Workday launches a 'system of record' for AI agents - is this HR or IT?

Are AI agents part of the workforce or are they just another IT asset? This week, Workday, not previously known as an IT service management vendor, weighed into the 'digital labor' debate with the launch of the Workday Agent System of Record. This is billed as a tool that helps organizations "manage their entire fleet of AI agents in one place." Does this mean AI agents now come under HR? Workday seems to think so, saying that it "is stepping in to provide a trusted, centralized way to manage digital and human workforces together."

Workday also launched a further four role-based agents to add to the initial four unveiled at last year's Rising conference. These sit firmly in Workday's traditional domains of HR and Finance, and comprise a contracts agent that monitors contracts, a payroll agent that carries out various payroll actions, a financial auditing agent for enterprises and their audit firms, and a policy agent that keeps employees and managers up-to-date on a company's corporate policies.

But the main focus was on the Agent System of Record, which Workday sees enterprises using to manage not only its own agents but also those of other vendors across their entire lifecycle. Its functions include processes for defining the roles and responsibilities of new agents as they are onboarded, controls to turn them on and off, monitoring and analytics to track their impact and optimize the cost of developing, deploying and maintaining them, governance procedures to support data privacy, bias, access control and other compliance factors, and feedback to foster continuous improvement.

IT service management or part of the workforce?

While it's obvious that agents will need managing, Workday isn't the most obvious choice of vendor to provide that capability. As part of the IT estate, it might seem more natural to look to established IT service management vendors such as ServiceNow, which launched its own agent strategy last month; or integration vendors such as Boomi, which is developing its own agent registry for release in the spring; or some of the up-and-coming pureplay AI governance players such as Holistic AI or Credo AI.

But Workday argues that, since AI agents will be taking on responsibilities that have previously been the preserve of human workers, there are common processes where it does have experience, such as onboarding, training, compliance, security, access controls, and other aspects of workforce management. AI agents may not be human, but as they take on roles that interact with customers, partners or employees and their data, this 'digital workforce' will have to adhere to the same ethical standards, company values, codes of conduct, compliance rules and performance expectations as their human co-workers. As David Somers, Workday’s Chief Product Officer, explains in a Workday blog post accompanying this week's announcement:

Accuracy, consistency, bias, speed, and cost are all things we need to consider carefully when implementing agentic AI within organizations. That's why it's so important to set clear boundaries and rules for how they operate. Just like we wouldn't give a new employee free rein over sensitive systems, we need to control what our AI agents can access and do.

Many companies today lack a central way to do this, which is a big risk. That's where an agent system of record comes in. This acts as a central AI system of record for governance, management, auditing, monitoring, and role-based access controls, ensuring AI agents are used responsibly and efficiently.

He also notes that Workday's Skills Cloud already has a detailed mapping of the skills that make up each of the roles employees perform throughout an enterprise, which has helped inform the creation of agents to perform certain aspects of those roles. He writes:

These role-based agents may start with a skill and they will, over time, have many configurable skills. Role-based agents are more valuable than task-based agents because they mirror the multifaceted nature of actual roles, going beyond simple task completion.

He concludes:

[D]igital workers are now joining human workers as one combined workforce, and we need a unified platform to manage them as one. This is a fundamental shift, and companies that don’t embrace managing this complexity will fall behind. Workday is uniquely positioned to lead this transition. We understand human capital management, skills, and roles better than anyone; managing digital labor is a natural extension.

My take

I really don't buy this argument. Yes, AI agents need careful management and governance. But that's a job for IT, not HR. No one ever argued that bank CHROs should have ATMs under their wing, or that Chief People Officers in the retail industry should have ultimate responsibility for the security and operation of self-service checkouts. Why does AI, simply because it's a novel digital technology that people often anthropomorphize, suddenly need to be treated differently than any other machine that takes away people's jobs?

So what possessed Workday to get into the IT service management field with its Agent System of Record? Here's why this move makes sense for Workday:

• AI agents with the wide-ranging capabilities we're discussing here are a new phenomenon. Therefore there's as yet no established, commonly accepted way to manage them, or to contain the significant risks they pose if they aren't properly controlled. Workday clearly recognizes these risks and is acting responsibly in offering a solution.
• In their ability to act on behalf of humans, these agents go further than other types of automation and require more extensive controls. Since those controls already exist for humans when carrying out these actions, it makes sense to adapt the same controls as guardrails for agents. Workday already has the infrastructure in place for controls such as access rights and permissions, spending limits, performance expectations, etc, and has simply adapted it to work with agents. But as IT assets, agents may need other controls that would not be applicable to human workers.
• In incorporating agent capabilities into its platform, Workday has had to create a framework for managing agents and connecting them safely into its AI assistant interface — not only its own agents, but also third-party agents that connect into its platform, as well as agents that customers and partners build using its Extend development platform. Since Workday had to build this anyway, why not market it as a tool for managing every agent across an enterprise?

All of this makes far better sense than arguing that AI agents should be treated as another segment of the enterprise workforce alongside permanent employees and contingent workers — that AI should be employed, not deployed. What Workday has built is actually an IT service management tool and should be under the control of the IT function. At the same time, the IT and HR teams need to collaborate closely to ensure that the controls and guardrails that originate from the people management domain are correctly applied when managing digital agents. But HR's primary focus remains keeping an organization's people happy and ensuring they function at maximum effectiveness — they'll have plenty to do on that score, which we'll come to in a moment — rather than getting distracted by attempting to manage something that is outside of their domain expertise.

One other aspect of the HR domain where the introduction of agents will have an important impact is in workforce planning. As Somers signals in his comments noted above, the Workday Skills Cloud has prepared the ground here by creating a map of skills and roles across an enterprise that could help HR leaders identify which jobs are most likely to be impacted by the introduction of role-based agents. Julie Sweet, CEO of Skills Cloud user Accenture, spoke at Davos about this process at her own company:

I have a database of almost 800,000 people and their skills. We're systematically re-defining the skills needed at Accenture - who needs gen AI, what kinds of technology? - but also, as we are replacing some of the things that they're doing with gen AI, we're able to identify who could be upskilled.

Some observers noted the irony of this week's announcement from Workday celebrating the growth of digital labor, after last week's news that Workday is laying off 1,750 employees, amounting to 8.5% of its workforce, in a move linked to AI investments. But Carl Eschenbach, Workday's CEO, told Fortune this week that there was no link between the two announcements, saying:

This is a restructuring of our company to make sure we’re aligning our workforce at Workday around the biggest opportunities... It’s about restructuring to get more dollars to spend on innovation and meet the customer demand.

Nevertheless, it seems clear that widespread deployment of agents will cause many employees at Workday and elsewhere to reconsider their career plans. Some may themselves find it useful to consult the Workday Skills Cloud platform, which the vendor only recently revealed it is beefing up for its own employees in collaboration with TechWolf, an AI-powered skills intelligence tool. The wider picture is that workforces are going to experience a lot more turbulence and HR teams will have their work cut out just managing all of that, without having to add responsibility for the growing population of AI agents at the same time.

The other takeaway here is that agents are coming and CIOs face the novel challenge of managing them. There will be no shortage of vendors offering to help with that task — Workday is one among many that sees an opportunity to either expand its footprint or defend its turf, depending on your point of view. It is not only people's jobs that are going to be reconfigured or replaced by the advent of agent technology. Enterprise technology vendors face a challenging time too and are already jockeying for position in this new world.