Dan Pell is the general manager of Workday in the UK and Ireland, with a background at Salesforce and Microsoft, and a law degree in his hinterland.
At the recent London edition of the HR software-as-a-service supplier’s Elevate customer conference, he gave the interview that follows, and which majors on how digital labour – agentic and human – can and should be managed. It has been edited and supplemented by comments he gave in a media panel at the same event.
Something that has struck me amid all the noise about the rise of agentic artificial intelligence (AI) has been the question of how organisations will do the identity and access management of these agents. They are non-human identities – like machines or workloads – but they are also, by definition, autonomous, and could be subverted by hackers, for example. How is Workday thinking about this problem?
We often say we manage two assets in any organisation: people and money. And now we talk about managing agents as well. Because they are a digital workforce. They are going to have skills, roles and identity permissions.
That becomes really interesting. So, we are building an agent system of record. Think about how we manage people – we understand your permissions, your role, your hierarchy, your pay, everything about you. So you have a profile, which gives you access to certain things within the organisation, and certain privileges, and so on.
The same is going to apply to agents. I think it will be an interesting space, and we are right at the beginning of this right now. But we think there will be this digital workforce that we naturally think would sit in a Workday [instance] as a place to manage, control and understand what they’re doing and how they’re behaving.
And, in some sense, will they get promoted? Will they become bigger agents, or will they stick to a certain bit of functionality? It is an interesting dynamic, and we are right at the beginning of this. I think everybody is.
It is a very interesting space, and I think you’ll see us and Salesforce collaborate and, potentially, we’ll manage lots of different agents on our platform for different organisations because we have that way of understanding how to manage people, and we can apply that to managing agents who will be your digital workers.
“We are the last generation of leaders who will only manage physical people. As agents become more prevalent in the organisation, as leaders, we’re going to have a role in how we manage those agents”
Dan Pell, Workday
We are the last generation of leaders who will only manage physical people.
It’s natural that, as agents become more prevalent in the organisation, as leaders, we’re going to have a role in how we manage those agents. You might be managing a legal department, dealing with contracts, and you will have contract agents. They’ll do part of your work for you. They are going to augment your workforce. They will have certain skills, and in some ways, you can look at how those skills augment the skills within your business.
Let’s say you are lacking a certain skillset, let’s say literacy or creative skillsets, you can augment your existing workforce with a set of agents to have those skills, rather than train your workforce. So, leave them as they are. Give them the skills to work with the agents, but the agents have the skills. That is really interesting. You are adding to the skills matrix of your business, not necessarily by changing the fundamental skillset of your people, but you’re augmenting them.
All right. You studied law. I know you didn’t practice, but let’s say you did. There is a lot of grunt work in law, especially as a trainee, that is formative. It is the same with journalism. You typically do a lot of basic press release filleting before you get to the higher-order stuff. Doesn’t AI take that away?
I think AI, particularly generative AI, gives you a good starting point. But it’s definitely not the final outcome. This is why we always talk about humans in the loop with AI. You can find sources of information and summarise those, but you still have to read and process them. You need to apply your creativity to the material. You need to pull out those bits of information that are relevant.
It is something we’re going to have to think about. When we look at the skills organisations are still looking for, digital skills come up number one. Using software and AI, project management and planning will still be very important. But then, social skills, communication, networking, problem solving, conflict resolution, emotional intelligence – these will be more important than ever.
With AI, we will have an IQ of 200, apparently, but with all those skills, such as interrogating data, there is no quick fix. That’s got to stay, and it would worry me if we were skipping all that because the quality of the content in the brain and the learning is going to have disappeared.
Finally, I am curious to hear what you think about the shift from generative to agentic AI that has been spoken about a lot in the market over the past six months. Especially what your customers and the leaders at your customer organisations feel and think about this evolution. It seems to me that people could see the value of generative AI so easily and intuitively, but it’s more obscure with agentic AI. For most senior executives, AI will mean generative. How do you see the difference?
In our world, we have a Workday assistant, which you can ask questions in natural language about, say, how many days’ holiday you have to take. People love it, it’s super simple.
With GenAI, creating job descriptions is an obvious use case. But that is just the starting point. The job description agent knows that you want to hire, it’s got the budget and headcount information, it can start the process. It can look for people internally in the business with the skills and notify them. Let’s say people with three years’ tenure who are probably disengaging.
Likewise, you can go externally and notify people, messaging them. Equally, if you are doing a job description for a very senior executive, it won’t do any of that. It will have learned the thresholds, and it learns based on organisational knowledge. That is the difference.