon: right now.

Here is the fatal part.

Skills platforms are sold to System 3. The problem lives in System 1.

The pattern is almost boringly predictable
Phase one is enthusiasm. An executive says we need to understand our workforce skills. Nobody disagrees, because who would? A platform is bought. An initiative is launched with a name and a logo.

Phase two is collection. Employees are asked to list their skills, update their profiles, complete the survey. Participation is decent in week one. No one remembers in week six.

Phase three is decay. Twelve to eighteen months in, the data is stale. There is no incentive to maintain it. Managers do not look at it. The people who do look at it do not trust it.

Phase four is abandonment. Decision-makers go back to what has always worked: their personal networks, a manager’s recommendation, or a req posted externally. Nobody kills the platform. It just stops being opened.

I have seen this in companies with excellent HR teams and generous budgets. Competence was never the issue.

AI does not rescue this
The current pitch is that AI fixes the collection problem. Read the resumes. Read the project data. Read the messages, the repos, the documents. Skip the survey entirely.

That is an improvement.

But it solves the wrong half of the equation. The constraint was never how the data got in. The bottleneck is that no one with real decision-making authority is required to use it. You can create a perfect skills graph and still have a CFO who funds headcount off last year’s spreadsheet and a business unit leader who hires the person he golfed with.

Better data into a system with no authority just produces a better dashboard nobody opens.

In other words, garbage in, garbage out

What survives
Systems that live for a decade instead of eighteen months have one thing in common: They happen anyway.

Professional services staffing systems are the clearest case. People use them every single day, not because anyone is enthusiastic, but because you cannot staff a project without them. The work does not move otherwise.

Engineering platforms are the other case. GitHub. Nobody at GitHub ever asked you to declare your skills. Your skills are simply visible as a byproduct of doing the work.

Two lessons fall out of that.

First, skills data must be derived from work, not stated by people.

Second: deriving skills from work is necessary but nowhere near sufficient. The system has to touch an actual operational decision.

Pick one. Project staffing. Budget approval. Strategy execution. If your platform influences none of those, it is a reporting tool, and reporting tools get abandoned.

The data is not in HR
There is a second reason these things struggle, and it is awkward for the vendors.

The most valuable workforce data does not live in HR systems. It lives in the CRM. In project management tools. In delivery systems. In engineering repos. In the sales pipeline.

HR knows job titles. HR rarely knows capabilities. Those are different things, and the gap between them is exactly the thing everyone is trying to buy.

Which means the right architecture is not another HR module. It is an intelligence layer that sits above HR systems and pulls from operations. The smarter startups are figuring this out. But the implication is bigger than most of them are willing to say out loud.

If that is the architecture, the category is not HR technology at all. It is closer to enterprise planning software. Operational intelligence. Strategy tooling.

Different buyer. Different budget. Different conversation entirely.

Capability Operating Systems
I think a new category will emerge over the next decade. I have written about this often. l Call it the Capability Operating System.

It answers questions like these: Do we have the expertise to enter this market? Which teams can execute this strategy? Which of our people can be redeployed as AI eats the middle of our work?

Almost no organization can answer those today. Not with evidence. They answer with anecdote and a hopeful slide.

That is a very large gap, and it is widening fast as AI reshapes what work even is.

The question every founder in this space decides is which company she is building
You are building one of two companies.

Company one is a workforce analytics platform. Low strategic weight. Slow adoption. Crowded market. Sold to HR, budgeted as a nice-to-have, but cut in the first bad quarter.

Company two is strategic capability infrastructure. Sold to the C suite. Tied to whether the company can do the thing it just told investors it would do. Much bigger prize. Much harder build.

Those are not two positionings of the same product. They are two different companies with two different buyers, two different price points, and two different fates.

Where am I?
The market need is real and growing. The concept is sound. The technology finally works.

The positioning is the whole ballgame.

The biggest risk in this category is being mistaken for a skills platform, because skills platforms have a graveyard, and buyers have visited it.

The biggest opportunity is becoming the system that tells an organization whether it can execute its own strategy.

Nobody owns that yet. Somebody will.