What we listened to this week...AI in HR simplified
It all sounds so complex but in reality artificial intelligence (AI) in HR is not as scary as you may think. So says Megan Marie Butler, AI research analyst at CognitionX. Speaking on a Tap’d Talks HR podcast, she says AI is split into three areas: automation, cognitive engagement and cognitive insights.
What is confusing many people is the talk of bots or robots, says Butler. “In our minds we have this idea of a little mechanical man or machine, or a robot arm. But in automation, when they talk about robots it's a software robot. It's like a giant macro. In Excel, some people know how to create macros and record. Automation records the process that a person would go through in the program. And it's able to do that on a large scale with different programs within your computer.”
Very few employees will give the thumbs up. So don't take it to heart
In other words, being able to do things automatically is as simple as responding to emails or adding a new employee to multiple databases at the same time. It’s just replicating processes that we already do.
When it is complex it just means you need to know your processes well enough to be able to automate them. So Butler recommends starting small. Find a provider for desktop automation or robotic process automation to help you get started. You can even download software for free sometimes and play around with it to see what you can already do with it. Don’t get caught up with the bigger picture.
“We can go a bit too big with it, bite off a bit more than we can chew,” she says. “Or we see the end goal of where we want to be instead of thinking of just the little steps that we need to get there. We just need to start with the small steps. So, do one process.”
Let’s look at chatbots. You don’t need a chatbot to answer every question perfectly from the start. You can use what Butler calls “semi off-the-shelf” chatbots that are already designed to know the standard types of questions, or how people ask certain questions around, for example, recruitment or HR help desk queries.
“If you already have your list of kind of commonly asked questions within your HR department, or even if you don't, you can just brainstorm and come up with them, you can start to program the chatbot. Then release it to your employees and they'll ask questions. Within chatbots generally you'll see a little thumbs up or thumbs down. It's important employees know that it's okay to give it a thumbs up or down; that's part of the training process. Very few employees will give the thumbs up. So don't take it to heart!”
Butler says companies find that, when they initially release a bot it's about a six week process when they get about 65% accuracy, and within a few weeks just through testing and letting people use it, that jumps to 80%. If you try to programme it with all the questions you think people will ask, they won’t be the questions people ask. So best to just programme it enough to get going.
Also make sure the bot has a personality, as people will ask silly questions, swear at it and say “I love you, will you marry me?”
This is the most exciting area as it could radically change how organisations make decisions about employees. Uses include natural language processing (NLP) to analyse employees to see what is really going on in an organisation and what they really think. Candidate selection is another big area.
“This is traditionally based on a lot of gut feel. If you really kind of start digging into how we actually select candidates, it is a terrible process. You look at somebody's CV, judge their personality based on it, judge their skills based on it, judge their communication skills based on it. Then we pick up the phone and maybe put them through a screening interview. We have a face to face interview – the best ones are structured, but even then it's probably going to be unstructured even if we have set questions to it. The predictability of those methods are absolutely terrible.”
The problem is that it is difficult to get to more predictable ways of understanding job performance and success in a specific role – who’s been successful and who not, why they have not, links to their personality and mindset. You can use psychometrics, which can be controversial, expensive and time consuming.
Now, however, there are game-based assessments and others that analyse through the words you use in answering a handful of questions to give a better understanding of the candidate. It doesn’t say who to hire but gives you the top 20 or 30 candidates you want to speak to first.
“It doesn't mean that we don't need governance or that we won’t need to use other methods, but it will help give us more information and more data on an employee, on a person, to help us understand who's going to be more likely to be successful in a role,” says Butler.
This approach is also used in learning and development, helping employers to see what an employee wants to do in their career and what skills they need to develop, for example.
“This is where AI is coming in, being able to make recommendations clearly based on a real individual person. So this is where we're getting that hyper personalised experience within an organisation for employees.”
AI does not make decisions for us. It's using statistics and maths at a high level and helping us provide information about data that we couldn't otherwise have done ourselves without loads of time or a “little army of mathematicians”.
We know you can’t use past performance as a predictor of future performance, yet we use CVs and past performance all the time. AI can help us to get a better all-round picture and find what's best for people.
But we need to be careful that we don’t shoehorn people into roles they may not want to play. People can change and learn and develop into the new roles. So there’s a role here for HR in terms of ethics and AI.