Herizon logo

Skill system revamp

Today I’m rebuilding completely how Herizon system handles skills.

Mari Luukkainen
Summarize with AI:
Skill system revamp

Today I’m rebuilding completely how Herizon system handles skills.

The core problem: when someone says "I know Python" it currently means the same thing whether they passed a quiz, shipped a real project, or got paid for it professionally. Companies don't care about the first one since they hire on the last two and our system couldn't tell the difference.

I'm building skill levels: Knows it, has done it, has shipped it, has been paid for it. A quiz pass is level one - this is what is already built in our system so community members can vacuum the basic understanding on employability improving topics. A course with a deliverable is level two. A GitHub repo or portfolio piece is level three. An actual hire is level four. Same skill, completely different signal strength.

Here's what's changing on top of it:

Skills now decay. A 2022 OKR certification looks the same as a 2026 one in our current system and that’s broken. After 12 months without a refresh (new course, new project, new work entry) the skill downgrades. Floria (our AI) prompts you: "you haven't shown anything for this in 18 months, here's how to refresh it." School system is horrible in this and teaches outdated content for years, while skillset is currently changing every few months.

We're closing the loop with hiring outcomes. After every placement we ask the company which skills actually mattered and which were missing. After every rejection we tell the candidate the gap and recommend the path that closes it. That feedback loop feeds back into the AI matching weights and course recommendations. Right now we credit skills and never check if they predict hires. That's changing.

Courses are becoming paths. Instead of 30 standalone courses you browse and pick from, you tell the system "I want to be a frontend dev in Helsinki" and it generates a path: these courses, these assessments, these portfolio tasks, a mock interview, in order, with a percentage tracker toward "ready." Same content, opinionated framing. Completely changes engagement.

Skills are also going public. Right now our verifications live in our database. A candidate can't show them to anyone outside Herizon. We're building shareable profiles, LinkedIn skill push, PDF certificates, and a "Verified by Herizon" badge that links back so a recruiter can validate independently. Our verifications are worthless if they're invisible outside our pool.

The job market data layer already exists. On Työmarkkinadata we source in-demand skills from 100K+ global jobs weekly. Now that feeds directly into skill recommendations. "This skill is in 30 active postings in Helsinki right now" next to every course.

Finally Floria stops being a CV assistant and becomes an actual coach. Not "I'll help you write your CV" but "You've applied to 4 backend roles this month. Three rejected on system design. Take this course, redo this assessment, then we re-apply." It should be continuous, outcome-aware, opinionated.

The whole system moves from "did you learn about this" to "can you prove you can do this, and does the market care." That's the difference between a badge and employability.

Want more insights?

Subscribe to our newsletter for research updates, community news, and employment insights.