IBM: ‘AI agents will change your job – but we want you to lead that change’

When AI is introduced to a workplace, it sparks excitement for some, and anxiety for others.

While some employees may see technology as a means of enhancing their ability to complete their role – improving productivity and reducing repetitive tasks – others may see it as a step towards making their job obsolete.

At IBM, the answer is neither simple nor threatening because the company is putting its people at the heart of its AI strategy. 

In an exclusive conversation with IBM’s Jon Lester, Vice President of HR Technology, Data & AI, UNLEASH explores how the technology giant is using AI agents not just to automate tasks, but to empower its 270,300 employees to learn, adapt, and thrive in the future of work.

Placing employees at the heart of AI transformation

For Lester, “openness and transparency” is the cornerstone of ensuring employees feel empowered by emerging technologies. 

However, HR leaders also need to demonstrate a solid understanding of how these technologies work, to then confidently translate this to the workforce.

Lester highlights that, when used correctly, agents can bring the workforce a number of benefits, such as increasing productivity and freeing up employees’ time. However, it can also raise the question: What will happen to us when new technology comes in? 

“One of the big things we talk about is that as a headline number, IBM has given back $3.5 billion in the last two years of productivity back to IBMers. We think agents will accelerate that even more,” he shares.

“This can feel threatening. But at IBM, the message is clear: your job will change – but we want you to lead that change.”

Agents can therefore be used to complete tasks that employees tend to find undesirable – such as moving data around and responding to the same queries multiple times a day.

This is where transparency comes into play, as Lester insists leaders must make their workforce aware that “constant change is going to happen”, but that they should assure the business wants to “invest in the workforce that lead it”. 

He says: “This is a really important first step from our leadership – putting them in the middle of the change and investing in people to give them time to rethink and retrain.”

A practical way IBM has made this concept a reality is by creating AI solutions to make learning easier.

In fact, in 2016 the average IBM employee spent 31 hours a year learning new skills. In 2024, that figure had almost tripled to 87 hours. Part of this change has been from shifting the approach from ‘it’s nice to learn a new skill’ to ‘it’s actually on the individual – but the company will facilitate it’. 

Lester highlights that this approach has played a “big part” in creating new roles that didn’t exist a few years ago, but which “have now evolved in every single part of the organization”. 

When I look internally at HR, 80% of the new roles that have emerged are HR people learning completely different skill sets, always based on their core capability, like great employee experience,” Lester reflects. 

“We are seeing traditional roles go through this, and we’ll see more roles change with Gen AI.

“But people will evolve into roles that we probably don’t even imagine today, and I think that’s the bit that we have to make sure we, as an organization and as a leadership, are very clear around. There are going to be some great opportunities.”

Integrating LLMs into the workforce

The rise of AI agents will force a rethink of how work is done. Unlike traditional assistants that guide people through predefined processes, agents can deliver task-driven outcomes end-to-end – changing what’s possible.

By implementing AI agents, IBM has tangibly improved its HR processes, for example, through the development of the concept Ask HR – a digital employee support capability. All FAQs and policies were moved to Ask HR, meaning employees could engage through the chat capability to have any questions answered.  

Employees can also transact through it. It becomes an open API, so they can link to the underlying systems and do a job transfer or book a day off, for example, simply by chatting to the digital assistant. 

As this was traditional AI – natural language processing – IBM has now shifted the assistant to LLMs, meaning it’s evolved from an assistant to an agent.

Now, when an employee prompts Ask HR, an agent will classify what they’re asking for and will route them to a specific domain within the HR content. The domain could be compensation, payroll, or learning. 

When this was set, Lester highlighted that they then needed to ensure it was compliant, admitting “we’ve learned that the hard way at times”. 

To achieve this, IBM created individual domains, “each with their own LLM”, to get 90% compliance. 

“‘Chunking’ your content up into smaller groups enables you to go quicker.

“But the agent at the top then has to understand what you’re asking, to then send you to the right domain with the particular version of the LLM to generate the right answer. That took us a while to set up,” he explains.

Now, HR is split into 27 of these domains, with Lester boasting IBM has “very, very good compliance within the responses”. 

Content that is created for AI-ingestion has also been a key learning for Lester and his team. 

“When we first tried to ingest our 7,000 policies, only half of them could actually be ingested into a LLM because we hadn’t figured out what makes good content for AI to read,” he shares.

“We have now worked out how to do this and have built a template that all policy documents must follow and these can now be automatically ingested.”  

Lester explains that although this is a common issue for teams, it also highlights the conversation around the convergence of HR and IT.

Most content within an organization will never be consumed by a human anymore,” he says. “It will be consumed by an agent.

“You have to therefore shift the way you’re designing your policies and your content for agents to read and ingest it into what we call a vectorized database.”

Finding the right HR levers to pull

By speaking with Lester, it was clear that his one key take away message was: educate yourself. In fact, he believes that it is down to the individual to “understand the art of the possible around what AI can do today.” 

He adds: “You’ve got to understand LLMs, you’ve got to be able to engage with them, and you’ve got to be able to start to understand the implications of this on what it is that you do and what you’re a specialist at.” 

From here, he advises leaders to “figure out the value that this can bring”. To help IBM employees achieve this, leaders promote the mantra: eliminate, simplify, automate. 

This stops employees from “doing work that adds no value”, while “simplifying the rest, and automating a great process” to “build a value case”. 

“Start small, but just go for it,” he notes. “Probably one of the biggest things we’ve shifted in IBM is we’re removing the stigma of failure – allowing employees to fail.

Hopefully fail fast and fail cheap, but fail and then learn, iterate, and then succeed hugely.”

Lester also reiterates the importance of implementing agents and Gen AI that is “relevant to the business.” 

“Make sure it’s relevant to your function, to you, your team, and to your work, because I think we spend a lot of time doing stuff that doesn’t have any relevance.”

Concluding, Lester highlights the importance of relevancy, as he doesn’t “think every company has that”. 

“I don’t think we’ve [IBM] always had relevancy,” he admits. “I think we sometimes try things and realize it was interesting – but not relevant – and continued with it. We’ve then had to learn to stop doing it.

“Having those checks and balances as a leader, in particular an HR leader, is really important to a value case. Everything we’ve done before has enabled us to adapt to agents.”

When HR leaders are faced with a problem, they therefore shouldn’t just “throw the tech at it”, but rather work to “figure out the relevance of the value and the transformation, not just the transference”. 

“People just transfer a bad process onto different tech, and they just waste their money,” Lester says.

Understand there are all these levers to pull. If you pull all the levers, you can change work. You can reinvent work, and then you get the relevancy.”