HCM Customers Set Conditions for AI Agents

Fusion Claw and Oracle’s agentic applications took center stage at Oracle AI World in Las Vegas this week, but four customers said organizations need to address business processes, data and legal approval before adopting AI agents in human capital management (HCM), the software used to manage workforce functions. The customers—IHG Hotels & Resorts, UL Solutions, Deutsche Bahn and Acosta Group—are at different stages of their HCM strategies and have not all begun introducing agentic AI.

IHG deployed Oracle Cloud HCM in 2022 and has since been moving to Oracle’s Redwood user experience. Matt McDonald, IHG’s director of HR technology, said the company plans to start with talent acquisition, which its leadership identified as an area for major change from 2027 onward. Deutsche Bahn went live with Oracle Cloud HCM two years ago after leaving a heavily customized PeopleSoft system; Daniel Passarge, head of HR IT systems, described the implementation as both an IT and business transformation that included a new HR operating model. UL Solutions has used its system for three years and is adding Oracle Learning Cloud, Oracle Grow and Oracle Dynamic Skills. Its HRIS and analytics director, Kelli Scroggin, said the company is reviewing earlier implementation decisions and data gaps as it expands into other modules and prepares for AI and other business initiatives. Acosta Group has adopted Oracle Recruiting Cloud and AI-assisted and generative tools, and is now moving into core HR, compensation and benefits.

The panel’s central lesson was that AI can expose weak processes. IHG had to narrow a long list of proposed AI tools to five core areas and revisit how recruiting worked across regions. McDonald said the company is first addressing missing interview scheduling with Oracle Recruiting Booster, then considering an interview assistant agent. Scroggin said UL Solutions wants faster employee survey follow-up, fewer approvals and fewer handoffs before adding AI, rather than carrying those process problems into automated systems.

Legal review is another prerequisite, and IHG assesses use cases by their intent and potential consequences. McDonald said legal approved an agent for interview scheduling but asked how a tool called Talent Best Match would work and where human oversight would sit. Deutsche Bahn operates under EU data and AI regulations, which Passarge said add time and effort, while the company’s works councils are asking when AI will arrive. He said the rules also make the company examine consequences, build a technical basis it can explain, and plan carefully. Acosta Group’s Doug Johnivan advised involving legal teams from day one, citing extended discussions over AI language in job postings.

Scroggin cautioned that deploying Oracle agents requires substantial testing and possible debugging, and recommended working with Oracle partners for guidance. She said UL Solutions is still trying to understand costs: token use varies with the agent and the large language model (LLM) it calls, and testing agents also consumes tokens. Deutsche Bahn began cleansing and reviewing data before implementation and continued through go-live, Passarge said, recommending shared responsibility between business and IT. Scroggin said keeping data within one technical landscape helps UL Solutions add modules and functionality; fragmented data makes that harder.

Fusion Claw and Oracle’s agentic applications took center stage at Oracle AI World in Las Vegas this week, but four customers said organizations need to address business processes, data and legal approval before adopting AI agents in human capital management (HCM), the software used to manage workforce functions. The customers—IHG Hotels & Resorts, UL Solutions, Deutsche Bahn and Acosta Group—are at different stages of their HCM strategies and have not all begun introducing agentic AI.

IHG deployed Oracle Cloud HCM in 2022 and has since been moving to Oracle’s Redwood user experience. Matt McDonald, IHG’s director of HR technology, said the company plans to start with talent acquisition, which its leadership identified as an area for major change from 2027 onward. Deutsche Bahn went live with Oracle Cloud HCM two years ago after leaving a heavily customized PeopleSoft system; Daniel Passarge, head of HR IT systems, described the implementation as both an IT and business transformation that included a new HR operating model. UL Solutions has used its system for three years and is adding Oracle Learning Cloud, Oracle Grow and Oracle Dynamic Skills. Its HRIS and analytics director, Kelli Scroggin, said the company is reviewing earlier implementation decisions and data gaps as it expands into other modules and prepares for AI and other business initiatives. Acosta Group has adopted Oracle Recruiting Cloud and AI-assisted and generative tools, and is now moving into core HR, compensation and benefits.

The panel’s central lesson was that AI can expose weak processes. IHG had to narrow a long list of proposed AI tools to five core areas and revisit how recruiting worked across regions. McDonald said the company is first addressing missing interview scheduling with Oracle Recruiting Booster, then considering an interview assistant agent. Scroggin said UL Solutions wants faster employee survey follow-up, fewer approvals and fewer handoffs before adding AI, rather than carrying those process problems into automated systems.

Legal review is another prerequisite, and IHG assesses use cases by their intent and potential consequences. McDonald said legal approved an agent for interview scheduling but asked how a tool called Talent Best Match would work and where human oversight would sit. Deutsche Bahn operates under EU data and AI regulations, which Passarge said add time and effort, while the company’s works councils are asking when AI will arrive. He said the rules also make the company examine consequences, build a technical basis it can explain, and plan carefully. Acosta Group’s Doug Johnivan advised involving legal teams from day one, citing extended discussions over AI language in job postings.

Scroggin cautioned that deploying Oracle agents requires substantial testing and possible debugging, and recommended working with Oracle partners for guidance. She said UL Solutions is still trying to understand costs: token use varies with the agent and the large language model (LLM) it calls, and testing agents also consumes tokens. Deutsche Bahn began cleansing and reviewing data before implementation and continued through go-live, Passarge said, recommending shared responsibility between business and IT. Scroggin said keeping data within one technical landscape helps UL Solutions add modules and functionality; fragmented data makes that harder.