share on
While it can take away the administrative aspects of certain tasks, it can't bring out empathy in your messaging. "Essentially, you can make it more human – but it isn’t human," shares the HR veteran.
There have been plenty of buzz and conversations in recent months, around ChatGPT – the AI chatbot that was launched in November 2022, and has piqued the interest of millions worldwide.
While some have found it handy in generating ideas and content at work, others have turned to it to tackle more administrative tasks, and as a whole, for personal planning.
For Fermin Diez, Adjunct Professor at National University of Singapore, it has come in handy in more ways than one, and in his (and our!) personal favourite: writing a song to commemorate his time conquering Mount Kilimanjaro! He reflects: "My first time using it was when I had come down from Mount Kilimanjaro. That evening, we were back in the hotel after eight days on the mountain, having a drink and playing around with ChatGPT.
"So I said: 'Write me a song about conquering Mount Kilimanjaro, in the style of Bob Marley, and in the style of Queen.' It came back, and it was pretty accurate in how it absorbed the styles!"
On the work front, the tool has also proven useful in tackling various HR tasks, in Diez's most recent role as Group Director, Sector Capability and Transformation at NCSS – for instance, performance appraisals.
In this interview with Priya Sunil, he tackles the theme of: What exactly does ChatGPT mean to HR - what can it do and what can't it do? Would it replace the function?
Read on for his full insights below.
Q How can ChatGPT be used in HR - what parts of the job can it do, and rather well?
I think the main thing that you can do with ChatGPT is tackle some of the most operational and time-consuming tasks that can be done more easily by AI types of tools than by people. Just like now, where we are using chatbots to go through resumes, ChatGPT can also analyse resumes and job applications, with the added advantage that it can quickly write up reports on what it has seen. Bots cannot do that, as bots are only able to be programmed to check on keywords in the resume.
So, with ChatGPT, it may be a lot easier for [recruiting teams] to write back to people to say ‘sorry, thank you for your consideration, but we cannot proceed with a job offer.’ Additionally, it can help automate the whole process, and helps tackle the operational part of scheduling the interview and writing the thank-you notes after the interview, thus a much better way of using people's time.
On the training & development front, ChatGPT is quite good at helping employees identify potential avenues and paths, as well as potential development tools and development courses both within the company and outside the company. If you bring ChatGPT into the company so that it has access to your own data, it can provide very useful advice to people as to what are the likely jobs that they could go into next, the competencies that they would need, how they are doing, and more, because you would have all the information on the employee and the information on the jobs so they can actually help you map these things out.
This could also make it very easy because then HR doesn't have to do all this - the employee himself or herself can ask ChatGPT these questions, and even get an answer as to which courses, programmes, or tools they can use.
You can also use it for all kinds of employee relations, but you would need to be careful that you make it a bit more ‘human.’ So, don't just use the standard ChatGPT language if you want to use it as a way to engage with staff. Make sure you personalise it. It's not that hard to do; just you tell your ChatGPT, for instance: ‘I want you to write to Priya and thank her for taking the time to do this interview with me’. You can get ChatGPT to write it, but you need to remember to ask it to be personal, otherwise, it will be quite impersonal.
Even for things like resources for mental health, I think it's not too far of a stretch to get ChatGPT to do them. It can also help you spot trends in data, although I'd be honest with you, with the preliminary testing that I have done, where I have asked it to calculate, it doesn’t calculate math properly, and it doesn't tell you exactly what you want. It could be that not enough machine intelligence has gone into the current version.
Another way I believe ChatGPT could be useful is, when we are doing research into industry trends, best practices, benchmarking, and so on, it could easily go through them. So, if you want to ask it, for instance: ‘I'm thinking of maybe introducing a no-limit vacation policy, can you tell me the trends? How are other companies implementing it, and what are the pros and cons?’ It could easily do that.
One particular way that I have already used ChatGPT, which I find quite useful, is in performance appraisals. You can ask ChatGPT to help you write the performance appraisal so you can do it instantly. For example, ‘write a note to Priya to thank her for doing this today’, or ‘help me write Priya’s performance assessment. Make sure to mention these four things that she's done great, and mention these two areas for improvement.’ This will help tremendously in something that bosses usually hate to do, which is writing up these paragraphs. It also saves a lot of time.
Q Keeping all these in mind, wouldn’t people’s jobs then be at risk?
I don't think so. I think, mainly, what we need to show is that ChatGPT is a tool to help us do our job better, not necessarily to replace us. Although, potentially, in some departments where certain staff are only tasked with things such as scheduling interviews or screening CVs, then yes, they may lose their jobs to ChatGPT.
However, if we take it in the context of, say, a report – it is true that ChatGPT can now write articles – well-researched ones – faster than you, but will it have the nuances that the individual reporter can bring to them? Will these articles have the insight that an individual reporter might bring to them? Probably not. So, essentially, if you want to make your writing adequate, you could write it using ChatGPT; but if you want to make your writing interesting, you can use ChatGPT as a tool so that you can do your own writing.
On the other hand, because ChatGPT is going to be prevalent and a lot of people are starting to use it, if you don't know how to use ChatGPT, then somebody that does is going to get ahead of you.
Q You’ve shared so much on how ChatGPT can help HR. On the flip side, what part of a CHRO’s job can it not do?
The biggest thing that ChatGPT cannot do is personal interaction.
You can ask it questions that can provide guidance and policies, but not add personal touch to your interactions. To take an extreme example that HR has to deal with – an employee has lost a family member who was tagged to their insurance plan, and they have approached you with questions or you need to now take the family member off the plan. ChatGPT can help you with writing the details on how this works, the policy information, and the steps in the process, but it can’t show empathy for the affected employee.
One problem so far is that if you rely on it for legal compliance matters, it can give you guidelines but you would not take its opinion as a ‘legal’ opinion. Lawyers, for instance, can use it for research, but not to write their opinions.
Apart from that, I also think that the software will have a problem if anything is too complex and has many interactions. Particularly, things that have to do with organisational dynamics, politics, and so on. To take another example: if I want to look at the next role an employee can take, it can give you all the answers. However, we as HR know that this employee would never get along with a certain boss and so one of the roles may be an issue – now this is something ChatGPT cannot do.
So, for that kind of nuance or dynamics, you would need someone who gives proper advice.
Another thing is – you can use ChatGPT to look at trends in the market, but you can’t really use it to plan the future.
The key point is that it should be a complement, not a replacement, for HR. And in maintaining the human touch, it is important to use personalised language, as mentioned earlier, and add empathy to your messaging.
Essentially, you can make it more human – but it isn’t human.
Q Such software can raise the issue of the credibility of the content provided - how can we put barriers or checks to ensure the people-related data it's feeding us is genuine? For eg, a job description.
To the degree that you will not use open-source AI in your company, but you will be paying for a service, you can point it to the data that you want it to look at. Thus, before you get that far, it should as much as possible be verifiable, or at least reliable, data.
I haven't tried this myself, but I expect that when you are paying for the service and incorporating it in your corporate toolkit, you can tell it to only check the sites or places where you can find the right kind of information.
It also allows you to cross-check the details. You can probably tell the software ‘I want you to do this only by looking at these job portals’ or instructions along those lines, and see if it comes back with more or less the same information so that you can have a better sense by checking on multiple sources.
Once you get into using it internally, machine learning itself will help you make sure that you are using the right language, you are writing in the right way, and that it is providing the right information.
But I think in the end, you will always need two things. One, is HR’s oversight, wherein somebody from HR should look at it; and the second is the user's oversight.
This is no different than the way we write job descriptions today. We either get consultants to write the job description, or we get the incumbent to write the job description. In either case, you are never 100% assured that they got it right because the consultant may not know the company that well, or the incumbent may be biased towards trying to make their jobs look bigger. So, you will always still need a pair of HR eyes, and an additional pair of supervisory eyes, to look at whatever ChatGPT produces.
Q Will data privacy & contextual integrity be an issue, when you compare with, say, data you find on Google Search?
Regardless of the technology, data privacy is a concern. The tool you use is not going to make it easier or harder. It's going to make it exactly the same, so it is about how you manage it.
With Google, you search the open internet on the same things as everyone else. It has cookies that track your activity, and now you can customise which cookies you allow – although I find this haphazard at best. If you forget to do it on one website, that website can sell your data to other parties without you knowing. So, the tool does not really make much of a difference except for one thing: it has more of you than goes into Google, or tools like Siri.
What I mean here is, when you ask these tools to search for something for you, they pull up all this research and so on. But if you ask ChatGPT questions, there is a little bit of you that it remembers, because even your way of asking questions is data that it collects. It remembers everything.
So, I would say, there is additional care needed when using ChatGPT – especially for more sensitive information such as calculating your budget expenses, credit card expenses, etc.
Alongside that, there may be one added problem: the legislation on cookies has more or less caught up with data privacy issues with the bigger tech companies, but ChatGPT is so new, that it will take some time for the legislation to catch up. In the meantime, we will need to rely on common sense. So, I do think if you bring an AI tool in-house, you can design it in a way that you can protect the information, the encryptions, and all of that; but if you are using an open system, at this time, it’s best to rely on common sense until it can be regulated.
Q Do you see AI chatbots like this one being contextualised by companies that can afford to reengineer the algorithm to suit their own industry/company needs?
I do see us heading in that direction. Companies will eventually bring it in-house and link it to their own databases, and make a chatbot personalised them. It learns from and searches in your databases.
I expect this open/free version of the software to remain, because companies such as Google and Microsoft will build it into their search engines. And very soon, within a year and not more than two, people will start ‘ChatGPT-ing’ - just like how we are now ‘Googling’ online.
Q Can ChatGPT help to raise the profile of AI in aiding human behaviour or tasks? Will CHROs be more inclined to try AI through this?
I think it will work better than current chatbots. We’ve already seen many HR departments incorporating standard chatbots into either their employee search or their recruiting processes. So yes, I can fully expect that this is just a better tool to do that. Plus, as I said, I think search engines will eventually work this way. I think it will help to do a lot of automated work, and give you a chance to do more value-adding work. So yes, that's why I made the point earlier that you may not lose your job to ChatGPT but if you don't know how to use it, you may lose your job.
What I hope starts to happen is, as more companies start to use it, they start generating best practices that can be shared. They can even get shared through ChatGPT, so if someone types ‘what are the best things AI can do for HR?’ we will have more information based on what these companies have been doing. I’m quite bullish.
Q Outside of the HR space, what have you used AI chatbots most successfully for?
Planning my holiday! I wanted to see what ChatGPT could do. I told it: ‘I have two weeks, I want to spend them in Europe. These are the three countries I want to go to. These are the dates I have. I want to travel by train in Europe. Please tell me how to organise my trip. How many days I should spend in each place, how do I arrange for transportation and what should I do in each place?’ And it came back with a whole lot of useful information.
Again, it’s not perfect, I also learned from experience that I added the countries, and in the first iteration of it, ChatGPT gave the information in the order in which I wrote the countries, and that was not necessarily the best order to travel within the countries.
Thank you for reading our story! If you have any feedback, feel free to let us know — take our 2023 Readers' Survey here.
Lead photo: Shutterstock / Kaspars Grinvalds
Fermin Diez's photo: Provided
share on
