Albemarle, a global chemicals and materials company, started its AI transformation with the goal of solving one problem: the current process for evaluating and weighting compensation was overly manual and too time-consuming.
In our webinar this week, Ani Huang sat down with Autumn Gagarinas, Chief People and Workforce Transformation Officer, to learn how her team built one AI agent called Comp-Pilot and then scaled it across the enterprise.
How it worked: Comp-Pilot was built in-house and trained on the company’s own compensation philosophy, internal equity data, and external market benchmarks. This step was critical because fairness, internal equity, bias, and compliance carry real stakes in compensation. The agent evaluates the data, then generates an offer recommendation.
Human oversight accelerated adoption: Comp-Pilot handles repeatable pricing and policy logic, while Total Rewards and HR professionals review recommendations, apply judgment and override the agent where appropriate. Humans retain final decision authority. “Building that organizational mindset and culture of acceptance was as important as building the agent itself,” said Gagarinas.
By the numbers:
-
Time spent to calculate offers reduced from days to minutes.
-
10%–20% routine work savings: Total Rewards capacity was freed for higher-value work (workforce planning, capability modeling and executive decision support).
-
Manager satisfaction increased: Hiring managers gained faster answers and more visibility into the reasoning behind compensation recommendations.
-
Global consistency improved: The same compensation logic can now be applied more consistently across roles and geographies.
What’s next: Comp-Pilot has already expanded from new-hire offers into promotions. Albemarle is now pursuing deeper Workday integration and is planning for a multi-agent Total Rewards ecosystem that connects compensation with broader end-to-end employee workflows.
Takeaway: Scaling AI may be less about launching more tools and more about creating a repeatable operating model for how humans and agents work together.
