Six months from signature to go-live: how Red Hat scaled skills intelligence at builder speed

Business context: a builder culture meets the speed of AI

Red Hat is the world’s leading provider of enterprise open-source software. Its culture is built on a simple instinct: do not just consume technology, experiment with it and iterate. That instinct is what makes the company’s approach to its own workforce worth studying, because it applied the same builder mindset to a problem most large enterprises stall on for years.

‍

The pressure was real. As AI reshapes how engineering work gets done, and as Red Hat integrates more deeply into IBM, leadership could not plan the technical workforce on top of static job descriptions. The company carried more than 1,600 roles in an architecture where thousands of associates shared generic titles like “Senior Program Manager” while doing very different technical work. Leaders called this “title-blur,” and it made even an intuitive call, such as merging Software and Quality Engineering into a single role, a guess rather than a data-backed decision. The team needed a dynamic, task-level map of its engineering DNA, and it needed one fast.

Challenges: the gap between a skills vision and an operational reality

Leadership had a clear ambition for a skills-led enterprise. Four practical hurdles stood between that ambition and day-to-day reality.

‍

• Title blur across 1,600+ roles. Generic titles hid the actual work being done, so the move toward a unified engineering role had no evidence base behind it.

• Experience fragmentation. Associates had to bounce between separate modules for career goals, learning, and job roles. The friction led to experience fatigue and low engagement with the career tools that already existed.

• A metadata problem at scale. With tens of thousands of learning titles, most of them sourced from external providers, manual skill-tagging was never going to keep up. The recommendation engine was effectively blind to the content associates needed for specialized upskilling.

• Legacy systems that move like trains on tracks. Traditional enterprise systems are stable, but slow to pivot. Red Hat needed skills intelligence in associates’ hands at the speed of its AI roadmap, not at the speed of the next platform release cycle.

How we helped: a headless intelligence layer the team could build on

Rather than wait for legacy systems to catch up, Red Hat used TechWolf as an intelligence layer it could build directly on top of. The engagement followed three pillars.

‍

‍Trusted data. TechWolf harmonized data from several fragmented sources, including Workday, Credly, and Skills Base, and used AI to infer skills from the work itself rather than from surveys. That produced a validated source of truth for the roughly 2,500 associates in the pilot, fit for real talent decisions instead of another framework exercise.

TechWolf’s ability (and agility) to connect our diverse systems -HR, learning, business- truly impressed us. Moving beyond traditional silos of data to connect the dots with critical, relevant information opened a whole new world of possibilities for us to truly understand the skills we have and the skills we need."
— Jaime Tobar-Espinoza, Senior AI Specialist at Red Hat

‍Executive insights. The team translated raw skill data into a decision leadership could actually act on. Task intelligence showed that Software and Quality Engineering roles shared over 70% of their work, more than 28 hours a week, which gave the engineering leadership team the evidence to finalize the Unified Software Engineer role architecture and to flag where AI augmentation would land first.

‍

‍Embedded actions. Red Hat brought skills data to where associates already work. TechWolf’s Skill Assistant sits inside Slack, so associates validate their skills in the flow of work. On top of that, Red Hat’s developers used TechWolf’s API to “vibe-code” Skills Builder in a matter of weeks: a conversational front door that lets associates see their skill gaps and prompt their way to learning paths without touching a legacy interface. The pilot proved the pattern; the go-live scaled it.

We’re building our skills strategy like an open-source project - transparent, shared, and designed to connect value across our talent ecosystem. The Skill Assistant global launch is our version 1.0 - a clear starting point we will continuously improve, refine, and scale with all Red Hatters to unlock long-term agility and potential."
— Mona Yadav, Global Skills Strategist at Red Hat

Outcomes: a pilot that earned the right to scale

From signature to a company-wide go-live in six months

Red Hat signed its expansion in December 2025. By early June 2026, three systems were live in production at once: the Skill Assistant in Slack, the input integrations feeding HR and job data from Red Hat into TechWolf, and a Workday skill sync pushing validated skills back into Workday to power internal search. On June 3, more than 18,000 associates worldwide were prompted to validate their profiles, a sevenfold jump from the pilot population. Standing up a credible skills foundation typically takes large enterprises two to three years; Red Hat did it in six months.

Adoption without the friction

Scale usually comes at the cost of trust. Here it did not. In the first week after go-live, associates accepted 89% of the skills TechWolf inferred for them, and 28% of the full workforce had already engaged, with no manual form-filling involved. That built on what the pilot had shown: when skills surface in the natural flow of work, associates validate them and the awkwardness of career self-assessment falls away.

One use case already live

Validated skills are not sitting in a dashboard. The Workday Career Hub was already available, but with the newly populated skills profiles, associates can use them to find internal mobility opportunities right away. That closes the loop the company set out to close: from a more accurate picture of who can do what, to associates acting on it themselves.

I’ve never seen a vendor that has been so in lockstep with us through the journey."
—Tomislav Vujec, Director, Skills Solution Center and Red Hat co-sponsor

Governance: trust earned in an engineering-first culture

In a technical culture, data only wins if experts vet it. Red Hat ran a multi-layered validation model so the foundation stayed credible as it scaled.

‍

• Human-in-the-loop SME review. Subject matter experts across functions used the TechWolf Console to refine taxonomy clusters, so the model understood the specifics of Red Hat’s own products, from OpenShift to Ansible, rather than treating them as generic keywords.

• Associate validation by design. Every inferred skill is presented to the associate to confirm or correct. The 89% acceptance rate is not a passive metric; it is the governance mechanism, with people checking the data as they use it. For instance, the associates themselves proposed to increase the importance of certifications in the inference process, as those are part of their professional development.

• A privacy stance grounded in existing policy. To clear adoption friction, the team mapped every data object back to existing employee privacy statements, showing that the program made more intelligent use of data Red Hat already held rather than collecting anything new.

Project challenges: the honest parts of the work

• Data delivery timing. Early in the pilot, delays in delivering source data put pressure on downstream milestones. The team worked through it by tightening the integration format and pulling the Workday team in earlier than originally planned.

• Works Council scrutiny. Ahead of go-live, a works council raised a concern that the tool would automatically profile employee performance (i.e., that there may be a consent and high-risk issue under GDPR and the EU AI Act). Red Hat addressed it head-on with a communication to clarify that (1) the TechWolf Skill Assistant suggests skills, it does not score performance or make automated decisions about roles or pay, and (2) participation always stays voluntary with an option to opt-out.

• Taxonomy harmonization. Red Hat’s proprietary product language did not map cleanly onto a generic skills taxonomy at first. Iterating with SMEs on the taxonomy was the unglamorous work that made the inference trustworthy at scale.

Next steps: from one use case to a workforce strategy

With the foundation live, the roadmap shifts from proof to depth. The near-term moves are specific and already in motion.

‍

Business data ingestion is next, bringing Jira and Salesforce signals into the picture while the team explores on-premise processing. That richer signal then feeds a proficiency-level rollout, moving from “which skills” to “how strong,” so suggested proficiency levels can sharpen internal search and mobility.

‍

In parallel, Red Hat is onboarding its job architecture team to design a job governance flow, keeping the role architecture current as work keeps changing. The longer arc is to introduce work intelligence as a repeatable method for every business unit: measure how roles actually evolve, find the human capacity that AI frees up, and redeploy that talent with the same speed the company brought to the go-live.