Workday has launched Workday AI Research, a dedicated AI research team focused on developing reliable, trustworthy, and efficient AI for the enterprise.
The new team is tasked with addressing some of the technical challenges emerging around AI agents. Some of its researchers have already produced work accepted by major AI research conferences.
The launch comes as enterprises enter a more challenging phase of AI adoption, with businesses increasingly looking beyond experimentation and asking what is required to deploy AI agents reliably at scale.
Workday Targets Enterprise AI’s Toughest Problems
Workday said its research will focus on technical problems that become increasingly important as AI agents gain the ability to retain context, make decisions, and act on behalf of employees.
One area of research is agent memory. Workday researchers developed a more selective approach designed to retain useful information while filtering out outdated, duplicated, or unreliable details. The method delivered 12% higher precision and around 8% higher overall memory quality in testing, while retaining 97% of the memories judged important. It also operated around 31% faster than the AI-based comparison used in the study.
The research also examines whether multiple specialized agents can produce better outcomes than a single agent. Workday found that dividing complex tasks between agents with different responsibilities improved accuracy by 5.8%, while ensuring that every final answer in the study met its defined constraints.
The findings also raise questions about how enterprises will govern information held by AI agents. Workday investigated whether an AI agent can genuinely forget information when instructed to do so, finding that information remained recoverable from an old summary around one in five times after the original memory had been deleted. Completely removing the information required the summaries containing it to be deleted as well.
To fuel further research, Workday is establishing a PhD fellowship to deepen its links with academia. The program will provide $50,000 in annual research funding through an unrestricted university gift, alongside mentorship from Workday researchers and opportunities to collaborate with the company.
AI Vendors Face a New Adoption Problem
The launch comes at a difficult point for the enterprise AI market. Businesses and technology providers are increasingly focused on whether AI agents can deliver enough value, reliability, and control to support wider deployment.
Gartner’s forecast that more than 40% of agentic AI projects could be canceled by the end of 2027 highlights the uncertainty surrounding the market. The research firm attributes the cancellations to escalating costs, unclear business value, and inadequate risk controls.
That creates a problem for the companies selling enterprise AI. The commercial opportunity depends not simply on organizations experimenting with AI, but on them deploying it deeply enough to generate sustained value. For Workday, which is increasingly positioning AI agents across its HR and finance platform, making those systems reliable enough for enterprise use could therefore be critical to turning AI interest into sustained adoption.
Megan Barker, Senior Manager, Talent Acquisition at Workday, said: “In areas like HR and finance, enterprise AI needs total trust and precision, but off-the-shelf models simply aren’t built to solve challenges like privacy, auditability, efficiency, and accuracy.”
Continuing, Barker said that Workday AI Research was a direct response to that:
“That’s why we launched Workday AI Research, tackling the toughest technical hurdles, from persistent agent memory to multi-agent collaboration and explainability, all backed by peer-reviewed science.”
The focus of Workday’s new research team therefore suggests that the next phase of enterprise AI adoption may depend less on whether models can become more capable and more on whether businesses can trust them with consequential work. The company is effectively investing in the technical foundations needed to close the gap between AI experimentation and production deployment.
Workday Looks to Build Trust Into Enterprise AI
Workday’s announcement represents a shift in emphasis for enterprise AI. Rather than focusing solely on what AI models can accomplish, the company is putting research resources behind the problems that determine whether those capabilities can operate safely and consistently inside business systems.
That distinction is becoming increasingly important as AI agents move closer to making decisions and taking actions rather than simply generating responses. Gartner has separately predicted that by 2027, 40% of enterprises will demote or decommission autonomous AI agents because of governance gaps discovered after production incidents.
Workday’s research agenda reflects many of those concerns, from controlling what agents remember to coordinating multiple agents and ensuring outputs remain within defined constraints. The launch also gives Workday a way to build expertise around the technical problems that could ultimately determine how widely its AI products are deployed.
As enterprises become more selective about which AI projects make it into production, solving those problems could become increasingly important to vendors competing for enterprise AI spending. For Workday, the research effort is therefore both a technical investment and a bet on the next stage of enterprise AI.
