The Possibilities for Generative AI in Sourcing

thanks for joining us today we are  going to talk about the possibilities for generative Ai and sourcing and uh I'm  gonna I'm here Amy Fong I'm a partner with Everest group I lead our programs focused on  sourcing and vendor management stakeholders um excited to have uh my colleague Vignesh  here as well as a uh a friend Edmond and I'll let them uh introduce themselves and  and talk a little bit about their roles um Edmund you're the guest you want to go first  thanks Amy uh it's great to be he

re and uh exciting to discuss the topic of generative Ai and  autonomous sourcing today with uh with YouTube uh so my name is Edmund zagarin I'm the founder and  chief strategy officer of orchestro Orchestra is a predictive procurement orchestration platform that  uh uses artificial intelligence machine learning but also Game Theory and Behavioral Science to  help orchestrate complex procurement processes and we are um really focused on uh helping  companies really instrument and execute their s

ourcing and buying and procurement activities  uh using data ideally using real-time data um and uh my background is in strategic sourcing  and um I uh actually kind of began working on this presentation in concert with a number of people  on the ism thought leadership committee and that's one of the reasons that I'm so excited about this  topic and excited to share uh Vignesh do you want to do uh give a little bit of your background and  introduction yeah thanks thanks Edmund uh thanks Amy I'm

super excited to be on this call and  uh a very interesting topic to discuss a quick production uh Vignesh Khan and vice president  with the firm uh I lead our research and advisory practice focused on procurement supply chain  uh and Finance and Accounting I've been with average group for over six years now so yeah  looking forward to this exciting discussion is uh so excited to have two different  perspectives here these guys probably are way deeper in this than I am but I think the  evolution

and the potential for generative Ai and the sourcing Space is really uh exciting and  fascinating so this uh this format is really meant to be conversational and we want to involve all  of you in the audience as well um as a start would love to hear where our uh the attendees those  joining on LinkedIn where are you all today were you joining from um I know I'm in uh I'm in Dallas  uh Edmund is in San Francisco today and Vignesh you are in Bangalore right yeah hey so we have a  pretty Global Gr

oup here um so just add it to the chat and uh as we go through feel free to add your  comments or your questions in the chat and welcome Susan from the UK great to see you Amsterdam wow  these are all great places to to visit too I love the global uh Global audience and happy to see you  Kate from Dana Point another great location wow um so great to see that why don't we jump into  it uh so as a start we're gonna hit on a number of different topics today the first thing I want  to do is just uh

Define generative Ai and let's talk about what differentiates generative AI  from other Solutions and I know Edmund has done this several times and has a pretty good uh  pretty good Viewpoint uh from this so I'm going to ask him to go first uh and just give us you  know five minutes of of how you view generative Ai and really why should we be talking about  it today and why should sourcing care about it thanks Amy and uh always always  happy to share a perspective um you know I think a lot of us

in procurement and  and supply chain and in sourcing uh you know about four to six months ago depending on where you are  in the world got hit by this question what is chat GPT what is generative Ai and how is it going to  affect our business and that those questions might have been coming from our executive teams they  might have been coming from our stakeholders they might have been coming from uh people on our team  who said hey I'm starting to use this technology it's pretty exciting uh and

so I want to answer  some basic questions why are we talking about this how does the technology work and what does  it mean for our profession and for our industry so first uh this is just a little piece of data  which is that this technology became the fastest technology to reach a million people and this is  a metric in technology disruption you can see uh Netflix took 3.5 years Twitter took two years  Spotify took five months Instagram took 75 days and chat gbt accomplished that same uh reac

h in  just five days so really really dramatic growth and it's it's funny to look back on this because  it was originally released really as a beta just for people to play with and it just uh kind  of caught fire and took over the conversation and it didn't stop there it went on to a billion  pretty quickly uh thereafter a million and so the world is is paying attention to this technology  there are lots of people that are using it on a daily basis to perform various tasks ranging from  searchin

g the web to answer getting their basic questions answered and so it is um uh kind of this  dramatic shock to the system and the question is well what does this mean for procurement what does  this mean for sourcing supply chain and business operations The Wall Street Journal uh did a  a deep dive on this looking at some ways that um generative AI through a chat bot interface  could change supply chain and one of the um things that we do a lot in procurement is we  ask questions of our data we w

ant to know how much we've spent with the supplier where an order is in  the world how long it's going to take to get there and if there is a problem what the status of it is  and what our team is doing to solve it and these are all questions that potentially generative AI  could give us faster and more accurate answers to and I think that's something that you see  in the journalism around this technology that people are just really excited about one thing  I did want to mention is from um Orche

stra customers and one of the themes that have come  up in our discussions with Chief procurement officers and chief supply chain officers is  that this technology is not just computer to computer from a kind of traditional application  perspective you're also seeing applications that use as an input signal technology that's being  deployed at the loading dock and so one example um of this technology in production was instead of  relying on a person to put in information from a manifest or bill

of lading there is now a sensor  at a loading dock with a camera on it that will visually confirm that an order has arrived on time  in full and record that metric so a human does not have to enter it into a workstation and then  we'll actually use generative AI to make inference is the next steps from that ground truth input  what this means practically for us is that while a lot of times we've had to rely on imperfect  data unclean data data that takes a lot of time to enter and takes even mor

e time to confirm  and check we're now getting it kind of from The Source data from sensors from cameras that  really can be game changing for our profession um next slide please and uh this has dramatically  impacted AI adoption among procurement and supply chain teams as you can see from uh this  recent research there's only six percent of our cohort that is not considering this  technology everyone else is on a journey where they're beginning to get educated  evaluate different solutions stac

k rank their most impactful use cases uh begin  piloting and implementing uh technology in production with suppliers whether it's  managing communication managing stakeholder intake or managing real-time analysis of data  that that matters to them next slide please and so I want to talk a little bit about what  makes generative AI different and uh there are probably some people on the call and uh you know  I've I've now uh lived through uh two major hype Cycles uh around Ai and procurement uh so

me people  might have lived through four or five of these everyone gets really excited about AI uh and then  uh you know six months or a year later people are like okay what do we have to show for what are  the business impacts um and so I'm I'm going to I'm very very skeptical in general but I'm gonna  make an argument uh that's cautiously optimistic about generative AI having reached an inflection  point that has powered some of these seismic changes like you saw earlier first to a million  an

d first to a billion and it really relates to how this technology works in two fundamental  components so uh Transformers and reinforcement are kind of the fundamental building blocks  and one question that I always like to ask an AI vendor is what is your reinforcement learning  human feedback loop your RL HF Loop and you can see to some extent that Loop described in this  diagram you have an input there's a Transformer that turns words into zeros and ones transforms  them into inputs for stati

stical models back into words and then it shows an output and then  a human responds to that output and based on the response the machine says did I do it right did  I do a good job and it learns and it learns again um this is based on a paper that was written by  Google deepmind researchers called attention is all you need which essentially says using human  attention we can scale these reinforcement learning human feedback loops without limit across  many different disciplines simultaneously a

nd that paper and research around Transformers is  really what's responsible for the AI Renaissance that we're seeing today next slide please uh and  shockingly enough it comes from stuff like this uh so uh on the right we see Chihuahuas and blueberry  muffins which is which uh on the left we see um uh fried chicken and labradoodles and uh so  being able to answer basic questions and then have humans confirm that that classification is  a way of scaling classifiers which is a kind of primary tec

hnology and AI research in ways that  haven't been possible uh before and so you know kind of the the main application that generative  AI got started on and one reason it became so popular is it just became the best way to make  memes and so like who doesn't want to make a meme like people are are using the technology for that  but in doing so it also teaches the AI systems what people like and what people are interested  in next slide please a little bit of vocabulary I'm not going to dwell on

this but I just think  it's useful for the viewers to have access to this um you know we're we're going to get into the  conversation in in just a sec but I think having a little bit of background vocabulary is always  useful the main takeaway here is traditional AI required large amounts of data that was very  clean and consistent in order to do basically anything uh and the new generative AI data gender  of AI models uh do not and they don't because of the architecture of these Transformers a

nd so  you have an objective function we want the you know machine to engage us if it chats to us we  want it to chat back there's a training set it reinforces by how we respond to it the model has  weights and parameters we can say we don't want uh certain types of responses we do want other types  of responses longer versus shorter faster versus uh longer Tempo labels then you have supervised  learning which is where we have to label the data there's deep and unsupervised learning and  then th

ere's efficiency which is just really on the compute side and then tuning where we actually  take it and perform domain specific tasks this is some of the most exciting growth um especially  for the procurement side of the house where various bespoke models could be tuned to specific  categories of spend uh next slide please [Music] uh and this is this is coming right so uh  globality recently uh announced their chat bot uh called glow um which purports to be uh  as good as a co-worker which I t

hink is uh a uh bold but also big if true claim and I think uh  it will be exciting to see how that technology evolves and Tonkin uh has launched procurement  GPT uh which is essentially the same technology applied to procurement so this technology is  already being beginning to show up among many different uh uh solution providers and I think  it it falls on us uh helping um our teams and our stakeholders navigate the technology journey  and procurement to understand what this technology is how

it works and if there's leverage  uh that that we can gain next slide please um and this is uh the last point that I'm going  to make uh it's a point that's been made uh by by many people uh more more eloquently than I I'm  not going to talk about existential risks right now or uh the propensity for AI to cause supply  chain disruptions if you're interested in either of those topics um connect with me on LinkedIn  uh there's some some interesting work being done um through uh through a number o

f  different uh Association groups um but the final point I'll make is that this  technology is uh it sounds very confident um but it often says things that are not true um  and so again going back to the architecture of reinforcement and Transformers which is the uh  the the building blocks of of chat gbt and and these other generative AI Technologies um this  will this technology make stuff that makes that it thinks will make us happy and will make us uh  acting positive and excited ways and i

t thus has the form of truth but not the value of truth  there's a great uh article that that I think um called stochastic parrots that suggests AI is  like a parrot it's just trying to say stuff to get get a reaction from us uh some people  call it fancy autocomplete if you've used autocomplete but can imagine it completing like  an entire page of text but it also hallucinates it says things that aren't true it will make  up news stories that haven't been written um and it will confabulate whic

h is something that  uh Alzheimer's patients do where if they make up something and you challenge them on it um they'll  make up more stuff uh to support the things that um that that you know you've challenged them  on that that also aren't true uh so this is a little bit of a warning label on this technology  as you embark on your journey just be aware that um uh this can produce uh results that  that are not that are not accurate all right thanks Edmund um I always  love that Fried Chicken Lab

radoodle one um probably one of my favorite little uh little  AI comparisons and it's not necessarily not really new either it's uh I've seen that floating around  for many years now but makes me hungry every time so thanks that's a great overview and I think it  kind of level sets a lot of people I see there's a couple questions and we are going to get to  those actually some of the questions you're you're asking in the audience are questions we do  plan to cover especially around confidentiali

ty um the big National I'm going to give you a chance  to jump through an Everest group definition and kind of our perspective on the evolution as well  um Edmund's coming from one perspective I think our perspective as as research as a research firm  focused on you know Global Services and Technology um we have a bit different angle on the market  and uh so it's interesting to me to hear hear both of you talk through in your own words yeah  thanks thanks Amy and Edmund a great perspective befor

e we get into gen a uh probably let's  take a look at uh you know the evolution of a in itself right uh you know it started it's not  Jenny a fundamentally it's not a new New Concept right I would say uh it's a it's a significant  evolution of a traditional a models right if you look at look back uh I I I'm sure most of you  remember Windows 98 uh if you remember right this is that a paper clipping yeah that's that's the  basic form of a tool right it started from there and the sophistication ev

olved right and we we  got into you know different forms of a in the form of say Siri and a lot of chatbots came into  the picture and a lot of machine learning models came into the picture but the last I would say 18  months or or two years uh there's been significant development in terms of the social investigation  and the evolution of the a models in it in itself right uh looking at chat GPT or the GPT model gpd3  and gpd4 what makes this this Jenny very different from the traditional AI uh

it has been there like  you have a machine learning deep learning your NLP energies your Gans Transformer model diffusion  model it has been there right the key uh here is what makes jna so powerful is the volume of  data and the number of parameters that go into the model the sophistication of the uh a model in  itself right just a comparison uh GPT if you take there are 175 billion parameters that forms this  model and before GPT the largest uh ml model was from Microsoft Turing with 17 billio

n parameters  so you can see that 10x jump in terms of the the number of parameters that go into the model right  and other couple of factors that that makes this tool So Sophisticated is one that has been a  significant Improvement development in terms of computation power so now we have systems that  can process huge volume of data and lastly we have now data sets right and and even Jenny is  feeding into that in terms of you know creating your synthetic data to further train your model  right

so the quality of classified data sets and and the sophistication of the ml model and the  faster computation systems that we have is what is making change and what is making is very  different from your traditional AR ml models thanks Vignesh uh so I want to Jo and I think  that that's a that's a good overview and you know we all hopefully we all remember clippy or the  some of you may be too young to remember clippy I think both of you might be actually but uh that  was a pretty useless form

of AI to many people I think it's really fascinating how fast things have  changed uh to the point that I think most people we've got other webinars where we pulled and most  people have used you know chat GPT in some form or another and it's been way more useful than clippy  was so I want to turn now to uh how gen AI is helping sourcing or what the potential really is  in the sourcing space and when I think about this I think of you know when I look at the people I  work with right cpos their t

heir teams category managers they kind of need to think about this in  two ways they're they're learning their educating in two ways one is what is the business and I.T  going to need to support us so if I'm talking to an insurance company they're going what is  generative AI going to do to Insurance what are my competitors doing what do we need to start  partnering or source thing to support you know RIT our business functions all of that and that's  super important for us all if you're if you'

re a category manager you definitely should be thinking  about this um we have a number of different Everest group webinars on that and so I'm not  going to try to dive into that today because it's a huge topic that's evolving fast but if you go  through some of the the blogs and the webinars on our site we start to address you know the broader  scope of uh generative ad and the Enterprise for today I want to focus on really the  functional needs so specifically how is generative AG going to imp

act our own processes  so how we do work who we work with who we need to accomplish things how we work with suppliers  within the you know source to pay process the procurement function um so Vignesh if you  could talk a little bit about this slide and this is a framework I know we've used for a variety of  different processes as we look at generative AI um just understanding which processes you are  thinking right now are going to have the most opportunity yeah yeah uh before we get into that 

I'll I'll probably you know uh talk a little bit about the procurement ecosystem and how it is  evolving right so traditionally compared to other functions say your your contact center or your  Finance or HR function procurement has been you know uh kind of behind in terms of adopting Next  Generation technology it has always been uh cost Centric function but uh kovit changed a lot of  things right uh there are a lot of organizations thinking of you know procurement being a value  creation cente

r and not just a cost center so what kovit did was you know to accelerate the  adoption of digital in the procurement function so we have seen a significant adoption of multiple  Technologies be it your your s2p Suite or there are a lot of best of great Solutions coming into  the mix and uh uh specifically you know Market intelligence kind of uh category intelligence or  a market intelligence Solutions uh I've picked up adoption so in general procurement is accelerating  uh the adoption of digit

al and covet did that right now the way I see jna is covet accelerated  the adoption of digital and jna will accelerate the adoption of data and analytics right that's  how I see uh uh you know Janae in the context of uh procurement now if you look at uh from a  procurement angle and the application or specific use cases or processes within the sourcing and  procurement environment uh there are you know couple of perspectives to uh block from one uh I  I strongly believe the entire buying experi

ence uh is going to change if an organization is  adopting generally and traditionally if you look at some of the Legacy systems in in the  procurement World it has been designed in a you know process-centric mind circuit it always  has that process-centric architecture right now with the advancement of Technologies with new  platforms coming in it is that the technology ecosystem and the architecture is moving from A  process-centric View to a user-centric Persona Centric View and Genie will ac

celerate that so  to to take an example if you take the entire buying process right Jenny you will act as an  assistant in your buying process and looking at in the context of your b2c uh buying experience  and procurement world is now adopting the concepts and Technologies and process designs from the b2c  buying experience a lot of organizations have come to us and you know we want to move to more like a  Amazon uh buying experience right so I see this as unlike other Technologies uh DNA is mo

re like  evolved from a B to C kind of uh uh Evolution getting into an Enterprise grade B2B solution so  I believe this will accelerate the adoption and buying process will change significantly right  it will act as a assistant to say for example a buyer is buyer wants to order uh say 10 laptops  right it can act as an assistant uh you know give you prompts like hey if if timeliness is very  important then go with these suppliers of these suppliers or you know uh cost effective and are  complian

t with our uh contract policy and if Janae is able to integrate with your catalog system  with your contract lifecycle management system the power it can give and the assistant that it  can give to buyers immense right now of course you know some of you have raised some questions  around confidentiality security are those are still concerns that you know we need to address  to realize the potential of this right so one the buyer experience will change from from a buyer  angle second uh if you lo

ok at from A supplier Community it's going to change the experience of  the supplier itself right one common uh or high value uh quick win that I can see is uh help this  kind of a solution right yeah we have conversation anyway we have chat Bots uh but but the data and  and the sophistication of the model that Jenny is bringing in it the the uh the entire uh vendor  query process will be you know almost touchless right uh so it will change the experience of  the buyer it will change the experie

nce of the supplier and lastly it will make category managers  the entire procurement function intelligent and frees a lot of bandwidth to you know focus  on higher order strategic work right the the amount of data that the procurement function is  sitting on right here your supplier data your materials you need category intelligence you need  supplier intelligence right uh the ability of DNA to derive insights and synthesize uh in a manner  that is required for a procurement professional that w

ill save a lot of time and give lot of  value to this right and looking at yeah using this framework we consider different factors like  you know what are the factors that will influence the adoption of this technology uh what are the  factors that will determine the potential of this technology availability of data what is the level  of reasoning required for this particular process and is that data sensitive how critical is that  information so we looked at different factors and this is again

a preliminary view uh we see you  know a lot of use cases uh one around the creation of document right whether you are purchase order  creation your rfx creation or contract creation right so there it has lot of potential uh and  from a sourcing angle coupled one we discussed around uh say you know the helpless kind of  a use case the other is more around bringing intelligence you're reporting uh say if you have  to roll out identify a supplier for a particular requirement it can analyze supplie

r and in fact  it can you know monitor the performance of the supplier you know given the right data right  so I would say it's still in early stages yes there are a lot of concerns to be addressed there  are a lot of uh you know factors to be considered sensitivity uh all that and the developments  are happening but I would say uh that to to Echo Edmund's point this looks solid uh compared  to other Technologies like you know blockchain metaverse there was a huge bus and then trout  right but t

his shows some real potential there yeah thanks Vignesh uh yeah and I look at this  and I think about you know some of the other use cases that we're seeing um contact centers  right what can we learn from that b2c support uh you know how people can order I I know there was a  headline about Ikea Now using you know AI for all of their uh contact center agents and retraining  those people into being you know designers right which is a great example of Shifting talent  to a different skill set and

reskilling um but when I think about how that applies into  uh procurement you know those are all things that we could be doing and we've been talking  about like you said the Amazon like experience well what that experience is in other contexts  is it has evolved and will continue to evolve um Edmund uh comments on on that slide  or any anything you would add in terms of you know specific impacts and use cases  see a lot of questions about use cases yeah well I think just in terms of the quest

ions  to think through one of the things that Vignesh mentioned that I think is really important is  you know you can ask generative AI to write a document for you and it will produce a  document that I mean I've seen it write things that are a better first draft than what I  would have written or what uh you know especially someone who potentially uh uh less experience or  newer to their role might produce and so I think um you know it that's really interesting  because getting a first draft of

a RFP is a labor-intensive process it's not entirely  rewarding and a lot of companies have these template libraries a lot of people have worked  in procurement for a long time have like a bunch of templates on their computer they have  friends who they go oh hey have you uh you know have you sourced uh you know this type  of packaging before do you have some Excel sheet that you you can you can hand over so that's  something that generative AI I think is is going to be very interesting to see

how it impacts  that and then one of the things that came up um when we were talking at the the ism committee  was how do we feel about our suppliers using this technology and I think you know Amy to your  point uh this technology is in the contact center it's in consumer applications it's in a  number of different areas and I think it would be a mistake for us in working in procurement and  and sourcing and supply chain to think that this technology is not going to come to our suppliers  and it

might even come to our suppliers before it comes to us and so you know that's a question I  would put uh you know to you if you're advising a a an Everest client and and someone says should  we have a policy or guidance to our suppliers on using chat gbt to generate submissions to our  rrfps and how do you feel about that I mean does that feel like something hey no problem  like if they submitted it's legally binding doesn't matter who writes it or do you feel like  there are that that companie

s should have Guidance with with uh with their suppliers yeah I think  that's a really good question um in terms of writing things I actually use chat GPT to write  the description for this LinkedIn live um that is not my favorite task and I've been trying to come  up in the past I don't love doing and how I can Source them to the Bots and it did it pretty well  I mean I just needed a couple prompts and I wrote a whole a very long agenda and description  and I sent it to marketing and said go ah

ead work with this it's it's what chat GPT says we  should talk about to talk about Chad gpg so you know saved a little bit of a headache for me  but yeah I think it's a really good question um I my take would be the you know what you  said around if your suppliers are using chat GPT to respond to an RFP as long as what they're  responding is binding and they need to do I think Anita made the point it you know it's great for a  first draft I would advise suppliers to use it as a first draft and

then go through and review it  I you know anytime we submitted our fee response where you know we have two or three sets of eyes  on it to make sure that everything's accurate um I think that's very important but given that  you know as long as suppliers are are doing that and they're held to what they submit I think  it's fair game right I mean rmp responses can be very tedious no offense to our sourcing but  I've been on both sides right they're very tedious and frankly sometimes it's a lot of

  the same questions in the same you know in in different formats and I mean I've I just did  this yesterday here we asked the same questions last week as these these answers right and but  reformat them into the right spreadsheet right um and as we look at you know one of the  challenges I've seen from the supply side or even from the from the buy side in adopting Tech and  the e-sourcing side everybody's using different platforms and you have seven different versions  of the spreadsheet with t

he same questions asked different ways for every response on you know the  Ian voicing side everybody's using a different form or POS or whatever your supplier networks  are all you know I've talked to very large Cloud company several years ago and their AR person  was like uh or AP person I guess was like I don't care I'm not I'm not submitting to 17 different  portals and I'm like well if you're not going to automate your invoices then no one is right so  I think those challenges still kind of

haunt us and uh there are ways to get around that right  and automate that process and make it easier for both the customer the buyer and the supplier  and us um so I don't have a moral issue with that although we all are responsible for accuracy  and that's where things can go wrong right yeah um picnicity any other thoughts on  that or anyone any thoughts on that yeah I think uh I saw some comments around  uh data quality uh data security right uh and especially procurement uh it interacts wi

th  suppliers and the data from the companies going going out and recently I think a month back a  a very confidential information uh uh from a from a phone manufacturer went out right uh and  someone in the company used chargpt to redraft something right uh so there are you know very  it can have critical damage uh repetition wise Financial Risk IP risk yes uh but but the  key to accelerate the adoption and uh you know realize the potential is how can you you  know guard the data that a jni can

access and how you can utilize this data and I know a lot  of big techs and and service providers in the procurement world are working with big techs and  platforms to build security you know complaints uh around this but that is to me going to be  how the data is going to be managed when jna is integrated with the procurement systems  will determine the uh an odd option rate yeah I think I know there's a lot of questions on  confidentiality and I want to I want to dig into that in a minute um

I do want to talk about what  it means for suppliers and um I I think one thing that I just heard earlier today in a leadership  meeting you know we work with a lot of service providers as well and the outsourced Services  side and um some of my colleagues pointed out that there are service providers that are hearing  from their clients that they're not allowed to use generative Ai and I think that goes back to  Edmonds point I think that's really interesting um you know both from an internal pe

rspective  of companies saying you can't teams don't use it um and from a you know supplier perspective of  saying we don't want you using it at all and and certainly the risk part is a good reason but I I  think that that's something that we probably need to rethink because when you think about it from  uh you know from I.T from a coding perspective contact centers all of these different areas  of the business there are huge opportunities to do work that's really not easy for for  uh people to

do and it's much more efficient um so I think that's an interesting  path there to go down but uh I know you know you work with a lot of the service  providers on the outsourced procurement side um can you talk a little bit about this slide  and what some of the big providers are doing in this space and and also how we can you know use  procurement Outsourcing providers um to to learn to adopt more what are you saying yeah so uh the  last six months or I would say three four months uh you know m

ost of the leading uh you know  Outsourcing providers have started investing significantly in in uh you know bringing Jenny to  business process services in general right and uh on the slide you can see some some examples here  but uh you know a couple of areas where they are making Investments is one uh these large leading  uh service providers are setting up a dedicated uh Jane AI Coe and uh exploring different use cases  uh right across different functions uh different Industries so uh identi

fying the use cases where  China can uh play a role what where it can have a you know meaningful uh potential so we have seen  lot of reading providers Accenture is one example and uh you know EXL has also set up a Coe ibms set  up a Coe with uh dedicated thousand thousand and 1500 people uh who are you know purely focused  on exploring uh use cases and how can how can they bring jennai to the table uh the other type  of Investments are progress that we have seen is a very strategic partnership

with big Tech right  how can uh they bring in Chennai to their platform to their ecosystem uh so we are seeing a lot of  strategic collaboration uh partnership and a lot of this procurement Outsourcing uh providers  are uh actively working with big text to address two key concerns right the security and  confidentiality data data conferencing uh other kind of Investments that I've seen as uh you know  most of these leading providers have uh uh good AI capabilities right for example cognizant has

  a neural platform now they are embedding DNA into their you know neural platform right similarly  you know PCS have igneous the area platform web pro has homes as their ego uh you know platform  and IBM I'm sure you all heard of Watson now they have launched what's an X where they have  integrated Genie capabilities into their platform so in in short period of time in our own uh you  know six months or so a lot of these providers have uh you know showed that agility in investing  and exploring

Chennai whether uh if you're asking are there any real implementation use cases yes  to some extent most are in Pilot stages uh but uh compared to broader Enterprise side of the picture  we are seeing uh on on the supply side the service provided side the extent of investment is this  massive in the last uh five to six months yeah thanks yeah it's really uh it's really been  a huge investment fairly quick some things have been around for a while like Watson but  really yeah shifting track on a

lot of this as the technology is caught up um Edmund uh anything  that we miss here in terms of and then I'd love to hear from your perspective what sourcing should be  asking their suppliers for uh around generated AI yeah I think you know it's it's just it's an  exciting time to be to be doing this work and to be thinking about these questions and I think  that the best that procurement and sourcing teams can do in questioning the not just their suppliers  um from a service provider materials

uh and so on side but questioning from a technology side and  saying hey we want to know what llms you're using if you have plans to adopt this technology  what are some of the friction points that we have in using your technology and how are you  applying llms or generative AI to reduce some of those friction points and I think one of  the things that's really interesting is that you know to vignesh's point earlier about like  The Help Center being a real quick win like how much do we spend on

user adoption and training  for new procurement systems like anyone who's gone through a source to pay transformation  knows that like the change management um and you know just user adoption journey is  can be can be a pain and so like this has the potential to radically impact user adoption of  existing source to pay technology to me that seems like the the quick wins here are not going  to be in like radically rethinking our you know procurement process it's going to be in getting  additional

leverage off of Technologies uh we're already using that's one reason that Orchestra  overall is is betting big on embedded systems we don't think that companies are going to you  know rip out uh systems they've spent decades in some cases really really fighting to to get  into their Enterprise it's more how can we use emerging Technologies to use what we have  better get more Roi from some of our existing technology investment so that's one of  the themes that that we've been hearing um and I

I want to um uh there's a comment from  Anita whis that I I think is is really uh uh powerful and I just want to um kind of elevate  that for for the group discussion which is um you know and this actually uh Echoes something  uh that uh Sacha Nadal of Microsoft has said which is like think about how each role can get  leverage for higher value activities and it's so funny because one of the earliest professions  that was like meaningfully impacted by this technology is software programming and

like you  know if you're an AI programmer like your job has changed like and it's changed not just because  the technology is different in terms of what we work on but because um it's like pretty easy for  AI to generate code so like literally write code um and so the last uh thing I would  say Amy to to your question of like um you know how how can procurement use generative  AI now to for for a business case and like what questions would you ask your suppliers I  would challenge everyone on th

is call who has a relationship with an I.T services provider  to ask how generative AIS can produce your bill um it's the next few years because there's no  reason it should and I think I think it's uh very reasonable as you see Infosys making this  investment uh some of the Accenture of course some of these other groups if there's a provider  that's that's doing outsourced I.T service work um or or providing integration support  uh generative AI should be part of their value proposition or at v

ery least uh task that  once took a certain number of hours should take dramatically less and I think that that should  begin to show up in statements of work and even in even in social proposals yeah 100 and uh I think  what you're jumping in there is a lot of what our pricing Assurance team is doing right now which is  you know one observation is that we have right now in the market is that a lot of people are moving  to a managed Services construct instead of the tne construct or TNN um and u

h you know there's  a lot of reasons for that and it does tend to be more efficient when you give a little bit more  control to the provider and in a market like this that's that's a good way to shift things  around but it's another out come there that it gives providers more flexibility in  how they deploy these new technologies um to save save costs to be more efficient to  shift some more strategic activities ship their resources around fill in gaps and talent that  we've been dealing with fo

r you know the last couple years so it's a really good point um if  you want to understand that a little bit more we have analysts who can go very deep into what the  pricing models should be like um so before we go a lot further I feel like we need to go into the  elephant in the room which everybody's been kind of calling out the challenges the limitations  the risks of generative Ai and this is a huge issue especially from sourcing where we deal with  a lot of kind of private proprietary info

rmation confidential information that we don't want  getting out of the Internet to our competitors we may not even want it getting out internally right  a lot of the projects that we see the sourcing projects are you know going to affect people's  jobs and the the non-private nature of chat GPT has a lot of people worried there's obviously also  concerns around accuracy and response there's a lot of concerns about ethics and and Edmund I  know you're particularly close to the ethical um topic u

h but before we get into any specific  are these reasonable concerns and I'd like to hear from you how you see especially the privacy and  confidentiality point being addressed over time yeah I think it's a huge point I mean the Vignesh  raised this earlier with a with a certain phone company having um uh you know Secrets kind of get  out into the wild that you can you know if you're on this on this LinkedIn live you can Google  that to check out some of the stories around that um but it is a re

al risk and I think that you  know especially in in procurement I mean the I feel like the non-disclosure agreement is really  a very very standard part of so many interactions that that that we um work on and so to have  a a disparate group of employees with access to Central systems that also have access to  this uh to generative AI systems that could potentially circulate uh that that confidential  information outside of our four walls or outside of our trusted networks is I think it's scary 

for a lot of people one of the things that um just anecdotally I can share is that companies  a bunch of companies have gone out and banned this technology outright and said we don't want our  employees using this well you know I would say like imagine if a company banned using Amazon at  work like all the employees might still use it when they go home and so and like people generally  have their phones with them and so I just think like it maybe a little bit unrealistic to say  that outright b

ands are going to be completely effective as a as a solution long term I think  that companies have done them reactively a little bit to say hey let's just like have a policy that  says we don't want you know anyone using this for anything I think one thing that we'll begin to see  is I think we'll have you know um can you confirm that this response uh that you did not put rrfp  into chat GPT as an RFP question I think that will be like an interesting one in addition to I think  it will also cha

nge non-disclosure agreements I think legal departments are already beginning to  update those to say hey by the way like disclosure means like did you put um information in our  conversation or in our documents into an llm or into a generative AI system if you think  about that that's really new because you know like if you take something from an RFP and you  do a Google search on it you're not disclosing it like that's that is within the bounds of what  non-disclosure typically means however g

enerative AI Works differently it that information can  travel and so that's something that I think uh people are just beginning to to kind of  understand of how public is this information um and I think um you know one one thing I would  be really curious about uh Vignesh uh or or Amy if if you have have a comment on this um is uh you  know some of the the the questions in the chat it you know seem to suggest that um you know I see  one almost zero privacy and security from arnab Chowdhury uh n

o wonder Enterprises or blocking  access that should come with a warning label a statutory warning use at your own risk and so I  wonder um what kind of warning label um do you think you know people that are are implementing  this technology should have to their employees stakeholders or if this technology is being used  with suppliers should suppliers have uh you know essentially have to to agree to some of the risks  up front in order to even use this technology and just to make it even more c

omplicated do  you think bands work are you pro-ban are you anti-band I think that's how you answer that  yeah uh so to your last question I don't think uh bands work like to your point uh I I was  talking to uh one of our clients couple of days back uh she was saying that you know it's  banned but I'm just going to use my phone so I've seen those cases I don't think banks are  going to work and uh that that shouldn't be the way an organization should be taking this  to uh to to their you know e

mployees to me uh for like you mentioned about change management  right uh that's that's critical here uh it is it is very important to educate uh how to use this  right educate the risks and concerns right rather than putting hey don't use it it's here all the  potentials here are the cases that you can use right so that educational and mindset shift is  very important than you know setting some controls or putting a band from a system standpoint  so to me uh mindset change is more important th

an uh you know placing some controls in the  system and secondly uh you're hearing uh you know exploring you know let's control the access to the  data let's let's only you know let jna access data in our private Cloud right let's let's limit  the use cases let's limit the access of DNA to particular data right that's counterintuitive  to me right once you're limited then you're you're basically suppressing the potential of  DNA so I'd say I don't have a clear answer uh but the way Investments a

re happening and uh that  the kind of uh exploration and Pilots that we are seeing uh I believe soon uh as an industry we  will figure out how to you know manage this confidentiality data but putting in security  I think the big text hyperscalers will have a huge huge role to play here yeah I think I  think that's right um I think we all tend to go to chat GPT which is sort of a public platform  and we'll see more of you know what what I would you know years ago referred to as a Walled  Garden i

n the early days of the internet but um you know there are so many applications where  you can have your own internal to your company um chat GPT or sorry not touch it but generative  AI Solution that's you know just Gathering internal processes from a supply chain perspective  or all of the sort of judgment related things that we're doing on a regular basis um aggregating that  information that sort of tribal knowledge as we've often called it there's a huge potential there to  do that internal

to a company as well and you're not facing those risks um Edmond I laugh a little  bit when you talk about Amazon because there was a point and there still is where a lot of companies  say if you want to buy anything you have to go through our Erp or through our I'm not going to  name names but you know whatever P2P platform for requisitioning um and how well did that go right  I've seen all the Rogue spent data out there and very few companies can get above 70 or 80 percent  of their spend thr

ough the contracted processes um so it didn't work people still went to  Amazon and bought what they wanted or called their friend down the street or whatever and  it's not going to work here either that's just not how very few companies now are in a command  and control major that's going to work that way so I do think we need to figure this out um I  want to put a plug-in that we do have a paper that you can download here a generative  AI viewpoint on the new age of artificial intelligence and

the link for that is in the  comments or in the chat um so uh feel free to go click that it's a pretty interesting paper  and it goes you know goes through this from a broader perspective kind of that how do we  help our organizations view for sourcing um so I still see a lot of questions we don't  have a hard stop on this but I do want to ask um you know let's wrap up or start to wrap up a  couple minutes uh with kind of a just a general question in terms of how much of this is the oh  current

hype is kind of buzz and it's going to settle out it's going to be you know people say  metaverse is dead I'm not convinced it's dead I still think there's a lot of applications but the  b2c part maybe is not as exciting as as it might have been made out to be um but how much of it  is real and will gain traction and I think that there's a question here on the pros and cons and  I think that kind of aligns with that like are the pros big enough that this is going to be something  that sticks uh

who wants to go first yeah I'll I'll I'll share my perspective I think uh to that  question definitely you know pro software cons uh in terms of whether this is a buzzword going  to settle I think yeah Edmund mentioned earlier right we are in the hype cycle it will settle down  but it will not drop off I don't see Jenny as uh you know all together a different technology that  that's going to change the procurement landscape again echoing admin's point I see Jenny as a value  amplification layer

right which will amplify the value of the Investments that you have made  if you have invested in analytic solution it's going to amplify the value of that if you have  invested in an automation platform it's going to amplify the value of that so it's not going  to fundamentally change your process change your uh ecosystem it's a value amplification  technology so it's it's going to have impact Advent yeah I I would agree with uh with Vignesh um it's uh it's funny uh our Orchestra is uh  like o

ur company mission is to amplify the value wrong person to ask word tries to have but  yeah no I look I I agree I think it you know procurement's been on a digital transformation  for a while now uh with I think some fits and starts and and varying degrees of of success  the only way that you could have been wrong consistently for the past 20 years would have been  to like bet against digital and in favor of like we're gonna keep doing it on paper and it took  much longer than a lot of people th

ought but if you look at source to pay adoption where it is  now um you know I like most companies are very comfortable doing business electronically doing  business online uh in certain industries there's there's still uh opportunity but I think that you  know to vignesh's opening point about Clippy in in you know Windows 98 like this like like you  know what's what's old is New Again like this technology has already been around it's already  been transforming the way we work for a while and I

think the question like in in my mind I kind of  have three questions that I I would I would leave is like open think about for for people who have  been marinating and digesting on on some of the the information here one is does generative AI  like if you have a co-worker that you've never met and like you interact with them over zoom  and you email with them and like they have to you send them information and they analyze  it and they send you answers back like could generative AI essentially

perform some version  of that for you and be more cons like responsive and consistently available then certain not all  but certain co-workers that you've had like the experience of working with and so like that's a  it's a little bit of a provocative statement but I I ask that because I don't know what the answer is  I don't think anyone does and I don't think anyone even the people working on uh generative AI fully  understand exactly how or why these models work or what precisely what causes

them to mature we have  some sense that like the number of parameters is important like yes a trillion parameters is like  better than a billion great and like you know more uh more Loops more like reinforcement learning  human feedback loops is good like so if we have have it order of magnitude more the model will  be better generally people are confident in that but the the inflection points and breakthroughs  are very non-linear and so I think um you know if that happens it will happen for  d

ifferent people in different areas and different Industries at different times and different  geographies uh and and and so on I don't expect it to be like a massive everyone adopts it all  at once kind of wave but I asked that question like would you would you have an AI co-worker  I don't know I think that really depends more on emotional comfort and culture than it does  on um the technology the second question is um Can AI help me do something sooner than  I would otherwise and I say sooner

because I think we often get it focused on efficiency and  Effectiveness and to an extent speed is part of both but I think for procurement and supply chain  getting involved earlier is so valuable and so if there's something where where generative AI could  be in a a warning system for something that we would otherwise have to monitor manually or find  out through data analysis I think that could potentially be disruptive for procurement um and  so I think like you know different Industries wil

l look at this different ways we have some customers  in the automotive space who really care about the availability of certain components it could be  easy to set up a very simple monitoring system for those components that can be very custom built  for for that need chemical industry uh same same deals and then the the third question I would ask  in this again these are just questions to think about and and take away is hopefully useful um  but um you know just just kind of think about is um i

s there like a task that's really unpleasant  that that you have to do all the time you're like you have to pay someone to do all the time that  involves document creation and if there is then like could AI begin to do that because like the  number one job that we're seeing get disrupted in this space is copywriting which is so strange like  if you ask people 10 years ago what jobs are going to be replaced or disrupted by AI um I think a lot  of people would have said oh like things that are ver

y rule space like accounting or um you know  kind of back I mean purchasing was on the list um but like it turns out AI is really good at  writing poetry and making memes and like doing weird art and so like that was not something that  people 10 years ago really foresaw as being like an application because those are all really  creativity involved practices so I I like if you're paying someone to create copy it's part  of your function like that's something to look at and I think there's um Joh

n Oliver did a great  uh segment on AI where you know they interviewed lots of different Industries and they talked about  AI legal and one of the things there was a comment from that episode which is I don't think AI will  replace lawyers but I think lawyers who use AI will replace lawyers who don't and I think that  that's a really valuable frame like I don't think we should be like afraid of our afraid that AI  is going to like take our jobs but I do think that there's so much leverage and Ti

me Savings  there that I think the procurement teams and procurement organizations and of course Outsource  service providers that leverage this technology will have a compelling value proposition um and  potentially uh be able to get more Market sharing foreign yeah I think those are all  really good questions to think about um I I have heard the advice given you know if  someone's coming out of college now um will they have a job because of AI and then the answer was  well they need to learn t

he right prompts to use AI to do their job better right so it's not them  against the AI it's like how do you use it better um and I I think the concept of you know  think of the task that you most hate doing um and see how you can use that that kind of goes  back to my writing descriptions for webinars and Linkedin live things hate doing that it's  not that hard just always the last thing on the list but there's a lot of other things  cleaning my house I agree Susan I've got to uh robotic vacuu

ms now that do that so that's  a huge win for me um but so I'm gonna wrap up um any final comments from both of from  either of you before we we close out today thanks for a great thanks for a great discussion  yeah this is it's very great to see what what Everest perspective on the topic is and uh yeah  just lovely to to chat with with uh with both of you yeah again I got that great discussion  and uh thank uh all the participants uh very engaging comments and discussions I thoroughly  enjoyed

this probably if you do this maybe six months from now a refresh of this it might be  very different yeah it's fascinating because six months ago we would have been talking  about it um and I'd love to hear where we are in six months I want to thank everyone who  attended we had really good uh participation um great comments I think there's you  know this could go on for three hours questions about costs questions more about  you know the ethics uh accuracy all of that obviously there's no end t

o the conversation  I want to encourage you all you can reach out to any of us uh you've got here through Linkedin  so you have our our contact information that way um also for the average group site or the  orchestra site um I see there's a question for orchestra and I'll let Edmund take that  offline um but uh thank you everyone appreciate the participation and uh look forward to  continuing this conversation in the future and thanks uh Edmund and Vignesh for giving us  your time thanks Amy th

anks so much thanks so much yes