How GSK turned a skills foundation into smarter learning and workforce planning

TL;DR

GSK is a global pharmaceutical company with around 66,000 employees, spread over more than 80 countries. Recently, they have rebuilt their skills and learning foundation, collapsing more than 20 fragmented learning systems into a single AI-powered L&D hub. What started as a commercial problem, the rising cost of buying skills externally and slow time to value, became one connected data foundation that now powers personalized learning, skills intelligence, and GSK's first agentic workforce-planning pilots. More than 80% of employees have rated AI-suggested skills in the hub, and the significant decommissioning of legacy systems unlocked millions of dollars to fund the transformation itself.

Business context: a commercial problem in a learning problem's clothing

GSK's R&D engine had scaled significantly. Investment had doubled since 2016, the company had grown at double digits for several years, and more than half of the pipeline was now shaped by business development and strategic partnerships. That growth pushed GSK into areas that demanded capabilities it could not build internally quickly enough, so it bought those skills on the open market at a premium cost. Internal mobility was limited, and in a regulated, complex business, external hires take a long time to reach productivity. The result was a higher workforce cost and a slower time to value.

When GSK interrogated why, the answers were uncomfortable but clear. There was no enterprise visibility into skills, no common language for what good looks like across roles, and a decentralized development offering that was generic rather than targeted. Underneath it sat a strong complexity: a learning technology estate of more than 20 systems. Named as a business capability and mobility problem rather than a learning one, the imperative was obvious.

The trigger wasn't really a skill problem, and it wasn't even a learning problem. To be honest, it was a commercial one.”
— Zaka Farhat, Global SVP at GSK, speaking with David Green on the Digital HR Leaders podcast

Challenges: building the conditions before the platform

Buying a platform and switching it on is the easy part. The real transformation was the plumbing underneath, and several conditions had to exist before any platform could land.

• No enterprise visibility and no common language. Skills were dispersed, there was no shared definition of what good looks like across roles, and development was decentralized, with potential duplication of investment across business groups & geographies.
• A fragmented, costly technology estate. More than 20 learning systems were expensive to run and impossible to connect into a single usable data foundation, the very thing GSK needed most.
• A job architecture built for grading, not for work. Like many companies, GSK used its job architecture for salary benchmarking and grading rather than to capture the actual work. That had to be revamped before AI could infer skills reliably.

Buying a platform and switching it on is the easy bit. The real transformation was building the plumbing underneath, and the conditions that have to exist before that platform even stands a chance of landing.”
— Zaka Farhat, GSK, on the Digital HR Leaders podcast

What GSK built: one data foundation under the whole talent system

TechWolf became one of the key supporting layers for GSK's people data, inferring skills from the job architecture and market signals, and connecting skill, learning, workforce, recognition, and mobility data so it tells one story rather than sitting in separate silos.

Trusted data. AI inference mapped skills to jobs and people and cut roughly 90% of the time the traditional manual approach would take. The technology did roughly 80% of the skills mapping, with subject matter experts fine-tuning the final 20%. The taxonomy carries both technical and behavioral layers, moving the organization off scattered competency models toward a common capabilities-to-skills language, grounded in a future-proofed job architecture.

Executive insights. GSK’s intelligence layers shows where to reallocate learning investment, where to double down on capability gaps, and where content sits unused. That same skills data will reshape workforce planning, moving GSK away from a once-a-year headcount allocation exercise toward more continuous, capability-led decisions about whether to build, buy, borrow or bot a given skill.

Embedded actions. Inside GSK's single L&D hub, employees see and maintain their AI-inferred skills, pick the skills to focus on, and receive personalized learning, opportunities, and job recommendations, with an AI coach layered on top. Skills conversations are embedded into the mid-year and year-end check-ins that already matter to employees.

Skills inference cuts the time by 90% of the investment you'd make doing it the traditional way. I wouldn't imagine us doing this in 18 months without AI.”
— Zaka Farhat, GSK, on the Digital HR Leaders podcast

Outcomes: a foundation that funds itself and moves into planning

Adoption first

One of the clearest early outcomes is adoption. More than 80% of employees went into the learning hub, rated themselves on AI-inferred skills data, selected their focus skills, and started learning more than ever, with strong returning usage driven by the experience. Development conversations are now equipped with data they never had before.

More than 80% of our people went into the hub, rated themselves on their skills, selected their focus skills, and started learning more than ever.”
— Zaka Farhat, GSK, on the Digital HR Leaders podcast

A self-funding transformation

GSK retired legacy systems and rationalized its vendor portfolio significantly, cleaning up assignment profiles and unlocking millions of dollars. That money was reinvested into the new AI-powered ecosystem, and GSK still came out saving. A transformation that funds itself changes the conversation with finance, which makes it far easier to keep adding capability.

Having a self-funding transformation with the savings makes the conversation with finance easier, because we want to add more and more. This is not our end.”
— Zaka Farhat, GSK, on the Digital HR Leaders podcast

From learning to workforce planning

With the foundation in place, the highest-value use shifts from personalized learning to planning. Skill intelligence signals let GSK understand how work is evolving and where the gaps sit, replacing an annual headcount collation with continuous, capability-led decisions. Early signs include an uptake in internal hiring versus external. GSK tracks progress across three layers: foundation health, whether skills show up in decisions, and business outcome, and is candid that the deepest outcomes, skill growth, time to productivity, mobility and retention, will take longer to measure.

Governance: councils, standards, and a taxonomy built to evolve

To keep a single foundation trustworthy across a decentralized global business, GSK stood up governance early rather than bolting it on later.

• A global learning council and skills council. These centralize the employee experience in the L&D hub, keep the skills taxonomy agile rather than built once, and set minimum standards for the quality of learning that enters it.
• A standardized skills vocabulary. GSK moved the organization off scattered competency models toward a common capabilities-to-skills language, including the behavioral and cultural skills that define how the company works, not only technical ones.
• Agile by design, tolerant of imperfection. The foundation is built to evolve with internal and external signals. GSK chose to act on data that is roughly 80% right and continuously fine-tuned, rather than waiting for a perfect snapshot that would be stale on arrival.

The hard parts, in Zaka's words

• Decommissioning is the part most transformations skip. The conversation is almost always about the new tool, rarely about what to stop. Retiring systems that were embedded for years meant working through genuine emotional attachment, with a clear business case and leadership backing.
• Revamping the job architecture mid-flight. GSK assumed it could infer skills straight off the existing architecture and have everything fall into place. It had to go back and future-proof that architecture first, an unplanned but pivotal piece of rework.
• Outcomes take time. Zaka pushes back on anyone claiming a short-term transformation delivers business-level impact quickly. Adoption and savings are real now; skill growth, time to productivity, and retention will take longer to prove, and GSK is open about that.

Next steps: go deep, go broad, go agentic

Zaka frames what is next in three moves: go deep, go broad, and go agentic.

Go deep means optimization: rationalizing more content, rolling out AI-supported skill assessment, and unlocking the AI coach to support specific business capabilities. Go broad means extending beyond learning into internal mobility, hiring and talent, so that when a leader asks what happens if a capability grows by 20%, or what the risk profile of an organization looks like over three years, GSK can actually answer.

Go agentic is the move Zaka is most excited about. GSK is piloting a Workforce Planning Insights agent that helps HR business partners and leaders redesign their organization conversationally, drawing on internal and external signals, alongside AI coach and talent agents now in testing.

HR is becoming more and more a science and tech function in its own right. It's not a back office anymore.”
— Zaka Farhat, GSK, on the Digital HR Leaders podcast

Hear the full conversation between Zaka Farhat and David Green on the Digital HR Leaders podcast.