How to build an AI-ready organization: Cultivating a Data-Driven Culture

Key Takeaways

hello everyone welcome to AI anytime Channel and in this video we are going to have a very important discussion okay on a topic and the topic you are you have already seen uh from the thumbnails and the video title the topic is how to become an AI ready organization now in this video we are going to outline the building blocks you know for developing an AI ready and capable organization and you know with this video uh you can jump start your AI strategy nowadays you will see there has been lot of development you know in the AI community in in this industry you will see AI is booming like anything you see multiple genitive AI models for example large language models and vision models are being you know developed and launched you know on you know daily basis you know in the last few months for example you have seen uh revolutionary chat GPT you know and there are a lot of other open source model llama was launched by Facebook you know you see recently Microsoft has launched a Jarvis so there's a lot of development you know in the with this technology and the technology has evolved over a period of period of time you know so we'll discuss this in detail that how an organization can you know leverage AI as a technology to innovate further okay innovate you know for innovate in future for their clients and partners how can they innovate with AI okay and how can they become an uh AI ready organization okay this is what we are going to discuss in this topic so when when I say a company is AI ready you know they are able to take an idea from conception to implementation you know to ripping benefits all with minimal friction so it's not easy to become an AI ready uh organization because I will discuss that further why I am making this statement because what I believe and I also have discussed with my fellow you know uh researcher and friends in in my peer that there there's a huge Global AI illiteracy okay and that's what I'm going to discuss you know today that how you know we can consider few of the fundamentals you know Tropics or fundamentals points basically that will help you you know transform your organization and the first thing that you see you know currently on my screen and I'm talking about some points here you can see I'm talking about budget culture infrastructure data and skills and on top you see something called be kids that's how it gets pronunciated because we're calling it B kids then we're talking about budget culture infrastructure data and skills these are very important if you really want to make your organization an AI ready okay and all of them are too big to discuss and in this video I'm going to discuss something on culture okay we'll discuss a culture in detail that what I mean by culture why a culture you know establishment is uh very fundamental okay if you really want to make your organization an AI radio organization you know and you don't get uh to this stage overnight when you have to become an AI ready organization like you see uh companies for example Google okay they have become an AI uh driven organization I will say that they have this ability that they can you know develop uh or they can develop or they can deliver large-scale AI projects of course for businesses and of course directly for you know consumers as well and they haven't become this you know uh haven't acquired this ability overnight you know they have developed it because AI ready requires a level of preparation you know and that is rarely discussed okay uh you know within community and when I say that it requires level of preparation it's not only the cost okay it requires you know investment in your company culture you know talent pool you need the right Talent you need the uh data you need the infrastructure it's very important and it will even eventually change the way you do the business okay and it takes time it's a long term investment and now is the time now it's now it's very crucial time for an organization to focus where they want to innovate in future because AI is developing with a rapid pace and it will be here AI is to stay and this is what you know we are going to discuss so let's discuss you know at a very high level you know uh the pillars that you uh you have seen in the previous slide that budget culture infrastructure I in this video I will discuss culture so when I discuss culture the first thing that I would like to discuss is establishing an AI literacy because I believe that there is a huge Global AI illiteracy and these organizations the the group of developers or you say researchers they take you know advantage of this illiteracy okay mainly the uh this large organization so it's very important to create this cultural and you know mindset shift okay for embracing Ai and that's where this comes establishing an AI literacy okay so you know literacy is the key uh for sleeping employee attitude towards AI this is not just about training your technical teams it about about most employees in your company if even if they are not part of your technical teams you know they should have a base understanding of AI and B able to answer what is this thing okay how does it work and what is the company planning to do with AI okay or how does it impact job security for example where because you know currently in the industry if you see there is a lot of discussions going on that AI will take over human or AI will kill you know some set of jobs you know it's you know it's debitable we can debate on that but I'm not going to debate uh on that topic but this is so important because you know it reduces fear when you create this kind of literacy when you create uh this AI literacy you know it reduces the fear and closes the door for false interpretations you know there should know there should uh there shouldn't be any false interpretation within your team or within your organization across organization that you know uh whatever that we are discussing that AI will take over humans or AI will take your jobs okay without this base AI literacy you know once employees hear about your company's plan to integrate AI they may take their own assumptions you know they can assume you know they take actions that can negatively impact your company's reputation you know just to give you an example now if if your employee think that they are going to lose their job they may plan a protesting you know in reaction they can start protesting against you know some male League uh information to the media based on their own interpretation of what you are really going to do within with AI what as an organization you are going to do with AI this information can be leaked and this has happened you know recently uh uh in you know in in uh in the United States there are a lot of data that has been exchanged to Media you know based on employees interpretation that what's company is really going to do with artificial intelligence given the fears and confusion around artificial intelligence it's important to address uncertainties with company you know a wide education and planning is very important so that's a very fundamental and the very important point before we go into uh this further discussion that it's important to establish AI literacy if you have you know uh if you have a team of five to ten people and they they don't understand AI at least you can you can arrange some workshops you can arrange some kind of you know boot camps you know for your team members so at least they can understand the fundamentals and basics of you know artificial intelligence which is of course a very vast subject a very vast branch of computer science but at least to create this literacy it's very important now excuse me so the next point is to make leadership data Savvy okay you know we have this we just discussed about AI literacy okay data is uh uh the core of AI it's very uh very important you know to because we have abundance of data and just to give you number you know on day to day basis we kind of generate 453 exabytes of data you know and it's abundance of data are sitting you know on the cloud or wherever whatever kind of knowledge or databases that you manage now data literacy is also very important okay when I say make leadership data Savvy you know it's important to understand that data literacy does not stop with data scientists or senior leaders or you know HR managers or sales and marketing persons anyone making business decisions and managing Innovation should know how to use and interpret data data is the most fundamental thing you know in this current world you know if even if you're not talking about technology just take uh this you know this universe as an example data is the most fundamental thing currently and it's very important for people to understand how to use and interpret those data only will this help with you know with strategic decision making anchor on data but it will also facilitate the adoption of AI you know as many decisions around AI are inherently data driven okay when we talk about AI most of the time we talk about this data driven intelligence or data driven insights you know when it comes to data literacy industry based practices to exemplify from the top okay as more and more leaders use data uh to drive decision based on just gut feelings the time has gone where you know as a leader you go with the gut feelings okay uh x x will work or the Y will work just because you know I'm getting this message from inside most of the time it will not work through the time it might work you know of course experience helps you know but that's also a data okay so it's very important for you to understand the use of data how to use data and how to interpret data how to make decision based on those data as a leader I think it's very important in in any organization whether it's a technology driven organization or any organization you know it's very important to understand if I go next it's be ready to experiment uh it's one of my favorite you know because be ready to experiment and brace for uncertainties okay uh experimentation and iteration are critical parts of you know machine learning which is a subset of which is subset of artificial intelligence and nope the ML and AI has been used interchangeably people have you know have some misunderstanding about the both the keywords but you know uh experimentation and iteration are extremely critical part of the machine learning life cycle or the ml development life cycle you know it can evolve it can involve going back and forth multiple times to tweak your model so if you are training and a machine learning model on uh vast amount of data and you might not get the desired response when I say response again it's a uh that can be any evaluation metric the way you evaluate your model if you are not uh getting desired response after training a machine learning model you have to perform several kind of tuning uh you have to tune the model again and again you have to go back and start doing it uh have a look at data again and then you have to make better decisions while training the model you have to tune the parameters so it's a it's an iterative process okay it takes uh time you know and you fix issues during development as well and you know it can uh also include iteratively making changes based on your testing the way what uh doing this testing you may also perform you know a b testing of completing solutions to decide which one to use now you have you can have multiple versions of your model and you can perform this a b testings to decide which models uh which model to use and you know all this takes time okay uh so you have to be openness to uncertainty okay and the ability to interpret and act on data that we discussed earlier so you know building a culture of uh experimentation in your company requires uh that you provide sufficient time you know for multiple rounds of test and iterate most of the time as an individual and also when I work with my team now I have failed most of the time rather than you know getting success now this has happened and that's how you know we all learn uh and mainly when we are working with uh Ai and machine learning models most of the time we failed we don't get the desired response you know and then we go back we tweak it iterate it and over a period of period of time we get this uh desired response that we want okay so and most of the time you know you should you should also remain open uh to the possibility that not all projects will pull through okay most of the projects will you know get failed and only few of them will get uh get this get succeed and you know given this uncertainties it's it's crucial to incorporate early test thing okay when you are you know creating this machine learning life cycle you it's crucial to incorporate early testing and always have a plan B mapped out you should always have a plan B when you are trying to work on a large Ai and machine learning driven project and you should have a plan B in place and maybe you can parallely also keep a track on that as well with your team members so and now we come to build cross functional team uh which is very important you know across uh organization I'll discuss it why because integrating AI into uh your products or your canned products to your partner's product or the business processes is truly a team spot okay uh just you know a bunch of data scientists will not be able to uh not be able to do the Justice you know to take the models in production so thousands of people or millions of people are using your product at the same time and you have uh Ai and machine learning models working on back you know let's take example of you know Twitter has recently released their uh uh recommendation engine algorithm that how they have built this recommendation model okay uh for Twitter and there are a lot of other recommendation models available on Tick Tock for example having one of the world's best recommendation models now when you see this uh apps like already platforms like hotstar or Netflix for example you know millions of consumers millions of end customers are watching the same content at the same time how do they handle all these you know latencies and traffic it's very important to understand and also in the machine learning life cycle only when I talk about ml Ops for AI Ops in general you know so you need set of people you need data scientists you need you know a business stakeholder now let's start from a business stakeholder you know this can be a product manager this can be an SM in your subject matter expertise or a domain expert or an AI project manager who understand in and out okay so you need a business stakeholder you need an AI implementer or an AI export and this can be a machine learning engineer or a you know senior data scientist or an AI consultant and you also need one or more Engineers this is very fundamental you know if you if you're really finding any interest in this video you make note that you need data Engineers to work on your product or whatever project I'll say that you are working on you know with AI and machine learning if you're really looking to deliver a scalable solution you need data engineers in your team guys okay who can work with this data scientist or software Engineers you know who really help with the data pipelines development of these pipelines and model integration you know with the data feed that we are getting so it's very important and also it at a minimum you should have this set of people or business stakeholder uh set of data scientists or ml Engineers or AI consultant and also couple of data Engineers or software engineers in your team Okay so you know if you if you are used to operate operating in silos then AI initiate initiative will remain as prototypes in the hand of data scientists building a prototype is a very simple thing you take machine learning or AI models you have so many open source models now if you want to really build a sentiment analytics or summarization you know POC a proof of concept of prototype for your client it hardly takes you a day or two to build it with so much of Open Source support now we have you know uh in this community you go on hugging face which is one of the largest deep learning and NLP Community you know uh you go and use utilize their open source models and you create some kind of POC and you show it to your client it's very easy to do that but when you are really building a solution which requires multiple Ai and machine learning models to work you know in work in a single platform or a single tool it's very important okay so you know and you need this kind of collaboration if you have a large corporate corporates if you have a large organization you need collaboration to work on okay so lack of collaboration is a big contributor you know to AI initiatives failure guys especially in large organization you know either no support is available from data engineering or software engineering team they are too busy to incorporate uh new models but if you generally want to become an AI ready and capable uh company you need the right team to function right team to work and you don't need a hundred of people if you take examples of company like mid Journey you know mid Journey that AI image generation company there are only nine people eight to ten people hardly who who is working with a mid-journey okay and you see that they have created this you know revolutionary AI image generation platform okay and the service that they provide through their Discord server it's it's amazing to see you it's only eight to ten people you see when open AI started you know a few years back there hardly few people were working for them now take example of Deep Mind they don't have thousands of people working for them they have very limited people but they have the right team to work on they have ai researchers they have they know they know clearly what they have to do and they have a clear vision of how they have to work with this technology as an AI first company or an AI radio organization I'm talking about deepmind and open AI who are fully AI based companies they are AI companies by the way so they know what they have to do they know their research areas they have those set of team they have ai researchers subject matter expertise they have data Engineers they have senior machine learning engineers and of course they have this uh people of you know technologists who are having vast uh technology knowledge of of you know more than 10 to 15 years as well so this is very important to have the cross-functional team now most uh most important I will say you know in the last few years uh is having an ethical committee okay uh at your uh organization it would happen uh ethical and accountable committee okay and there can be you know multiple case study that can be discussed you know to to prove this that you need this an Ethics Committee you know mainly you know most of the time you talk to cxos they might not like the idea of having an you know ethical committee but it's important the the way AI is being developed over the uh last few years if you see now I believe that it's now the high time that we need regulations and if you see example that how Italy has banned chat GPT it's no brainer that governments or these authorities now Europe European Union have their own protocols and standard guidelines in U.S that dapa has their own you know different organization has their own uh ethical guidelines and it every organization you will see a large corporate they would have their ethical committee but you know I'm talking about your team because in large corporates you might not directly get you know any interaction with this committee so you need people who understand uh ex these components of ethical AI I'll just give you an example you know it's very famous example and you you will you will remember what I'm talking about so in early 2018 18 we saw how you know uh Facebook now meta has mishandled data you know from you know millions of users I'm talking about Camry's analytica case study the cameras analytica case you know political data analytics form okay uh had harvested the personal data of Facebook users at that point of time without consent and Landing meta in a whole lot of trouble okay and the whole drama I'll say cost uh meta more than you know the funds were in billion okay I think it was around 5 billion US dollar that meta has to pay uh in fines and you know they didn't learn their uh uh Facebook uh now meta in late 2020 you know they uh meta settled a lawsuit of uh around 650 million US dollar you know over another data privacy issue and it was more related to their uh biometric Privacy Law you know in in North America and United States that we have you know of from their photo tagging feature now Facebook or meta they have this photo tagging feature and you just facial recognition software to identify faces in your photos and you know they have a law against businesses collecting biometric data I think it's in state of Illinois if I am not wrong and so these are the example and I I know I as an AI uh practitioner I do believe that you know ethical use of data is extremely tricky I know I discussed with my peer that we talk about buyers we talk about you know uh transparency mainly then we talk about bias I always say that bias is inherent in my thought okay it's very difficult to uh you know take care of bias you know it's not impossible to eliminate buyers when you are uh working with uh this technology but what I believe that our first Focus would be on bias in data rather bias in modeling I think the data is in bias rather than directly going and jumping into the modeling part and saying okay I will eliminate the bias I will minimize the bias let's have a look at the data you never know you know when they label or annotate data how have they annotated it you know how are you validating those abundance of large set of data so they're all discussion that we can have it separately but you know what I was saying that it ethical use of data is extremely tricky and when you combine that with AI okay you know with little regulation it gets even trickier Okay so you know so it's very important now to make to create these guidelines okay if you really want to maintain good standing with you know customers and Society at large okay because we talk about AI for good AI for society AI for social good you know I'm also talking about uh that your organization might focus on Society at large right there's a contribution that you have to do it's important for companies to take a strong stance on where they stand in terms of data privacy ethics and accountability okay one data will you know the company absolutely not use for AI or any type of decision making what kind of models will the company not build you know should users opt-in or opt out of features that uh feature that use their data you know so it's very fundamental it's very important to understand uh that uh rather than let's take example of deep learning models or large language models which are on trains nowadays we have this large language models coming in for example you take GPT three uh uh you know gpt4 also which had been released by open AI it's very difficult for AI researchers or the Developers explain those models explain those outputs to the end consumer or End customer that can be anyone that can be one more organization as well that can be an individual as well how do you explain the response or the output that you are getting from these large language models you know deep learning models are you know you know complex to explain okay people have seen the cxus have seen AI mainly as Black Box you know years ago but uh how would you explain okay so how on who will be the accountable who will be the responsible how responsibly you are using uh you know AI so you know these are the few questions that we first have to answer you know uh internally or the team that you have you first have to decide what kind of data really you you will use what kind of data you will say no and these set of questions so the other thing that you know we I wanted to discuss create an Ethics Committee build cross-functional team be ready to experiment make leadership data Savvy establish AI Literacy for all the images that you see is all these images are again being generated by a mid journey and this is all Air Generation and culture is at core guys okay if you really want to you know make your company an AI ready organization okay it's very uh fundamental to focus on culture you create AI literacy you help uh you Empower your colleagues or your employees so they become more literate you know with this technology you know make leadership data savvy your decision should be based on data it should be data driven it should not be you know emotion driven most of the time okay when you are making decisions a few other times the things are not in your control that you you don't have data at your disposal or you don't have these insights at your disposal so you make decisions with you know that comes from your experience you know and that will always be there and now be ready to experiment you have to be you have to keep an open mind when you're working with AI and machine learning and you know it takes time you know if you really want to work with this technology and then we build cross-functional team as I said you have to you know build a cross functional team you need instead of software Engineers you need data Engineers you need data scientist or senior machine learning machine learning Engineers you need AI consultant and you need a business stakeholder that can be a subject matter expertise or domain expertise okay and then create an committee who will look after uh who will set the guidelines or look after your ethical and accountable or ethical components in general that we have multiple component that we can discuss that can be you know explainability that can be accountability that can be transparency buyers governance that can be anything okay that can be discussed so you know I discussed this five elements you know for establishing culture you know or cultural readiness uh for AI in your organization and you know with this five elements you now have a head start uh in that journey and once you get into this uh you know get into this thick of each of the element okay you will find more areas okay uh for cultural improvements and you have to tell me okay because this are all my findings uh you know I have been part of this AI machine Learning Community for now almost six years now I have been learning and these are all my learnings you know I've seen this across startups across you know these large organizations and now if you are you know considering this you know factors and if you find more areas for cultural Improvement please reach out to me please let me know your thoughts and feedbacks in the comment box you know um no I'm open to feedback as well I would like to understand what kind of challenges you are facing and how you overcome those challenges to when you're working with this Ai and machine learning projects large scale Projects please let me know I think that's all for today videos guys okay I hope you like the video if you like the content I create please you know subscribe the channel if you haven't subscribed yet and please do share this Channel with you know your friends and to peer that's all for this video guys see you in the next video