Automating recruitment | Todd Carlisle | TEDxLASalon

alright how's going everybody so y'all the robots are coming you know the robots are coming but are they coming for us humans the humans and HR will they displace the people behind the very name of Human Resources I think actually they will be coming but it won't be as bad as we think and those be some advantages but first let's take a step back here is a hundred years of HR history in about 30 seconds so we started with a guy named Frederick Taylor who had this concept of the principles of scientific management which is basically the idea that you could take Adri individual movement that somebody does in his case steel workers and optimize them even better after some time this moved to the concept of personnel and this was kind of the the dark ages of HR where where humans were basically files to be moved around large groups of masses to be put here and there but eventually we figured out there's humans and human resources we rebranded ourselves we started to do things like training and development leadership organizational planning and things got better during this time also statistics got better and we started to do things like factor analysis and structural equation modeling and ultimately things like machine learning in big data and this brings us to the 2000s and for me about 2004 when I started at a little Internet start-up called Google which was a great place to be if you were kind of a data nerd like I was and wanted to bring kind of analytical rigor to the HR process so Google at the time was filled with executives lots of data and lots of questions that they wanted kind of data-driven answers too it was a great place to kind of start what started there which was kind of this concept of people analytics and you can find lots of different definitions of people Analects out there and you can do your own research but essentially for me back in 2004 what this meant was taking data adding insights and getting better people decisions out of this so I've since moved on from Google and gone to other places and also been on the other side of it which is how do people interpret that data but people analytics has taken off as it field this is a graph of all the search term for people analytics in Google over the last 14 years and see it's up into the right this is a concept that's we are seeing a lot of uptick on so what can you do in this kind of field you can turns out you can do all sorts of interesting things for example if you have 300 different things you might know about somebody that are like we call predictor variable so where you went to school but maybe where you worked but also how job Hoppe were you what things were you into as a kid what hobbies do you have how many of these things will actually predict performance once you actually get there so performance variables like your first performance score on the job or even altruistic behaviors how much you help other people at work so when you throw all these into interesting analyses you find out that most of them actually aren't that predictive but you do find some gems in there so for example as I'm able to find things like the age at which you got into computers compared to your peers was a really good predictor of whether you're a good software engineer later on in life and also found things like people who are highly neurotic make better product managers draw your own conclusions on that one you can also do something like the fabulously named Cox proportional hazards model for survival time so in this analysis picture two people blue line person green line person where you can measure them over time time is on the x-axis and you want to measure their happiness which is on the y-axis so I have a theory that you're happy say on the job is your your first day on the job because just excited to be there you don't know any better and then your happiness kind of trails down over time and I know that sounds depressing but it turns out the data kind of supports that a little bit but what happens is your happiness doesn't trail off at a steady state rate it goes in drops big drops and in our field we call those drops organizational shocks and those shocks can be these are some of the shots that I found in the research if you go from a top-tier manager to a terrible manager that's a huge shock and your happiness will drop down as soon as you find out that that's the case if your reorg for the third friggin time you will be less happy on that third time so you have these things that happen to you and as HR professionals if we can get ourselves in there and insert ourselves before that shock happens we can prevent some of the decrease in happiness and it turns out if you get these things it's way more predictive of whether someone will leave then if you ask people are you considering leaving people are not actually that good at telling you that even a year from now so you can do all these cool things you can get all this cool data and you want to have these happy employees how all these happy jumping people that are there the issue is you have the boss in the middle the decision-maker the executive and some of the fidelity or the power of that analysis gets degraded as you go through the West the messy sometimes well-meaning sometimes slightly clueless person in the middle so you know why is that the case so I've found there's a couple reasons where I think this happens the first is that your intuition you feel like it's way more important than the data you're shown so if you ask somebody I want you to interview somebody but I'm gonna give you a bunch of questions and you have to stay to this structured script and answer that everyone's like nah don't do that give me five minutes with it person I will tell you if we need to hire them and it's a little bit like being told that's great you signed up for match.com you don't have to go meet the person at a bar just go to the chapel the algorithm is so good it will tell you show up there this is the person needs to marry people don't like that they want their intuition in there and speaking of intuition it's it's one example of all these biases and heuristics that we all have that we bring to the table that are really hard to ignore so from hallo to horn similar to me bias confirmation bias we come with all this stuff and even if we know that it's there it's really hard to not have it be there and the third is we want an easy answer we want a top free list right we just just give it to me and top three things the issue is the robots are still coming AI still here machine learnings here so we need to find a way to prepare for this so what I think we need to do is as HR professionals we need to act as the Rosetta Stone need to have this data that's here we need to have people that can interpret it and and make it sing for our executives because they're busy and they're mostly well-meaning and they just have 5,000 other things they need to do and you need to help draw that out from them just a couple other things I think that will help along the way the first is we need to clean up our people data and it's a light study last year it was found that 69% of organizations are preparing their data for better people analytics which is great start but if you talk to any of the social scientists within your organization's they'll say we got a long way to go there's a lot more stuff we can get we need better attrition data better performance data a lot more stuff second is we need to upscale ourselves as HR professionals we need to bring social scientists on people with training and instead who can make this again this data sing interpret this for me how does this work and the third is we need to pass some of these repetitive tasks that we actually don't really want to do on to the chat box on to machine learning on to AI so you know what are some of those things so there are now 500 million people on LinkedIn half a billion people and you don't want to just give that data set to human recruiters who are going to go through this very slowly one at a time even they can write some basic code you should be able to train algorithms to find subsets of people in LinkedIn to do that better the second is scheduling and phone screens and even conducting some of those phone screens so it breaks my heart a little time every time I see recruiting coordinators trade emails 20 times back and forth just to set up one phone screen for then in a recruiter to then ask the same 10 questions or asking of all these other people we can train actually virtual humans to do this for us so that we're not wasting that that time and the third is answering FAQ so these are the kind of thing most HR people don't want to do when is open enrollment how do I sign up for my 401k where's the employee handbook like talk to the chat bot that person will figure that out so the good news is the robots can't do everything machine learning can't fix everything and there's lots of things that you would never ever send robot to do for example let's say you're asked to investigate why is the experience in your tech company different for women or people of color than it is for white tech pros you're not sending a robot in to figure that one out that's super nuance right or let's say you were worried that there's an employee there who's been stealing a few pennies from thousands of customer accounts over many years and the press is about to find out about it that's a tough one for an algorithm or you're given a hundred liters and you have to figure out who are the best 10% that you need to give extra compensation for and you have five days to do it so these are actually three challenges I've been asked to do that I can't ever imagine giving a robot to do so I just want to want to end with how we can blend all this together how can we be an AI a data-driven AI enabled HR function and what this could look like for the end-user so particular yourself as an employee in Columbus Ohio this is probably a live image of Columbus here on the snow and you're you're fairly happy in your organization there but you know you'd be open to other opportunities so you flick your little thing in LinkedIn that says you're open to new opportunities and that alerts something that's been crawling through LinkedIn that's been keeping an eye on your profile that now goes I think this person will be ready and that sends them a message to see if they would like to trade a few message back and forth to see if they'd be interested in your company which is a gaming company in California where it's 72 and sunny and the and to see if they'd like to do that and the algorithm figures out that they're gonna send this note when the temperature differential is the highest between where you are and where you're going right these are all things you can code into this and we know things like actually people are more responsive to LinkedIn in Mills I think it's a Tuesday morning so you could even pick the time to optimize the chances that this candidate will be more interested so you trade emails back and forth and at the end of this you have a conversation with a virtual human that asks you basic questions about yourself how much are you looking what are your salary expectations what do you want to do you know when you grow up and it can record all this stuff and score this and if you score high enough it alerts the human recruiter who gets in touch with you and eventually you fly out you fly out you know and you get the job and you start in the job but it turns out your managers not so good and so you go talk to your human HR person and say I'm not like in my my manager that much that human has a nice conversation with you codes that into a database which identifies a trend with this particular manager that they are not nice to multiple people that person can then get sent a nudge email for training that's being given by real humans so this is kind of the world that I'm I'm thinking we're moving towards where it's a combination of real live humans big data and AI working together to enhance the experience so I'll end with a quote from Frederick Taylor who we kind of started this with who said in the past the man has been most important in the future the system must be most important so in this case I think the system is the humans the big data and AI to make all of work better for all of us humans thank you you

Channel: TEDx Talks

Views: 10,177

Keywords: TEDxTalks, English, Business, AI, Economics, Robots

Id: ljmsrGdlo9A

Length: 11min 53sec (713 seconds)

Published: Wed Dec 12 2018