"If I could go back and revise the approach to this technology from the very beginning, I would not call it Artificial Intelligence. I would call it Augmented Intelligence."
Think of the early beginnings of AI, and IBM will come to mind: the builder of the computers used by some of the first AI researchers, and later the developer of Deep Blue and Watson. So it's hardly surprising that IBM has, from the start, been an early adopter of AI in the workplace – and the HR function was the first to implement AI, becoming 'client zero' in the company's AI-automation initiative.
People Matters met Nickle LaMoreaux, global CHRO of IBM who has been with the firm for a large part of its long association with the technology, during her recent visit to Singapore and asked for her perspective on why HR is so well placed to pioneer the use of AI for an organisation. Set against the context of IBM's decades of experience with the technology, here's her take on AI's evolution and its role in the world of people and work.
What led you to decide on HR as client zero for AI?
There are three main dynamics affecting the HR function in very real ways. One is cost pressure. HR has to optimise its operations as much as possible. The other is a very high standard for accuracy. You must get your paycheck correct, you must get your benefits correct, onboarding must be perfect. And the third thing, which accelerated during the pandemic, is that employees want consumer grade experiences. The experiences we have in our personal lives – shopping, booking a taxi, ordering food – they want to have a similar quality of experiences in the workplace. In our HR at IBM, all three of these pressures were converging on us, and we realised humans are always going to generate some level of error. And so this entire move was about how to make our HR department even more effective with technology and AI.
When you made the decision to implement AI in HR processes, where did you start?
We started with manual processes and high volume processes, of course. But we also started on processes where HR professionals were spending a lot of time to achieve accuracy, and we wanted those HR professionals to put that time into other, higher value efforts. A very specific example of one of the first things we put AI and automation into was transfers, where employees transfer internally and move to a new department and/or a new manager.
At IBM, most employees transfer to a new manager about once every three years. That adds up to 80,000 employee transfers a year, and managers do those transactions.
What we found is before we put in AI or automation, managers were spending on average 15 to 20 minutes on each of these transactions, and 14% of the time there was an error. They picked the wrong department to transfer, they picked the wrong effective date, or something even less easy to notice. And 7% of the time, they had to ask for help from an HR generalist. 7% times 80,000 transfers is over 5,000 transactions a year where our HR generalists are getting pulled in to answer questions or fix problems, instead of doing high value work like coaching managers and working on leadership development.
So we automated this process with AI and natural language processing, such that managers can now transfer an employee through a chatbot that asks all the right questions and checks the work for you.
Last year, all our employee transfers happened through the chatbot, and we had a zero error rate.
Because if a manager starts to give the wrong instructions, the AI tells them, and then it guides the manager to an accurate, acceptable decision.
And that creates a huge impact on the employee experience because it means that people get transferred in a timely way, without errors or issues. And it also means that my HR generalists can spend their time working on higher value work that they like.
Generative AI powering chatbots is fairly new compared to more established applications of automation – how do you view the progress of this technology?
As we move towards generative AI, the technology is becoming even more helpful, because it's more conversational and easier to use, and it's learning faster and adapts better to our needs. But in my view, it's really about the technology and the people coming together to drive an outcome.
I think a lot of times when we talk about AI in the workforce, we get worried that AI is replacing humans, and we miss the fact that particularly in HR, it's about augmenting humans. In fact, if I could go back and revise the approach to this technology from the very beginning, I would not call it Artificial Intelligence. I would call it Augmented Intelligence. Because we have found that in HR work, the magic really happens when our HR professionals, our managers, our employees, are working together with the AI.
Working with AI has the potential to be great but also to go greatly wrong. Coming from an organisation that pioneered the technology, how do you self-regulate its use?
How companies implement AI is really, really important.
At IBM, what we've done is put in place a set of principles to guide the way we approach AI. The first piece is, AI is never a decision maker. It may offer recommendations to a manager or an employee, but the human is always the decision maker. We think this is a very important principle to maintain.
The second principle is that we believe that the data AI uses is owned by the data's creator. So implementing AI is absolutely not about letting data out into the wild or spreading it everywhere. That's also very important to address the ethics and privacy issues which you hear about today.
A third principle is that AI must be transparent and explainable. So if AI is making a recommendation, you can ask it why, and it shows you all the data it used and why it's making that recommendation. We think that's really important not only to the end user – the managers and employees who get the recommendation – but also in HR, because as program owners, whether in compensation or in benefits or in payroll, we want to be able to monitor and watch what the AI is doing to avoid what's called drift (when the data changes and affects the accuracy of the output).
Also, the technology that underlies the AI must be secure. This is HR data we're talking about. We can't leave any possibility of cybersecurity attacks. And it must be robust, because these AI algorithms work best and come to the best conclusions when they receive a lot of data inputs.
That links back to why HR is client zero for AI. In IBM's case, we're using this technology across 280,000 employees, 171 countries, and that allows us to be a good test case.
You mentioned onboarding earlier. Can AI, especially generative AI, reduce the time to produce for a new hire?
When someone enters a company for the first time, or changes jobs within a company, there's always a learning curve. And one of the ways AI can help is by knowing your profile and surfacing to you the specific things that you need to know.
A good example of this is how we use AI in learning and training. We build roadmaps of what needs to happen if you want to be, say, an IT consultant. These roadmaps include the courses you need to take, the tests you need to pass, the certifications you need to acquire. And when you log into the system and say you want to be an IT consultant, the AI takes that roadmap and combines it with your profile to create a customised plan based on what you already know and what you still need.
Now this also works when you're new in a job. The AI will not waste time trying to teach you something you already know. Instead it will surface the information that you still need, improving the experience. And of course, you still have your team members and managers to help you in the areas where the AI can't. And so I do think it's very powerful in getting the workforce up to speed faster.
On the point about doing things faster – AI can be used to improve productivity by increasing speed, but it can also be used to replace head count. Where's the balance between the two?
I do think that when we talk about AI, too much of the narrative is worrying about head count replacement and lost jobs.
I think what's really going to happen is that in many companies, there are very, very few jobs that are purely administrative, purely transactional, where AI is going to replace them completely.
The bigger issue is that for 90-95% of all of us, myself included, our jobs are going to change. 10% or even more of my job might get replaced by AI. Some transactions I'm doing today will go away, and that does create the question of what I will use that time for. But there's also another 50% or so of my job that will still exist, but I'll be doing it with AI. And then the more important question is, do I have the skills to now do those same things with this technology that I've been doing manually or on spreadsheets or by with meeting with people?
Right there are the big shifts and changes that we all need to prepare for. And going back to the question of how to balance these two models, I think that some of the culture around supplementing productivity is really about the question: Do you know what portions of your workforce are creating incremental value? And what do you want them to do?
In HR, coaching is a big skill that we want professionals to have – the ability to give advice, to counsel, to do leadership coaching. That's what I want more of my HR professionals doing, not transactions. And so the question is, how can I help shift them? Do I have a clear point of view on what those high value skills are? Do I have a path to train them in it? I think that once AI starts to free up our time, more companies will be able to do that.
Now on the question of reducing head count, I agree that that is not the ideal path for AI, but maybe with one exception. There are countries around the world that are struggling with very low birth rates, including some right here in the Asia Pacific. Governments are predicting massive labour shortages out into the future. I actually think this is going to make the existing workforce even more important, because your current population isn't going to be able to sustain labour needs in the future. So while reducing head count might be a bad outcome in some places, in other places, it's actually going to allow the economy and the existing workforce to continue to thrive.
The million dollar question: with all this change coming, what can we do to prepare HR professionals?
I think HR professionals are going to have to play two very important roles here. First, they're going to have to think about how AI might be able to enhance their own jobs, and how they are going to develop the skills for it. Maybe 10 years ago, we would have said all HR professionals need data analytics skills; I think AI is now replacing that. I would go so far as to say all HR professionals, regardless of their domain, need to start learning a little bit about AI. Some might decide to code, others might go into how natural language processing works, still others might be heading programmes and have to think about where to fit AI into the process.
And the second really important role for HR is to prepare the rest of the organisation. I've talked about how everybody's job is changing. I think HR needs to lead the entire organisation, from finance to sales to product development, and really helping those functions think about what's going to be the impact of AI. How will they train people? How will they onboard people differently? How will they forecast what type of roles they need in the future? HR plays a really critical role in getting all these in place, and will be the primary value driver of getting this right in organisations.