Learning Library

← Back to Library

AI Restores Humanity to Hiring

Key Points

  • Job seekers face overwhelming rejection and irrelevant opportunities, while employers struggle to sift through massive volumes of resumes, creating a frustrating and impersonal hiring experience for both sides.
  • The episode introduces a discussion on how generative AI can be responsibly leveraged across the entire hiring pipeline—from job posting to candidate attraction, evaluation, and offer—to improve outcomes for recruiters and applicants.
  • IBM’s HR technology leadership, represented by VP John Lester and advisory engineer David Levy, explains that a combination of AI and automation, rather than a single solution, is needed to address the nuanced challenges of modern hiring.
  • While AI promises efficiency gains and better matching, the panel highlights that the evaluation stage carries the greatest ethical risks, underscoring the importance of responsible implementation and oversight.

Full Transcript

# AI Restores Humanity to Hiring **Source:** [https://www.youtube.com/watch?v=V-zTsDVqnfQ](https://www.youtube.com/watch?v=V-zTsDVqnfQ) **Duration:** 00:21:37 ## Summary - Job seekers face overwhelming rejection and irrelevant opportunities, while employers struggle to sift through massive volumes of resumes, creating a frustrating and impersonal hiring experience for both sides. - The episode introduces a discussion on how generative AI can be responsibly leveraged across the entire hiring pipeline—from job posting to candidate attraction, evaluation, and offer—to improve outcomes for recruiters and applicants. - IBM’s HR technology leadership, represented by VP John Lester and advisory engineer David Levy, explains that a combination of AI and automation, rather than a single solution, is needed to address the nuanced challenges of modern hiring. - While AI promises efficiency gains and better matching, the panel highlights that the evaluation stage carries the greatest ethical risks, underscoring the importance of responsible implementation and oversight. ## Sections - [00:00:00](https://www.youtube.com/watch?v=V-zTsDVqnfQ&t=0s) **Reviving Humanity in Hiring with AI** - The hosts introduce a discussion on how generative AI can alleviate the flood of mismatched resumes and unresponsive applicants, restoring a personal touch to the hiring process while outlining the technology’s capabilities, limitations, and responsible implementation. ## Full Transcript
0:00you're looking for work you've sent in 0:02cover letter after cover letter hundreds 0:04of applications on every website there 0:07is and yet you still have zero responses 0:10or maybe worse hundreds of the wrong 0:12ones and on the other side employers are 0:15having the same problem too few of the 0:17right applicants and an avalanche of 0:19resumés making the right ones even 0:21harder to find it's a bad experience and 0:23a bad outcome for all parties nobody's 0:26winning and the whole process can end up 0:28feeling very impersonal and in human 0:31let's figure out how generative AI can 0:33help tackle these problems and start to 0:35bring the human back into human 0:37resources on AI and action in this 0:40series we're going to explore what 0:41generative AI can and can't do how it 0:44actually gets built responsible ways to 0:46put it into practice and the real world 0:48business problems and solutions that 0:50we'll encounter along the way so welcome 0:53to AI in action brought to you by IBM 0:56I'm Albert Lawrence and today I am 0:58joined by John Lester VP of HR 1:00technology data and AI at IBM hi John hi 1:04elott good to speak to you today good to 1:06see you and my other guest is David Levy 1:08advisory technology engineer at IBM 1:11what's up David hi Albert great look so 1:14I know you both are probably wondering 1:15why did I Choose You two for today's 1:17episode well John is the brains behind 1:19IBM's current HR Tech stack and David is 1:22a developer turned engineer who at his 1:24core is a builder so what better pair to 1:27talk with us about how you can use AI 1:30for hiring than these two because today 1:32I want to talk about HR but HR is a 1:35massive field right so let's get 1:37specific about it we're talking about 1:39applying and interviewing for jobs I 1:42think that we can agree that it really 1:43is hard out there people all across the 1:46globe are having a tough time finding a 1:48job and as I mentioned employers can't 1:50find the right applicants is there a 1:52tech solution for this distinctly human 1:54problem I think there's not just one I 1:56think there's many and I think if we 1:57look at things like Ai and automation 2:00that go across the whole end to end 2:01hiring pipeline from job wck to 2:03attraction to evaluation and finally 2:05hopefully offer you're going to try and 2:07find the right specific type of AI gen 2:10AI automation capability for the right 2:12World gotcha so are you also confident 2:15about this Tech solution for a human 2:17problem David I am yeah and it's 2:18necessary we've gotten so far with 2:20Automation and Ai and gener of AI now 2:22makes so many opportunities to make this 2:25process better for the hiring managers 2:27and better for the candidates across 2:30board okay so I mean that sounds like a 2:32potential win for both right the hiring 2:34managers as well as the candidates I 2:36want to break this down a little bit 2:37though I want to dig into the steps that 2:39the hiring pipeline can take and how 2:42exactly is AI changing those steps 2:44within the hiring pipeline when we look 2:46at 2:47evaluation um this is now that can 2:49really offer a lot of value for both 2:51employers and candidates but also has 2:53the biggest risk attached around ethics 2:55similarly as we go across the whole kind 2:58of um experience for hiring managers 3:01recruiters Talent acquisition 3:03practitioners we provided a gen digital 3:06assistant capability that guides you 3:08through that process gives you prompts 3:10gives you insights but also make sure 3:12that you kind of stick within the policy 3:15and then next step really is to then say 3:17how do we do a similar digital assistant 3:19that can help candidates through that 3:20process because it can be really 3:22confusing you don't quite know where to 3:24go having that assistant to guide you 3:26will completely chains The Experience 3:27John I really love the idea of having 3:29you know an assistant to help the 3:32candidate as an engineer it is a 3:34confusing process to get a job and it's 3:36a difficult and demoralizing process 3:38often and there's so many different 3:41things you have to know and routes you 3:42have to take in order to get through the 3:44hiring process having an assistant that 3:46could actually communicate with the 3:48candidate that's you know mapped to the 3:50data they have on their side maybe like 3:52info about the job rooll info about the 3:54company all this stuff put into the 3:56responses using something you could use 3:59like rag uh you know or you could using 4:01jbi obviously would be such a huge Boon 4:04for candidates and that's across all 4:06different Industries I believe you know 4:08whenever it really comes down to you to 4:10send those applications on out you 4:12really can feel alone so that's really 4:14encouraging to know that there can be 4:16some additional assistance that's really 4:18tailored specifically for you there 4:20we've done a lot of work with similar 4:22things for our employees which is why I 4:24think we know this would work because we 4:26have a an internal assistant called 4:27askhr that people like David use all the 4:30time and we just thinking to ourselves 4:31wouldn't it be great to take that same 4:33experience same automation same drive 4:36for productivity back to candidates that 4:38we do for employees so we we're pretty 4:40confident it should work if the first 4:42time you ever interact with a company is 4:44the hiring process and you make 4:46something that feels good you know it's 4:47a comfortable hiring you get hopefully 4:50you get the offer everything you're 4:51going to start you're going to get right 4:52in you're going to be you're going to 4:53like the company you're working for 4:55because it started off on the right foot 4:56and I think that's so huge the beginning 4:58the launch is so very important I want 5:01to know from you John where is it 5:03ethical to use AI in the hiring process 5:06I mean that's a brilliant question 5:08because a lot of the legislation that's 5:10coming out of the EU a lot of 5:11legislation that's coming out of the US 5:13around how we use AI with the workforce 5:16is really kind of touching on the what 5:18should AI be used for and what shouldn't 5:20it and I think it's worth reminding 5:22ourselves that AI is not new I think 5:24it's been around many many years but I 5:26think what's changed is that gen AI has 5:28suddenly moved AI into the public 5:30Spotlight at IBM I'm actually really 5:32proud of the fact that as an 5:33organization we put AI ethics at the 5:35heart of everything we do with AI not 5:37just HR but across all Enterprise back 5:39in 2018 we created an AI ethics board 5:42that brings together key principles of 5:44ethical Ai and I think certainly yeah 5:46I've been at first slightly daunted by 5:48this because I've taken a number of our 5:50recent Solutions around assistant around 5:53orchestration and automation to that 5:55board and actually they're there to help 5:58you know they're not there to to 5:59intimidate intimidate it's really the 6:01advice they give us around things like 6:03explainability fairness transparency 6:05robustness and privacy as well as I 6:07think a fundamental key principle that 6:09we have to keep humans in the loop with 6:11AI as in humans ultimately make 6:14decisions about humans what AI does is 6:16augment Us in making quicker and more 6:18informed ones I see it now as a real a 6:21real benefit for me to have that group 6:23of people there to to help me keep on 6:25the right side of the law but also make 6:28sure that we do these things in the way 6:30and I think that's something for me I I 6:31know massively rely on on all of any 6:33your Innovations around Ai and in 6:35particular geni so that AI ethics board 6:37that you mentioned John is that 6:38something that's unique to IBM or other 6:41companies also utilizing the same 6:44approach there are small number that 6:46follow similar approaches um I spend a 6:49lot of my time talking to peers in the 6:50market kind of sharing our story 6:52listening to theirs obviously because we 6:53we learn a lot from them I would say to 6:55them it's a relatively New Concept to 6:57the majority and what we say to them is 7:00as you move more and more into Ai and 7:02especially gen AI this is this is not a 7:05an option you have to bring this in 7:07because otherwise you're going to get 7:09people like myself out there going 7:11should I be doing this will it match 7:13with the the latest legislation or not 7:15should I stop doing something so for me 7:17it's it's not a it's not something that 7:20should be optional in any 7:21organization it should it should be 7:23mandatory for me with that in mind then 7:26David like how does that impact how you 7:28build I think Engineers have a 7:30compulsion to make everything efficient 7:32and automated like my compulsion to 7:34build something would just be like let's 7:36figure out the most efficient way to go 7:38through this entire process I wouldn't 7:40even think of a human in the loop 7:41because it if if I'm thinking as an 7:43engineer but if I'm thinking as a 7:45candidate that's totally different it 7:47makes so much sense to me the way I 7:49would think about it I I was thinking 7:51about it last night about big part of 7:53recruiters issue is that they have a 7:55thousand resum and so they have a 7:56thousand rumes in the docket and they 7:58have to be like who's the best who going 7:59to push forward to the next round and 8:01there's a way I believe you could you 8:02know obviously use NLP to entity 8:04extraction and semantic all that stuff 8:06you go to machine learning you start 8:08scoring but that's where we start 8:09getting into like I believe the Muddy 8:11Waters with the ethics board once we 8:13start scoring these resumés and like 8:16obviously J and I could like summarize 8:18their resume and all this stuff but John 8:20how do how do we solve something like 8:22that if we just give suggestions to the 8:24recruiter would that pass the sniff test 8:26for the ethics board or is it how does 8:28that work I mean it's it's interesting 8:29because a number of years ago kind of 8:31back in 8:32201617 we started experimenting with 8:35using kind of um traditional AI to 8:39actually start analyzing everybody's CV 8:42and as you say score and rank and we 8:44felt really uncomfortable about that 8:46because it was even though you may still 8:49say here's one to 100 a natural human 8:51inclination is just say that's so great 8:53I've only got to look at top 10 and 8:55figure out which ones I interview so we 8:57are very much led by a uring mechanism 9:00what we do now is we much more go to 9:04assessments um that basically tell us 9:07does this individual have the Baseline 9:09set of skills we need so we almost use 9:12it to knock people out who actually 9:14don't have the capabilities we do 9:16something similar around language 9:18capabilities in certain parts of IBM in 9:21certain parts of of the world you have 9:23to be able to speak English for example 9:26so we use assessments to kind of go yes 9:28no and what that tends to do is is 9:30actually drop quite a decent proportion 9:32out then unfortunately it is still down 9:35to the recruiter to start going through 9:36what's left and go okay these people um 9:41are now matching the basic capabilities 9:45how do I know start to rank them and 9:47figure out which ones I'm going to pass 9:48through to interview so even with that 9:51it seems like a human is always going to 9:53be an essential part of the loop there 9:55they have to be it's back to how David 9:57said how do I question you know if I get 10:00rejected if I know there's been a human 10:02in there somewhere at least I know 10:03there's been a a human based judgment 10:06call if I just think it's an algorithm a 10:08I'm going to go well hang another way 10:10that doesn't seem particularly Fair what 10:11about bias within the algorithm but B 10:14how do I then go figuring out how to 10:15beat the algorithm because we're all 10:17competitive we all want to get our CV to 10:19the top of that list and that's where I 10:21think some people are starting to think 10:22how do I beat these algorithms well 10:24don't because we only do algorithms that 10:28go yes no answers there is no inference 10:30there's no judgment in there and I think 10:33when we start talking about inference in 10:35hiring that's where we start the ethics 10:37board start looking us very closely and 10:39say yeah that's not something we want to 10:41do so David though like when you are 10:43training data how do you try to ensure 10:47that bias is being removed on the build 10:50side when you look at training data 10:52there's a lot of methods to try to 10:54totally remove bias and like this the 10:55obvious ones right take all demographic 10:57descriptors out of all the data use you 10:59don't have their name you don't you 11:00don't have their race you don't have 11:01their age nothing right that's one of 11:03the part of cleansing the data but if 11:05you try to get your data set you have a 11:08really broad set of data and you're 11:09using from a truly diverse uh set of 11:12people that's one of the best ways to do 11:14because you're you have anonymized data 11:16cleanse data but also a really wide net 11:18of data of a lot of different people a 11:20lot of different situation that's one 11:22way to deal with it it is a really 11:24really tricky thing to to fix because 11:27there's bias that you wouldn't say 11:29like you wouldn't see it and you'd have 11:31to check you'd have to audit your 11:32training data after the fact to see if 11:34there's any patterns of favoritism for 11:36any any type of biases you have to 11:38constantly be governing the data that 11:40you're using the models that you're 11:41using otherwise you will it will just 11:43pop up you can't really anticipate what 11:46it will be until it's already built and 11:48then you're following along you could 11:49try your best but I think governance is 11:51is that's the way to really really take 11:53a deep look into it with any technology 11:56we would always test it as much as we 11:58possibly could before go live the thing 12:00with AI and in particular gen AI is you 12:02have to do even more testing once you're 12:04live than you you do to get it live in 12:06the first place and you never stop 12:08testing and that's kind of quite a new 12:10thought to people it's almost like we 12:12have a model we build we test we do uat 12:16testing and then we get live now it's 12:19actually getting live is almost like the 12:20start of the journey so I think we we 12:22had a really interesting one back when 12:24we we were kind of experimenting with 12:25matching and things like that and we 12:28found that for a particular role within 12:30IBM the people from a specific 12:33University always seem to get to the 12:35head of the list and then and then when 12:38we looked at it um it was like there's a 12:41clear bias in the programming and then 12:43we looked at the engineers who' built it 12:44and four of them had been to that 12:46University so now fair play completely 12:49unbiased there was there was definitely 12:51no intent but they had build the 12:54algorithms based on what they knew to be 12:56good what they knew to be good had been 12:58taught by that University so and we only 13:01found that out by testing after the fact 13:04by continuously running the algorithms 13:06by Contin looking at the reporting and 13:08we caught it and then we were able to 13:09fix the problem but then we just sat 13:12back and went okay maybe this 13:14experimentation is not a good thing and 13:16then we again we started s to the ethics 13:19board he just said nope any inference 13:20any scoring just take it out that's not 13:23allowed in the hiring process at all 13:25right you know the funny thing about 13:27that is though people being concerned 13:28that AI might have bias but these are 13:31humans that are still we as humans we 13:33have our own biases as well too so we're 13:35working against that with it there's a 13:37lot of um a lot of publicity around that 13:41bias within the AI bias human bias that 13:43we build into the programs but yet if 13:46you were to just have 100% humans do 13:49this humans are more biased than some of 13:51the algorithms we build so then you get 13:54this kind of real challenge that says 13:56actually maybe if you put the both of 13:57them together you get the best outcome 14:00and that's where we're really kind of 14:01pushing heavily is where you can 14:04automate a response yes no or knock 14:08people out because they don't have a a 14:09designated skill or experience that's 14:12that's no there's no inference there's 14:14no uncertainty in there that's really 14:16good because that then gives the humans 14:18a less number of of candidates to look 14:20at less CVS and it can speed the process 14:22up it can reduce the time to higher but 14:25I think it's that balance between humans 14:26and AI that if you perfect it will 14:29really take us forward so this workflow 14:31that both of you were talking about it 14:33seems as though it really all does hinge 14:35on data and data preparation because 14:38these models are only as good as the 14:39data and the continuous preparation that 14:42they're actually given so be honest with 14:44me here and with the world that's 14:46listening do you actually see a world 14:49where your resume doesn't just get lost 14:52in some black hole once we click Send I 14:55do you do yeah I do there's automation 14:58processes all this stuff's going to get 15:00better and like John let me ask you this 15:02like the thought of getting no response 15:04first getting an automated response 15:06versus getting a personalized automated 15:08response like there's all these steps 15:10that you could take the worst in my 15:12opinion the most demoralizing is getting 15:13nothing just it's just a black hole 15:16right getting an automated response is 15:17not much better because it's just like 15:19you could tell it's a caned response 15:20sorry you know it didn't work out 15:21whatever but having having the 15:23automation tools and having gen and 15:25having all this holistic data about the 15:27candidate about the role about the 15:28company maybe even making suggestions 15:30for a separate role that maybe they 15:32weren't a fit for this one but using 15:34this automation process using AI giving 15:37them suggestions for another role within 15:38the company or or or telling them like 15:40these are the things that you were 15:41lacking and you could do this with you 15:43could you could have a paragraph written 15:46so that the uh the hiring manager 15:47recruited doesn't have to write anything 15:48but they have to just edit it and 15:50approve it and send it rather than just 15:52having an all automated I think that is 15:54a distinct possibility know it's a 15:56difficult question like to say which one 15:57I would prefer I know 15:59the one I would never want is just no 16:00response so I I'm hopeful and I do 16:02believe it I guess what we're starting 16:04to call intelligent 16:06orchestration is where you've now got 16:08automation that can actually almost like 16:11work with you through the whole program 16:14and you can train that orchestration to 16:16kind of look at events or look at 16:18timelines and say this individual sent 16:20their CV in on a 16:22Monday by Friday has anybody got back to 16:25them and the answer is no okay then the 16:27orchestration will go back and say thank 16:29you for your CV we are considering it 16:31we'll get back to you by a certain date 16:34and the orchestration can then look and 16:35say well has anyone got back by that 16:37future date so again we can use the 16:40alteration to really track that 16:42individual candidates process from from 16:44start to finish and constantly either 16:48send communication directly but 16:50personalized back to the candidate or 16:52nudge the hiring manager or the 16:54recruiter to say did you not realize 16:56this person is still waiting on a 16:58response for here and what they can also 16:59do is say okay if all the people have 17:01replied have we what did the metrics 17:04show us that we've actually got back to 17:05every one of those candidates with with 17:07an outcome the outcome may be sorry you 17:11know you're not a good match for this 17:12particular role but actually what you 17:13consider these others it could be you 17:16know what we are in the interviewing 17:18process next week and we'd love you to 17:20invite you to that process all of that's 17:22just data it's data points it's it's a a 17:25defined workflow that these this 17:28intelligent orchestration can now 17:31massively ensure that people don't get 17:32forgotten and that you get a timely and 17:35personalized response either from the 17:36technology or from a human whichever 17:38works I I got to tell you this actually 17:41I'm excited about this like I I mean 17:43look nobody wants to be rejected right 17:45like rejection really does hurt but I 17:47feel like ghosting is maybe the most 17:49hurtful form of rejection where you put 17:52all of your work on out there this is 17:54something that's very personal and then 17:56you just get nothing back at all so just 17:58hearing from the both of you you know 18:01this idea of not only just an automated 18:04response but something that does feel 18:05customized towards like my particular 18:08skills towards other ways that I still 18:10might be able to to join the company but 18:13that's me from an applicant perspective 18:15right how can a company actually benefit 18:18for that so basically like what's in it 18:19for a company in order to put this 18:20additional effort out there so I think 18:22from from my perspective we've got a 18:24vision of creating what we call a 18:25digital IB this digital ibma is one 18:29platform that includes ad hoc tasks 18:33using lar language models process 18:35automation through AI assistant and even 18:37program automation or orchestration all 18:40on one platform what that will do is 18:43take the the data that we need from from 18:46unfortunately still in IBM we've got a 18:47lot of HR systems a lot of recruiting 18:49systems we are massively simplifying 18:51that and that CRA this concept of data 18:53fabric but what we do is we have some 18:55really key metrics we look at things 18:57like time to hire 18:59uh look at things like quality of hire 19:01but we also look at things like time to 19:03productivity and how much productivity 19:04can we give back to hiring managers 19:07recruiters T acquisition partners and 19:10what we've just been talking about that 19:12for me is the big benefit is yes as a 19:15candidate you're going to get a better 19:16experience brilliant but from a 19:19employer's perspective all of those key 19:21metrics we are seeing gen and AI 19:24significantly improve them and managers 19:26now have time to actually do what we 19:29call high value work high value work in 19:31this case is those people that you're 19:34really focusing on that you would love 19:35to join IBM or any company spend time 19:38with them reach out to them talk to them 19:40engage them make them feel that you know 19:43what I as a person would love you to 19:45come and work in my team that's that 19:47empathy that's that personalization that 19:49we're all striving for but but managers 19:51in particular too busy David mentioned 19:52it earlier they're too busy to do what 19:55they would love to do because they're 19:56too busy doing admin the platform we're 19:59talking about when dig will take a lot 20:01of that had been work away and also even 20:05give you prompts and insights as to how 20:07you then approach that individual which 20:08is going to be brilliant you think like 20:10they they would be able to look at the 20:12resume look at the job roll look at all 20:13this and give suggestions for questions 20:16to ask like like like just lead them 20:18towards I mean obviously the human in 20:21this is the hiring manager she's still 20:23there he's still there but they get 20:24these suggestions that they might have 20:26not thought of that is so that is very 20:28cool cuz that would be huge I mean just 20:30giving them more time to be like 20:32empathetic like just having a personal 20:34conversation with the candidate or 20:35having and just removing the TDM from 20:38their job I think that's I'm very 20:40excited so a few takeaways here there 20:43are a ton of places in the hiring 20:45process that generative AI can help 20:47including job specs and job wrecks your 20:50very first interaction with an 20:51organization is hiring so make this 20:53process great it really counts a human 20:56in the loop is critical at all touch 20:58points so get a solution live but then 21:01never stop testing to continue to help 21:03preventing bias again I feel very 21:05inspired after talking with both of you 21:07today I hope that all of our listeners 21:08and viewers feel the same exact way I'm 21:10so thankful that you both have taken the 21:12time to be here so John thank you David 21:14thank you this was a fantastic 21:15conversation and friends that's it for 21:18this episode so we thank you for viewing 21:20we thank you for listening but don't 21:23worry there's a ton more where this came 21:25from and I promise you'll see us here 21:27again soon all right 21:28[Music]