AI will fade, says Dell Technologies' ephemeral Chief AI Officer

You're not alone in thinking that the long march of progress may be working you out of a job. Even a senior executive at one of the world's top technology companies sees at least part of his employment on the line, and John Roese in fact welcomes it. "The Chief AI Officer job is time bounded," he said, "There will come a point where having a Chief AI Officer makes about as much sense as having a Chief Internet Officer."

Roese is the CTO and, yes, also Chief AI Officer at Dell Technologies, and his interpretation of the role of those tasked with corporate AI adoption is noteworthy. It's more about the nature of work itself than it is about the technology, which becomes clear as he explained how Dell itself has undergone (or, perhaps more appropriately given he is still in a job he sees as temporary, is undergoing) at-scale agentic AI introduction.

In executing this mandate, Roese has applied engineering principles along with economic and social ones. He's also abundantly clear on what value creation looks like, and unafraid of shying away from skills and workforce changes, including those directly impacting his own employment. Most of all, Roese talks about jobs as adding value, rather than being worthwhile in and of themselves, while acknowledging the continual evolution of the nature of work.

Why every company should have an Chief AI Officer (for a while)

"Michael [Dell] and I have talked about this. AI is not a bottom up exercise. It's a top down transformation, and when you bring a new paradigm to an organisation, a named person in a senior role is essential. You can call that a Chief AI officer [but] sometimes it's the chief digital officer or the CIO. The absence of that is an enormous problem. [You need executive] oversight and governance to get into the AI era, but I hope it's time bound, and that in two or three years, we won't do that anymore."

One part of the point he is making is that AI is likely to become as pervasive and second nature as the internet is. The other is that many jobs come and go, and have done so throughout history. Technological change shifts and shapes employment; economist Henry Hazlitt forcefully made the point in the 1940s in 'The Curse of Machinery'. "So I have a job in which my job is to put myself out of a job," Roese quipped.

How Dell does it

Getting into how Dell Technologies, which considers itself 'customer zero' for AI adoption, has approached the task provides a handy blueprint for any organisation to consider. And Roese is the man who has spearheaded it (he said the company is considering publishing its methodology as a whitepaper).

Starting at first principles, Roese stressed the necessity of understanding basics which have little to do with technology, starting with the difference between jobs, work, and tasks. 

Precision in language matters; programmers know it, engineers know it, and communicators should know it, too. "Think about the effect of agents. What do they actually do? They are a vehicle that allows you to transfer work humans are currently doing, to the machine layer. That is different from a job," Roese explained.

Semantics? Maybe. But with a material impact on outcomes.

Containerisation of human effort

Roese said a job is a container of work performed by a person, while work consists of a series of tasks, and each task is an action. "Within a job, there are many types of work [and within work, multiple tasks]. Some of it you like doing, and a bunch of work you don't like doing. So there are different kinds of work. When you do agentic AI properly, you start from this basis that the agent doesn't do a job. The agent does work. And human beings do jobs that have multiple types of work."

This basis aligns with the principle of focusing AI efforts on the undesirable parts of any job. Roese described it as 'decoding which specific work you want to extract from the jobs, and moving those into the machine layer'. "That is the essence of an agentic strategy."

In Dell Technologies' case, this meant exhaustive analysis of 6,400 jobs, and 'decomposition' into the types of work (productivity, hygiene, coordination, expert, and human element) performed by humans and agents.

The result?

"Every job had work that agents could probably do, but no job could be completely replaced with a single agent," said Roese. "In fact, if we removed the work that agents could do from the job, the job itself didn't go away. But the person doing the job had more space for something else, more of the work they want to be doing."

That applied for people in Dell's lines of business as much as it did to the captains of the company's private jet fleet; detailed analysis showed that just 38% of a pilot's activities is expert work. "The rest is something else," said Roese.

It is in that 'something else' that AI could (or indeed perhaps, should) play a supporting role. "Get clarity on this. Work with your HR team and get all your job descriptions, your job postings, and go through and do this classification…which will not be 100% accurate. We were about 80% in general and that led us to a set of conclusions that are now guiding us going forward."

He said companies should think, and think carefully and clearly, at this level. "Stop using words like 'job' and 'work' interchangeably. Jobs are a container for different kinds of work. It sounds like marketing, it sounds like sociology, but it's so essential because the actual technology, the agents, are executing tasks and doing work. They are not doing jobs."

Being very deliberate about that relationship, said Roese, is where value starts to emerge. "What ends up happening when you apply agents properly is the hygiene and coordination work, which is not the exciting stuff, starts to shift below the machine line. Space is created for expert and human element work."

The future is bright…for everyone

Finally, Roese is confidently cheerful about eliminating at least part of his responsibilities, but has the obvious insulation of a career at the highest levels. What of the next generation, particularly those who imagine a jobless future?

He's highly sympathetic: "In no uncertain terms, this demographic is incredibly nervous," he said. "They are being told there are no jobs. That is not true, but they are nervous, they are concerned. I think we've been communicating the impact of AI on society in extreme ways that are not based on anything real, and it is causing damage. All jobs change, as they always have."

One need only look to the past, including reference to Hazlitt's freely available 'The Curse of Machinery', for confirmation. After all, one of the consistent features of the introduction of any revolutionary new technology is growth in the quality and quantity of jobs, and an improvement in the relative wellbeing of the species as a whole. This does, inevitably, come at a cost: as jobs change, some people are displaced.

Roese offers a comforting parting shot. "If the two things that matter in the future are deep expertise and human element capability, then what should you be doing? Get yourself a mentor who has those skills. Develop your interpersonal skills as fast as possible. That's 'the' marketable skill and that is the kind of person that will have a job in the future."