Generative AI and the future of HR (2023)

Generative artificial intelligence: it's powerful.is available. It promises to change the way we work. in this episodeMcKinsey talk about talentIn this podcast, talent hosts Bryan Hancock and Bill Schchaninger discuss the promises and perils of using genetic artificial intelligence in HR, from recruiting to performance management to chatbot support, with McKinsey Technology Council President Lareina Yee and Global Editorial Director Lucia Rahilly. An edited version of their discussion is included below.

What's different: So devastating

Lucia Lahiri:In recent months, there has been a lot of talk about generative artificial intelligence and tools like ChatGPT. Many seem to vacillate between marveling at the potential of these tools and fearing their inherent risks. Lareina, what makes generative artificial intelligence different and what lies behind its disruptive potential?

Larena Yeh:Several things stand out about generative artificial intelligence. In November 2022, OpenAI released ChatGPT 3.5, and within 5 days the number of users exceeded one million. So the adoption rate is unlike anything we've seen.

The most profound thing for me at that moment was that anyone, of any age, any level of education, any country, could go into GPT, search a question or two, and find something useful or interesting, like a poem or an article. There is an experience for everyone. Since then, we've seen great advances in technology, and it's only been a few months.

Another super cool thing is that you don't have to be a computer scientist to take advantage of this technology, it can be used for all kinds of jobs. OpenAI research estimates that 80 percent of workplaces could incorporate generative techniques and artificial intelligence capabilities into the activities that take place at work today. This has profound implications for talent and work, and it's not in the same way we talked about before.

Somehow, the genie is out of the bottle. Trying to get him back is probably not the best strategy. Lean forward and learn how to use it effectively and safely.

Lucia Lahiri:The immediacy of the use case feels so new and so fast. Please explain what generative artificial intelligence is, which is why we are working on a common definition of the term.

Larena Yeh:Generative artificial intelligence is a technique that suggests the next best answer. Many people use ChatGPT to collect information, to compile an answer to something by collecting large amounts of public data. But there are also amazing pictures. Maybe you want songs, audio, video or code. Code is a great example. The range of things that generative artificial intelligence can do in the world is amazing, and it's just getting started.

Brian Hancock:I asked ChatGPT about me and he correctly told me that I work hard on talent. However, he falsely reported that I went to Cornell because he felt Cornell was the most appropriate answer based on my background, rather than the University of Virginia I attended. I think it's very interesting that you don't necessarily get the right thing, but the logical thing.

Larena Yeh:In a way, this mimics the way we think. I'm not saying I think like a human being, but in many ways we use shortcuts and clues to make assumptions. That's why people say, "Man, that's really smart." But, Brian, that's not 100% true. There's a good term for this: "hallucinations."

What the AI ​​generation means for recruiters. . .

Lucia Lahiri:We'll talk more about some of the risks, but let's look at what these generative AI capabilities mean for talent in particular. Do you expect generative artificial intelligence to reshape or change?Recruitment processin any meaningful way?

Brian Hancock:I think it will reshape employment in two significant ways. The first is to help managers write better job applications. Generative techniques can actually demonstrate the skills you need to be successful at work. This does not mean that managers do not need to review the final product. They have to be vigilant to make sure that the job requirements are good. But the AI ​​gene can significantly improve speed and quality.

Another application in recruitment is candidate personalization. Now, if you're an organization with tens of thousands of applicants, you may or may not have a super personalized way to reach applicants. With generative artificial intelligence, you can further adjust candidates, jobs and possibly other jobs on offer if the candidate does not fit. All these things are easier and faster with generative artificial intelligence.

Bill Schninger:The best generations of AI applications are largeskills libraryYou are trying to fill a fairly well-known position. We need a more efficient way to navigate through all the passed configuration files. What makes me a bit concerned is any time it's a new job, a new role, or even, according to US law, a job change of more than 25% or 33%. In these cases, you should go back and reconfirm your judgment about the people in or out of the pool.

The challenge with validation is that you need performance metrics to go back and say, "What's the difference?" In some cases, this means figuring out how to get them out of the data lake without violating someone else's proprietary performance data standard. If you say, "Well, we're only going to use our data as an employer," then you're just basing the standard on the people you've hired. To check, you have to look at the people you didn't hire.

This does not mean that technology cannot be used. It just means that there will likely be more entry-level jobs to apply for new positions and a wide range of opportunities for a large skill set.

Lucia Lahiri:We often talk about over-indexing of qualifications and under-indexing of skills in the recruitment process. Can generative artificial intelligence work?accelerate changesFrom credentials like a college degree to the skills a candidate can actually bring to the workplace?

Larena Yeh:I'm optimistic. One thing this technology does very well is tagging: the ability to tag words for unstructured data. There are many companies that are thinking about applying it to e-commerce and different types of retail experiences. But you can also apply it to talent acquisition or skill hunting. Now you don't have to look for certificates or degrees. You can search keywords based on skills and abilities.

Look at social media, how do people talk about certain skills? You may find that there are better words to relate to someone with those abilities. Think of a world where you want to find candidates who have experience learning on the job, but don't have a PhD or college degree. I'm optimistic, this can open more doors for people like this.

Bill Schninger:This is an interesting trade-off in the business world, which favors proprietary datasets and profile groupings. The real power might be, "How much can you get in the public domain before you start attacking the paywall?"

A long time ago, when LinkedIn was acquired, the API was limited to job titles, not necessarily all the specifications below. There is power in these groups, especially in job profiles, because then tasks and skills are visible. I think there's going to be a race here to figure out how we can put this together to form an ontological cloud, so to speak, of "these 17 things that describe a capability."Skills, not credentials.

...what this means for professional development

Brian Hancock:You can also think of this as helping with skills-based transitions, not only from an employer's perspective, but also from a candidate's or employee's perspective. In today's world, if you may have some skills but are not entirely clear about your career options, you will be very dependent on a manager or someone who takes an interest in you and helps you navigate your "non-traditional" path.

But in the world of generative artificial intelligence, you can talk to a chatbot that's really smart and say, "Hey, here are my skills and experience. What jobs are available for me?" You can respond and say, "Well, most people with your skills will do these things, but some will do A, B, C" where "C" is coding. Then you can say, "Tell me what those coding jobs are," and you can show a coder job description that's not only applicable to IT people, but can also be translated into text you can understand. Then you can say, "Okay, this is great. I'm interested. What learning experiences do I need?" Generative AI can tell you what those learning experiences are.

So for people with innate but unknown abilities, generative AI could shed light on different career paths and begin to help people understand how to get there.

Larena Yeh:Imagine that ten years after my graduation I feel a bit stuck. What if I had an AI career assistant to help me think about things like, "What kind of job should I be looking for? What types of jobs are available at my company? How can I see them?" and "What courses should I take??" instead of waiting for someone to retrain me, that sounds awful. How can I proactively build my skill set and understand the scope of work that matches my abilities, ten years after my degree? That's very nice.

Bill Schninger:Depending on the regulatory environment you're in, you can't make selection decisions without human input. This is especially true for the European Union. This is a great way to increase human performance without reducing decision-making. Regarding personnel, it should ensure greater transparency; in fact, it shows how close you are to many things. I like the employee experience part. I'm looking forward to the selection part just because we're still not sure what's in the data lake and how good humans are at giving AI clues.

Larena Yeh:rectify. I'm happy to give you some options, but this is not a response or recommendation mechanism. Your judgment is important.

Brian Hancock:Another thing we've seen is that ChatGPT, and generative AI in general, is particularly good at onboarding new hires faster.

Erik Brynjolfsson of Stanford University and others at MIT recently conducted an interesting study looking at call center workers. They found that generative AI capabilities were not as useful for more experienced agents. This is very useful for newcomers as they can gain institutional knowledge faster. It's within reach. I can ask questions and get answers. As a result, new hires are significantly more productive. Generative AI can really get you to 80-90% full skill.

Larena Yeh:I like that, Brian, and I'm an optimist.

McKinsey talk about talent podcast

Bryan Hancock, Bill Schanyinger and other talent experts help you navigate the rapidly changing landscape and prepare for the future of work by turning talent into a competitive advantage.

watch more episodes

(Video) Generative AI Is About To Reset Everything, And, Yes It Will Change Your Life | Forbes

What's new in performance reviews

Brian Hancock:One of my personal favorite uses of generative AI for humans is actually performance evaluation. Hear me out: I don't want generative AI to actually generate someone's performance review. This requires informed human beings, human judgment and empathy.

But let me use this example to illustrate what I do as a McKinsey reviewer: I get written feedback from 15-20 people. They write it in the number system. I got a long feedback. I look at feedback results from the bottom up, including written reviews and ratings based on specific numbers. I observed how often people actually implement commitments. I review compliance-related measures. Did they deliver the items on time? all kinds of things. For me as a reviewer, getting the first draft was a very daunting process. I'm proud of the time and consideration that went into it.

But what if you could press a button and get an eraser? What if I already have an outline of every interview with the 15 people who know the person I'm evaluating best? It's not a substitute for everything, but the initial synthesis will help me understand more quickly the development and growth of the person I really need to explore.

I'm excited about this use case because it saves a lot of work. At first, many people will think, "I never want to generate an AI anywhere near a performance review." But if we think about it as something that increases productivity or helps us be better, that would be exciting.

Larena Yeh:Now let's talk about the employee you are evaluating. The employee gets the feedback, Bryan probably writes it down clearly, conveys it with empathy, so that the person feels, "Okay, I have some strengths, I have some development needs."

But if I, as an employee, can ask, "Who are five successful people who have my strengths and weaknesses, and what did they do next? How can I envision my career growing? How can I continue to work hard?" I also have an assistant who helps me plan my career development. That way, when we moved a year later, I really progressed and increased my ambitions.

What if Bill was someone you should emulate? Without Bryan introducing me to Bill, generative AI helped me realize that I had what it took to be Bill Schanyer. I can be inspired by it. I think there are many things that can improve what we have been trying to do for years.

Bill Schninger:We are talking about re-engaging managers in performance management. Every time you talk to someone about something good or bad, write it down. So, it's more about adding up and synthesizing towards the end of the year, which shouldn't surprise anyone. But it requires regular access. While I like what you're describing, it's not technology, it's technology. they are people working on generic data collection and a generic approach to achieve it.

Brian Hancock:Your opinion is good. Then, as an evaluator, I apply my human judgment.

Bill Schninger:Canonical data is fine. When we receive our sponsorship and training data at McKinsey, we see how we compare to other partners in a given region. But how do you know what's "good" if you don't have a reference point? Once you have the canonical data, you can start to get some guidance. I like it all and it's all backed up by loads of data.

If it leads to a more robust and healthy view of actual performance, havingDifficult acting dialogue..We have toPut managers back on performance managementBut can we make it easier for managers so they can spend their time managing instead of sloppily making a timeline or knitting 15 data points?

Bias and other risks

Lucia Lahiri:Let's talk a little more about risk. Generative AI learns from historical data whose historical patterns reflect historical biases. What is the risk that we inadvertently spread these genetic biases by relying on AI-powered generative tools?

Larena Yeh:Today, of course, generative artificial intelligence amplifies the bias.

Let's say I'm hiring and I'm describing several different qualifications. I was looking for a talent center and decided to look for basketball captains or conversely I said lacrosse captains would be preferred. They are team sports with captains and leaders, so it kind of makes sense.

But if you look at the demographics, the people who play basketball in the cities are very different from the people who play lacrosse. So by emphasizing lacrosse, you'll generally get more young white male leaders, while if you choose basketball, you might find more African-Americans or Latinos. What about softball, where do we see women? Conversely, what if we select the entire set of sports? However, simply selecting sports as a filter can reinforce the bias in the question. I think the power of this question lies in us humans.

Brian Hancock:Of course, there are also issues of intellectual property.

But I also think we all run the risk of becoming less interesting. If you're in a creative field and you're taking advantage of generative artificial intelligence to increase your output from 6 articles a week to 12, you'll spend less time on each article. You may have to do this in order to post it on time, but it also means you won't spend as much time showering, running or driving.thoughtAbout the article. Your productivity will increase, but you may not have as much time for creative thinking. We know that the most creative ideas come from free time, when you are doing other things and letting your mind wander.

This not-so-fun risk is an important one, and we probably haven't fully considered it yet.

Larena Yeh:exactly There are many risks. Let's also think about the managers who implement this technology. People often have a workflow where they think about technology and business ROI, only to end up asking, "Are there any risks we need to worry about?" I recommend thinking about risk in your workflow design beforehand.

The second thing is the real opportunity for what we usually call "change management". If you're not thinking carefully about how technology is changing work, workflow, or collaboration models, you're not necessarily spending extra time on more value-added things. You need to consider how this affects the work day and the rest of the work week.

Bill Schninger:In many cases, we prefer to blame the technology rather than highlight the poor handling of problems that occurred before it was implemented. Getting a better, lighter, faster, and wider tool doesn't relieve you of the burden.

Larena Yeh:More importantly, the three of us spent this time thinking about all the positive intentions and how we can put them to good use. But there may be people who think about this technology and wonder, "How can I use it to cause harm?" Traditionally, this is why government regulations, policies and international standards have played a fundamental role in our society. I do not think that the private sector can be completely left to self-regulation.

prepare for the inevitable

Lucia Lahiri:People's biggest fear is that these tools will lose their jobs or, perhaps worse, become their boss. What do you think people can do now to prepare for the changes that generative artificial intelligence will bring?

Bill Schninger:I will try to make it easier for them to learn and use. This is better than still trying to resist him. I don't think we should be held hostage by these fears.

Lucia Lahiri:Assuming HR and talent processes become increasingly automated, how can leaders ensure that generative AI doesn't get in the way of what Bryan calls "people in the loop"?

Larena Yeh:Leaders can make a big difference in two ways. One is that within the scope of his functions, he carries out modernization and rapid development of his own talented abilities. Second, if 80% of your workforce is being transferred, this will play a big role in how the transfer will occur and how the transfer will affect the company's employees. I think the leaders have a lot to say at the negotiating table.

Brian Hancock:This is a great opportunity for HR to increase the chances of a large number of employees getting the opportunity. this is an opportunity to winManagers achieve performance levels more consistentlyHuman resource managers have always wanted to achieve this rather than work in administrative positions. I hope HR sees this as an opportunity to routinize work they don't have to do and get rid of it. And then for the work they have to do, they can use this technology to find a way to get better answers faster.