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
you're looking for work you've sent in
cover letter after cover letter hundreds
of applications on every website there
is and yet you still have zero responses
or maybe worse hundreds of the wrong
ones and on the other side employers are
having the same problem too few of the
right applicants and an avalanche of
resumés making the right ones even
harder to find it's a bad experience and
a bad outcome for all parties nobody's
winning and the whole process can end up
feeling very impersonal and in human
let's figure out how generative AI can
help tackle these problems and start to
bring the human back into human
resources on AI and action in this
series we're going to explore what
generative AI can and can't do how it
actually gets built responsible ways to
put it into practice and the real world
business problems and solutions that
we'll encounter along the way so welcome
to AI in action brought to you by IBM
I'm Albert Lawrence and today I am
joined by John Lester VP of HR
technology data and AI at IBM hi John hi
elott good to speak to you today good to
see you and my other guest is David Levy
advisory technology engineer at IBM
what's up David hi Albert great look so
I know you both are probably wondering
why did I Choose You two for today's
episode well John is the brains behind
IBM's current HR Tech stack and David is
a developer turned engineer who at his
core is a builder so what better pair to
talk with us about how you can use AI
for hiring than these two because today
I want to talk about HR but HR is a
massive field right so let's get
specific about it we're talking about
applying and interviewing for jobs I
think that we can agree that it really
is hard out there people all across the
globe are having a tough time finding a
job and as I mentioned employers can't
find the right applicants is there a
tech solution for this distinctly human
problem I think there's not just one I
think there's many and I think if we
look at things like Ai and automation
that go across the whole end to end
hiring pipeline from job wck to
attraction to evaluation and finally
hopefully offer you're going to try and
find the right specific type of AI gen
AI automation capability for the right
World gotcha so are you also confident
about this Tech solution for a human
problem David I am yeah and it's
necessary we've gotten so far with
Automation and Ai and gener of AI now
makes so many opportunities to make this
process better for the hiring managers
and better for the candidates across
board okay so I mean that sounds like a
potential win for both right the hiring
managers as well as the candidates I
want to break this down a little bit
though I want to dig into the steps that
the hiring pipeline can take and how
exactly is AI changing those steps
within the hiring pipeline when we look
at
evaluation um this is now that can
really offer a lot of value for both
employers and candidates but also has
the biggest risk attached around ethics
similarly as we go across the whole kind
of um experience for hiring managers
recruiters Talent acquisition
practitioners we provided a gen digital
assistant capability that guides you
through that process gives you prompts
gives you insights but also make sure
that you kind of stick within the policy
and then next step really is to then say
how do we do a similar digital assistant
that can help candidates through that
process because it can be really
confusing you don't quite know where to
go having that assistant to guide you
will completely chains The Experience
John I really love the idea of having
you know an assistant to help the
candidate as an engineer it is a
confusing process to get a job and it's
a difficult and demoralizing process
often and there's so many different
things you have to know and routes you
have to take in order to get through the
hiring process having an assistant that
could actually communicate with the
candidate that's you know mapped to the
data they have on their side maybe like
info about the job rooll info about the
company all this stuff put into the
responses using something you could use
like rag uh you know or you could using
jbi obviously would be such a huge Boon
for candidates and that's across all
different Industries I believe you know
whenever it really comes down to you to
send those applications on out you
really can feel alone so that's really
encouraging to know that there can be
some additional assistance that's really
tailored specifically for you there
we've done a lot of work with similar
things for our employees which is why I
think we know this would work because we
have a an internal assistant called
askhr that people like David use all the
time and we just thinking to ourselves
wouldn't it be great to take that same
experience same automation same drive
for productivity back to candidates that
we do for employees so we we're pretty
confident it should work if the first
time you ever interact with a company is
the hiring process and you make
something that feels good you know it's
a comfortable hiring you get hopefully
you get the offer everything you're
going to start you're going to get right
in you're going to be you're going to
like the company you're working for
because it started off on the right foot
and I think that's so huge the beginning
the launch is so very important I want
to know from you John where is it
ethical to use AI in the hiring process
I mean that's a brilliant question
because a lot of the legislation that's
coming out of the EU a lot of
legislation that's coming out of the US
around how we use AI with the workforce
is really kind of touching on the what
should AI be used for and what shouldn't
it and I think it's worth reminding
ourselves that AI is not new I think
it's been around many many years but I
think what's changed is that gen AI has
suddenly moved AI into the public
Spotlight at IBM I'm actually really
proud of the fact that as an
organization we put AI ethics at the
heart of everything we do with AI not
just HR but across all Enterprise back
in 2018 we created an AI ethics board
that brings together key principles of
ethical Ai and I think certainly yeah
I've been at first slightly daunted by
this because I've taken a number of our
recent Solutions around assistant around
orchestration and automation to that
board and actually they're there to help
you know they're not there to to
intimidate intimidate it's really the
advice they give us around things like
explainability fairness transparency
robustness and privacy as well as I
think a fundamental key principle that
we have to keep humans in the loop with
AI as in humans ultimately make
decisions about humans what AI does is
augment Us in making quicker and more
informed ones I see it now as a real a
real benefit for me to have that group
of people there to to help me keep on
the right side of the law but also make
sure that we do these things in the way
and I think that's something for me I I
know massively rely on on all of any
your Innovations around Ai and in
particular geni so that AI ethics board
that you mentioned John is that
something that's unique to IBM or other
companies also utilizing the same
approach there are small number that
follow similar approaches um I spend a
lot of my time talking to peers in the
market kind of sharing our story
listening to theirs obviously because we
we learn a lot from them I would say to
them it's a relatively New Concept to
the majority and what we say to them is
as you move more and more into Ai and
especially gen AI this is this is not a
an option you have to bring this in
because otherwise you're going to get
people like myself out there going
should I be doing this will it match
with the the latest legislation or not
should I stop doing something so for me
it's it's not a it's not something that
should be optional in any
organization it should it should be
mandatory for me with that in mind then
David like how does that impact how you
build I think Engineers have a
compulsion to make everything efficient
and automated like my compulsion to
build something would just be like let's
figure out the most efficient way to go
through this entire process I wouldn't
even think of a human in the loop
because it if if I'm thinking as an
engineer but if I'm thinking as a
candidate that's totally different it
makes so much sense to me the way I
would think about it I I was thinking
about it last night about big part of
recruiters issue is that they have a
thousand resum and so they have a
thousand rumes in the docket and they
have to be like who's the best who going
to push forward to the next round and
there's a way I believe you could you
know obviously use NLP to entity
extraction and semantic all that stuff
you go to machine learning you start
scoring but that's where we start
getting into like I believe the Muddy
Waters with the ethics board once we
start scoring these resumés and like
obviously J and I could like summarize
their resume and all this stuff but John
how do how do we solve something like
that if we just give suggestions to the
recruiter would that pass the sniff test
for the ethics board or is it how does
that work I mean it's it's interesting
because a number of years ago kind of
back in
201617 we started experimenting with
using kind of um traditional AI to
actually start analyzing everybody's CV
and as you say score and rank and we
felt really uncomfortable about that
because it was even though you may still
say here's one to 100 a natural human
inclination is just say that's so great
I've only got to look at top 10 and
figure out which ones I interview so we
are very much led by a uring mechanism
what we do now is we much more go to
assessments um that basically tell us
does this individual have the Baseline
set of skills we need so we almost use
it to knock people out who actually
don't have the capabilities we do
something similar around language
capabilities in certain parts of IBM in
certain parts of of the world you have
to be able to speak English for example
so we use assessments to kind of go yes
no and what that tends to do is is
actually drop quite a decent proportion
out then unfortunately it is still down
to the recruiter to start going through
what's left and go okay these people um
are now matching the basic capabilities
how do I know start to rank them and
figure out which ones I'm going to pass
through to interview so even with that
it seems like a human is always going to
be an essential part of the loop there
they have to be it's back to how David
said how do I question you know if I get
rejected if I know there's been a human
in there somewhere at least I know
there's been a a human based judgment
call if I just think it's an algorithm a
I'm going to go well hang another way
that doesn't seem particularly Fair what
about bias within the algorithm but B
how do I then go figuring out how to
beat the algorithm because we're all
competitive we all want to get our CV to
the top of that list and that's where I
think some people are starting to think
how do I beat these algorithms well
don't because we only do algorithms that
go yes no answers there is no inference
there's no judgment in there and I think
when we start talking about inference in
hiring that's where we start the ethics
board start looking us very closely and
say yeah that's not something we want to
do so David though like when you are
training data how do you try to ensure
that bias is being removed on the build
side when you look at training data
there's a lot of methods to try to
totally remove bias and like this the
obvious ones right take all demographic
descriptors out of all the data use you
don't have their name you don't you
don't have their race you don't have
their age nothing right that's one of
the part of cleansing the data but if
you try to get your data set you have a
really broad set of data and you're
using from a truly diverse uh set of
people that's one of the best ways to do
because you're you have anonymized data
cleanse data but also a really wide net
of data of a lot of different people a
lot of different situation that's one
way to deal with it it is a really
really tricky thing to to fix because
there's bias that you wouldn't say
like you wouldn't see it and you'd have
to check you'd have to audit your
training data after the fact to see if
there's any patterns of favoritism for
any any type of biases you have to
constantly be governing the data that
you're using the models that you're
using otherwise you will it will just
pop up you can't really anticipate what
it will be until it's already built and
then you're following along you could
try your best but I think governance is
is that's the way to really really take
a deep look into it with any technology
we would always test it as much as we
possibly could before go live the thing
with AI and in particular gen AI is you
have to do even more testing once you're
live than you you do to get it live in
the first place and you never stop
testing and that's kind of quite a new
thought to people it's almost like we
have a model we build we test we do uat
testing and then we get live now it's
actually getting live is almost like the
start of the journey so I think we we
had a really interesting one back when
we we were kind of experimenting with
matching and things like that and we
found that for a particular role within
IBM the people from a specific
University always seem to get to the
head of the list and then and then when
we looked at it um it was like there's a
clear bias in the programming and then
we looked at the engineers who' built it
and four of them had been to that
University so now fair play completely
unbiased there was there was definitely
no intent but they had build the
algorithms based on what they knew to be
good what they knew to be good had been
taught by that University so and we only
found that out by testing after the fact
by continuously running the algorithms
by Contin looking at the reporting and
we caught it and then we were able to
fix the problem but then we just sat
back and went okay maybe this
experimentation is not a good thing and
then we again we started s to the ethics
board he just said nope any inference
any scoring just take it out that's not
allowed in the hiring process at all
right you know the funny thing about
that is though people being concerned
that AI might have bias but these are
humans that are still we as humans we
have our own biases as well too so we're
working against that with it there's a
lot of um a lot of publicity around that
bias within the AI bias human bias that
we build into the programs but yet if
you were to just have 100% humans do
this humans are more biased than some of
the algorithms we build so then you get
this kind of real challenge that says
actually maybe if you put the both of
them together you get the best outcome
and that's where we're really kind of
pushing heavily is where you can
automate a response yes no or knock
people out because they don't have a a
designated skill or experience that's
that's no there's no inference there's
no uncertainty in there that's really
good because that then gives the humans
a less number of of candidates to look
at less CVS and it can speed the process
up it can reduce the time to higher but
I think it's that balance between humans
and AI that if you perfect it will
really take us forward so this workflow
that both of you were talking about it
seems as though it really all does hinge
on data and data preparation because
these models are only as good as the
data and the continuous preparation that
they're actually given so be honest with
me here and with the world that's
listening do you actually see a world
where your resume doesn't just get lost
in some black hole once we click Send I
do you do yeah I do there's automation
processes all this stuff's going to get
better and like John let me ask you this
like the thought of getting no response
first getting an automated response
versus getting a personalized automated
response like there's all these steps
that you could take the worst in my
opinion the most demoralizing is getting
nothing just it's just a black hole
right getting an automated response is
not much better because it's just like
you could tell it's a caned response
sorry you know it didn't work out
whatever but having having the
automation tools and having gen and
having all this holistic data about the
candidate about the role about the
company maybe even making suggestions
for a separate role that maybe they
weren't a fit for this one but using
this automation process using AI giving
them suggestions for another role within
the company or or or telling them like
these are the things that you were
lacking and you could do this with you
could you could have a paragraph written
so that the uh the hiring manager
recruited doesn't have to write anything
but they have to just edit it and
approve it and send it rather than just
having an all automated I think that is
a distinct possibility know it's a
difficult question like to say which one
I would prefer I know
the one I would never want is just no
response so I I'm hopeful and I do
believe it I guess what we're starting
to call intelligent
orchestration is where you've now got
automation that can actually almost like
work with you through the whole program
and you can train that orchestration to
kind of look at events or look at
timelines and say this individual sent
their CV in on a
Monday by Friday has anybody got back to
them and the answer is no okay then the
orchestration will go back and say thank
you for your CV we are considering it
we'll get back to you by a certain date
and the orchestration can then look and
say well has anyone got back by that
future date so again we can use the
alteration to really track that
individual candidates process from from
start to finish and constantly either
send communication directly but
personalized back to the candidate or
nudge the hiring manager or the
recruiter to say did you not realize
this person is still waiting on a
response for here and what they can also
do is say okay if all the people have
replied have we what did the metrics
show us that we've actually got back to
every one of those candidates with with
an outcome the outcome may be sorry you
know you're not a good match for this
particular role but actually what you
consider these others it could be you
know what we are in the interviewing
process next week and we'd love you to
invite you to that process all of that's
just data it's data points it's it's a a
defined workflow that these this
intelligent orchestration can now
massively ensure that people don't get
forgotten and that you get a timely and
personalized response either from the
technology or from a human whichever
works I I got to tell you this actually
I'm excited about this like I I mean
look nobody wants to be rejected right
like rejection really does hurt but I
feel like ghosting is maybe the most
hurtful form of rejection where you put
all of your work on out there this is
something that's very personal and then
you just get nothing back at all so just
hearing from the both of you you know
this idea of not only just an automated
response but something that does feel
customized towards like my particular
skills towards other ways that I still
might be able to to join the company but
that's me from an applicant perspective
right how can a company actually benefit
for that so basically like what's in it
for a company in order to put this
additional effort out there so I think
from from my perspective we've got a
vision of creating what we call a
digital IB this digital ibma is one
platform that includes ad hoc tasks
using lar language models process
automation through AI assistant and even
program automation or orchestration all
on one platform what that will do is
take the the data that we need from from
unfortunately still in IBM we've got a
lot of HR systems a lot of recruiting
systems we are massively simplifying
that and that CRA this concept of data
fabric but what we do is we have some
really key metrics we look at things
like time to hire
uh look at things like quality of hire
but we also look at things like time to
productivity and how much productivity
can we give back to hiring managers
recruiters T acquisition partners and
what we've just been talking about that
for me is the big benefit is yes as a
candidate you're going to get a better
experience brilliant but from a
employer's perspective all of those key
metrics we are seeing gen and AI
significantly improve them and managers
now have time to actually do what we
call high value work high value work in
this case is those people that you're
really focusing on that you would love
to join IBM or any company spend time
with them reach out to them talk to them
engage them make them feel that you know
what I as a person would love you to
come and work in my team that's that
empathy that's that personalization that
we're all striving for but but managers
in particular too busy David mentioned
it earlier they're too busy to do what
they would love to do because they're
too busy doing admin the platform we're
talking about when dig will take a lot
of that had been work away and also even
give you prompts and insights as to how
you then approach that individual which
is going to be brilliant you think like
they they would be able to look at the
resume look at the job roll look at all
this and give suggestions for questions
to ask like like like just lead them
towards I mean obviously the human in
this is the hiring manager she's still
there he's still there but they get
these suggestions that they might have
not thought of that is so that is very
cool cuz that would be huge I mean just
giving them more time to be like
empathetic like just having a personal
conversation with the candidate or
having and just removing the TDM from
their job I think that's I'm very
excited so a few takeaways here there
are a ton of places in the hiring
process that generative AI can help
including job specs and job wrecks your
very first interaction with an
organization is hiring so make this
process great it really counts a human
in the loop is critical at all touch
points so get a solution live but then
never stop testing to continue to help
preventing bias again I feel very
inspired after talking with both of you
today I hope that all of our listeners
and viewers feel the same exact way I'm
so thankful that you both have taken the
time to be here so John thank you David
thank you this was a fantastic
conversation and friends that's it for
this episode so we thank you for viewing
we thank you for listening but don't
worry there's a ton more where this came
from and I promise you'll see us here
again soon all right
[Music]