AI Virtual Workshop Series - Preparing the Organization for AI Adoption (1 of 4)

0:02 Okay, good morning everyone. My name is Margie Graph and I'm the president and CEO of the Greater Rockville Chamber of Commerce and I want to welcome you to part one of a threepart AI series and I'm busy letting everybody in the room while I'm trying to talk. So anyway, I would like to um thank Pam Clark from Clark Communications uh for sponsoring our series. Thank you so much, Pam. We had our AI uh live kickoff at Bowero and now uh we're following up with a three-part uh webinar, AI series webinar. So, let me introduce you to Mia Caniti. I hope I

0:50 said that right. Mia is a data scientist, entrepreneur, and early stage investor with a track record of co-founding tech companies in India and the US. She is the co-founder and CEO of Venture Factory AI, a global AI venture studio and leads Beyond Z LLC, which is a US-based digital transformational firm. As joint managing director at Venture Way, she drives global tech investments and startup ecosystem growth. An active angel investor, she serves on the board of governors at Sparks Angel Network. Mia holds advanced degrees and certifications in AI and

1:35 data science from Sunni Buffalo, MIT, Stanford, and Harvard. She's also the founder of Can you say what foundation that is? this get AI uh it's growth and empowerment through AI okay which supports education for underprivileged children in India and is establishing the gete geti get foundation in Geneva to promote global empowerment through AI. So, we are thrilled to uh have Mia present to us and thank you so much for taking time out of your traveling schedule to present. So, I'll turn it over to you, Mia. Thank you so much, Margie. It's a

2:21 pleasure to be here. Thank you, Pam, for making this possible. Um, and great to see you all. So, I'm going to share my screen and we'll get started. All right. Is that good? Do you see the slides? Okay. So, today um I'm going to talk about building your AI foundation as an organization or as an individual. How do you start engaging with AI? Um and we we talked to a lot of people. We talked to a lot of organizations and what we see is that um you know people want to get

3:07 started with AI but they're just not sure how to go about it, where to start. So today I want to talk about three levels of engagement with AI depending on your level of um you know appetite for the technology, your budgets, the size of your organization. There are different levels at which you can engage with AI. So I'm going to talk about uh three levels of engagement. Um and then we'll get into lowering barriers for AI adoption. Why it has been getting easier to adopt AI, organizational structure when you start uh adopting AI in the organization, what kind of behaviors lead to successful AI projects and then we talk about the importance of a well-considered AI adoption roadmap. So

3:55 that's the agenda for today and um I'm looking forward to your questions and answering them. I think at the end of the session at as Margie mentioned um so here we are so three levels of engagement um so one is offthe-shelf applications that's the lowest hanging fruit um it's the easiest way to engage with AI there's not much investment there's not much upskilling or training required these are just tools that are out there that you can just start using off the shelf the next level of engagement is targeted AI implementations these Uh again localized implementations you have a certain business goal or a business issue and you solve for it. You custom build an AI solution for it. So it's targeted and

4:41 it's localized. And the third level of engagement is an AI first approach where AI and data analytics is at the core of what drives your company, what drives your business strategy. So, we'll look at each of these levels of engagement um just briefly and then we can get into it in the Q&A if anybody would like to know more about it. There are many off-the-shelf uh applications out there. I'd just like to mention a few. The chat bots of course we're all familiar with. Chat GPT is probably the most popular from OpenAI. There are also various other chat bots like Claude, Meta, Gemini from Google and more. And what these chat bots do today, they've evolved also over time. And what they do

5:27 is they allow you to reason over content. They allow you to summarize content. Um, if you have Excel sheets, you can do basic analysis on your sheets. You can generate images. There's a lot you can do. You can generate art, you can generate poetry. There's a lot you can do with it. Uh I know that a lot of people also use chat GPT or chat bots as search engines for web search. However, for web search uh perplexity is a great tool. Perplexity.ai uh it searches across uh websites and gives you a summarized response and the responses of perplexity are much more grounded. So uh you might have heard about hallucinations where these chat bots do give out information that is not

6:13 factual right but perplexities responses are grounded in the uh uh results from the web search. So for web search I think publicity is a better tool and nowadays in Google search results you might have noticed that before the links appear there's an AI overview that comes up that overview is also AI generated for marketing people for you know your marketing content generation for measuring your marketing there are various tools Jasper is a very popular one there are grammar checkers and rewarding tools that have been around for a very long time video creation and editing Runway is a very popular tool. There's Sora from OpenAI. These tools allow you to uh give text prompts and generate

6:59 video content for image generation. There's Dali, there's Midjourney, the stability.AI. In fact, Dali is uh developed by OpenAI as well and it works with chat GPT. Dali is the image generation tool behind Bing. Uh so you would have you might have engaged with Dali. Um there are a lot of design professionals who use midjourney extensively because it does generate highfidelity images. Uh so if you have marketing content for which you need images, midjourney would be a great tool for voice and music generation. Uh there are a lot of tools out there. Social media management which is becoming increasingly important for organizations today. Uh so for managing your social media feed, for generating content, for

7:44 generating blogs, there are a lot of tools. Knowledge management is another great application of AI. I'll be getting into it in detail a bit later. So these are some of the shelves tools that do exist, but I'll talk more about it later in the presentation. There are transcription and meeting assistance, um slide decks and presentation to generate beautiful slides to uh lay out your slides. Um there are tools that help with that. So what are some statistics? Uh Coca-Cola used a chat GPT based uh uh assistant for content creation and what they found is a 90% reduction in content creation cost. Uh and one of our portfolio companies, Venture Way, we use

8:31 a custom GPT as a communications executive. And what that allows us to do is to allow our uh our resources to be able to provide content at a much higher level than they would otherwise. Also in keeping with the company's tone, the company's voice, uh the kind of content that as an organization we would want to put out. Another portfolio company of ours, Lima, we had a code base that was in C and we wanted to port to Python. um and we didn't have a lot of Python resources but with chat GPT and the resource that we had they were able to do the code migration. Um so an MIT study actually said that using a chatbot as an assistant decreased the time it

9:17 took workers to complete certain types of tasks by 40%. And the output quality rose by 18%. So those are the kinds of benefits you can get by using just off-the-shelf tools and not really custom building any solutions. So this is definitely um lowhanging fruit but also uh good to engage with AI at this level as well. The next level of engagement is targeted AI implementations where we custom build solutions based on your data based on your organization's data. you build solutions uh for a certain specific goal and these solutions can come under various categories. They could be NLP which is natural language processing or predictive analytics, AI

10:03 agents, it could be computer vision. This could be a range of solutions and I'm going to just touch upon a few. So knowledge based management, I alluded to it earlier. Um, what this is is when you have an organization with a lot of internal documentation, it can be hard for your internal resources to know where to look for the information they need. And it can be difficult even for especially for new hires to find the information they require. But what you can do is build a conversational agent, a chatbot that sits on top of your documentation so that your resources can engage with this chatbot can ask questions. It will reason over your documentation and provide a well-grounded answer. So hallucinations

10:50 are reduced as well. Um so it's a great tool to have. Microsoft copilot does knowledgebased management. If you have the enterprise edition, it's a great uh it's a great addition. So, some examples, the International Labor Organization built Answermate, which is a chatbot that sits on top of their internal documentation, and they found really significant improvement in productivity. JP Morgan and Chase is giving its employees an assistant that's again powered by CHPT and they find that their workforce is better equipped on account of it. You can have NLP solutions that are external facing as well for your customers and your clients to engage with. You know, it could be product information. It could be service related

11:37 information depending on what your company deals with. You could have a chatbot that sits on top of your documentation that your external stakeholders have access to. And this can significantly reduce the load on your customer service team. Predictive analytics is another class of solutions that have been around for a long time and they have great benefit. Um, essentially in predictive analytics you are trying to predict future outcomes based on historical data. So it really transforms decision making because now your decisions are really grounded in the insights from your data and predictive maintenance is one of the main use cases that have been around for predictive analytics where

12:23 you intervene for maintenance as needed rather than doing scheduled maintenance, right? you monitor your uh equipment and you have data that's coming in and you have an AI uh model that's running on top of it on top of that data that then tells you which parts or which machinery might require intervention. And predictive analytics has use cases across multiple industries. In healthcare, it's used to predict patient outcomes. In supply chain, it can be used to predict and then prevent disruptions. In banking, it has been used, it is being used to detect fraud. And what this allows you to do is to monitor every transaction as it happens real time, 24 hours a day, 7

13:10 days a week, which is impossible to do with a human u workforce. In marketing, it allows you to understand your customers better and to segment your customers, to target your campaigns. And we'll be getting into the details of customer segmentation and targeted campaigns in uh in the following workshops a bit more. So this is a quote from uh the VP at Gartner saying that information is the oil of the 21st century and analytics is the combustion engine. Um so another thing that we see is that a lot of companies have a lot of data but they're not yet doing anything with it. So the idea now today is to run analytics on it and gather insights from your data and then let that drive your

13:56 business strategy and your approach to your consumers. A few examples of uh predictive analytics from industry. PNG uh used AI in in logistics and supply chain and they saved about a billion dollars. Um JP Morgan and Chase used ran a pilot program where they used predictive analytics to understand what kind of content their customers are likely to engage with and they found a 450% increase in click-through rate. Mckenzian company estimates that using predictive maintenance will reduce unscheduled breakdowns by up to 50%. So those are the kinds of um benefit you can have from targeted AI

14:43 solutions. Another class of solutions that is gaining popularity today is AI agents. Essentially AI agents are AI models which also have access to external tools. So these external tools could be email clients, it could be Google maps, it could be a weather service depending on your use case, depending on what you need to do. And these AI agents are trained to know when to call these external tools. So you could build AI agents with different levels of autonomy. Um and essentially they sort of supercharge your AI models by giving them access to additional capabilities. So an example of AI agents is uh Clara which is a financial services company based out of Sweden. um

15:30 and they built an AI agent as a customer service agent and they the agent has had 2.3 million conversations. The beauty of it is that these agents are available 24/7, 24 hours a day, 7 days a week in multiple languages across time zones. And they found that the quality of these conversations was at par with those of their human representatives. Clara estimated a $40 million profit last year. So what are some of the key considerations when you're looking at developing targeted AI solutions? One is what are the opportunities in your industry? What are your pain points in your company? Um the other is what are the opportunities in your growth trajectory? Where would you like to go?

16:16 And the most important consideration is what is the state of your data currently? Do you have the data that you need to build the solutions that you require? and and then as I mentioned um you need a well-considered AI implementation roadmap and we'll get into the details of that further in this session. And the third level of engagement that we could consider is an AI first approach. This is an all-in approach where AI is core and center of your business strategy. How do you get there? So this is um a statistic of sorts um that Forbes is saying AI first companies are outpacing rivals and redefining the future of work. How do you become AI first? You need a

17:03 centralized data strategy. You need a centralized data. You need to restructure and reimagine your operations to be able to fully leverage AI. And it allows you to build crossunctional and interdep department solutions. And we'll see how in the following slides. In this age of data, one of the significant insights is that the more data you have, you're able to train better algorithms, better models. And when you have a better trained algorithm, you're able to provide better solution, a better service. And when you have a better service, more people are likely to use it, which then generates more data. And this is a positive cycle that uh continues and repeats.

17:48 How do you go from legacy systems to an AIdriven architecture? So even uh established companies that have been running for years and they have legacy systems, they can make this transition. Um how do you do it? How do you go from a siloed data architecture which is what many companies have today? You know you have every department has its own data. The finance department has their own data. The marketing have their own data. Um R&D have their own data. Procurement has their own data. But these departments don't talk to each other as they should. So the idea is to go from this siloed data architecture to a centralized data architecture so that then you can run your AI ins AI and analytics on it drive insights from that

18:34 and then let that drive your business. So the idea is to go in phases. The first step is to run a pilot to choose two departments, get data from those departments and combine them. And in doing so, you will figure out what is the kind of data architecture that suits your organization, that suits your business, what are your data security needs, what are your governance, data governance needs, what are your access requirements for the data. So you get all those moving parts sorted and you combine the data from two departments um and run that as a pilot program and once you have that done you can then pull in data from all the other departments in your organization and then build the central data repository and on top of

19:21 this repository you would run all your AI analytics which would give you insights and then you would your strategy would be driven by that. So I wanted to give you an example of a company that has that is AI first and the kind of growth that that enables. So this is um it's plotting number of bank accounts in millions and we're comparing banks. Um and the green curve that you can see shooting up 2015 exponential growth since then is a company called Ant Financial. um it's a spin-off from Alibaba and the other companies there's Bank of America, Bank of China, City Bank, many of the banking giants. So you can see that Ant Financial has grown in terms of number of users significant

20:08 exponentially. And to put some numbers on it to Antfinancial services over a billion customers. It has less than 20,000 employees. In comparison, Bank of America serves 69 million customers, but it has over 200,000 employees, right? And Ant Financial also stands out in terms of the breadth of its service offerings to its consumers. The range of products that it offers its consumers is one and the second is the speed at which it's able to respond to its customers because it is datadriven because it is AI first the company is very quickly able to understand what the consu consumer wants and it's quickly able to spin out products to meet that need. So

20:56 that is something that has fueled this exponential growth. So we move on to the next section lowering barriers for AI adoption. So the barriers to AI adoption are falling. And what I mean by that is um you know three of the things you need hardware, software, data and algorithms. And all of these are becoming easier um to manage. In terms of hardware, you know, you do need uh high processing. You need a lot of compute to train AI models. Um, and today if you want an Nvidia state-of-the-art Nvidia card, there is a waiting list for it. But, uh, there are also cloud computing platforms where you can buy compute. And these are scalable.

21:42 You just buy as you need and you scale up as you go. Uh, so the scal these some examples are Microsoft Azure, there's uh, Amazon AWS, there's Google cloud. These all of these companies do provide compute. So you can take as much compute as you need and you can scale up as you go. The other thing is software. So there's lot of open-source code available today. There are a lot of open-source frameworks available today. So there a lot of the coding has been done and you can build on top of it. Uh which is not to say that there uh there's no room for innovation. And AI is one of the areas that innovation is going at the fastest pace but there is still a lot of code uh that is out there

22:27 that you can build on top of. So you do not need to start from scratch in many cases. And in terms of data and algorithms, AI training is very dependent on the data sets. Um any AI solution is only as good as the data that it is trained on. But there are a lot of data sets that are available. Many of them are open source. Some are available for purchase. There are marketplaces where you can look for data that suits your needs. Um and also there are pre-trained models that you can use asis or you can train further. You can fine-tune. Um so in that sense barriers to AI adoption are falling. And why I wanted to make a mention of it here is that because it's becoming easier to

23:14 adopt AI, chances are that someone in your industry is going to start adopting AI. And people who do make that step are going to have a competitive advantage simply because they're going to make be able to make better sense of their data and make better decisions for the business. which is not to say I mean there's plenty of room for uh intuition and experience but having this sort of datadriven um architecture will will feed into that will add to that um and because these barriers are falling it's essentially a call to action to say you know it's time to get started so the next one AI and organizational structure if you were to

23:59 build AI talent in the organization how is uh how would you go about it. What are some of the ways uh that you could build AI talent in the organization? So there are many ways to do it and I'm just mentioning three of them just to get uh you know get you thinking in that direction. Uh one is a center of excellence model. Uh a functional model, there's a dispersed model and we'll get into the details of each. So center of excellence model is where you build a center of excellence. you pull in domain experts from your uh various departments and you also have data analytics uh data science and AI skills um and you build a center of excellence. So what this enables is that

24:45 your domain experts can drive what kind of solutions might be impactful for your company and then the AI skills can build solutions and they can work together closely. So some of the advantages is that um learnings can be shared because this is centralized. Learnings can be shared. It's easier to establish best practices at an organizational level. Um some of the disadvantages is that when you have a center of excellence, when you have the AI skills removed from functional areas, um it can lead to a a lack of understanding of what's happening on the ground. It's important to have AI development very close to what's happening on the ground because otherwise you may build a solution and

25:31 then you give it to a department and they say that's not how we work. This doesn't help. So it's very important to stay close to what the department needs, how they work, what the consumers need, what the clients need. So this is a functional model where um you you know you have functional departments, functional um uh categories in your organization could be human resources, it could be sales and marketing, could be manufacturing. So in each functional area you also give AI skills so that they can then develop solutions for those functions. Um the advantage is that you're very close to uh you're on the ground with your AI skills. Uh the disadvantage is that because it's decentralized um it can be

26:18 harder to establish best practices. It can be harder to um share the AI learnings um that are happening across the organization. A dispersed model is where you know there is no real organization of AI skills. you have AI talent in the company and you know whoever has an AI idea and wants to build it can work with AI the AI talent in the company to build a solution um and there there are many disadvantages to this I think uh the advantage is that it's easier for small solutions to kind of bloom up you know to build small solutions that could have immediate impact which model might be best for you um and that really depends on your current organizational structure. Um,

27:06 and whichever model you choose, you do need to have an ongoing evaluation to make sure that the AI skills and the organization of the AI skills is best suited to your company and the needs of your company. And there are um Twitter is an example where they had a centralized model of um deployment deploying AI skills but they they then they dispersed it a bit because they felt they were disconnected from what was happening on the ground. So um what are some of the organizational behaviors for successful AI implementation? Right. The thing is that AI successful AI implementation also requires a cultural change, right?

27:54 The your workforce needs to be uh prepared to one is to take in the insights that are coming in from AI. Uh and you need to build trust in the system. You could have state-of-the-art insights coming in from your data but if your workforce is not using it um then it won't lead to any benefit for the organization. So one is the focus on insights and process. Um you need to focus on how your business runs. Where can you give these insights? Where would it be most impactful to give these insights and you need to build and deploy based on that so that you are getting the insights from the data at

28:40 the points that you need it the most and in a way that your workforce can use it easily. If if it's difficult to get those insights if they if there are four steps to getting uh a recommendation from an AI engine, chances are it's not going to be uh useful. It's not going to be used at all. Um and the second is you need strategic intent. It needs to be from the top from the CXO level all the way down in the organization. There needs to be a strategic intent to use AI. Um it needs to be communicated across the organization. The benefits need to be communicated. Um and it needs to be a very intentional approach, right? Well communicated intentional

29:25 approach. So that strategic intent needs to distill from the top down. You need to build trust in the system. Um very often uh you know when there are established ways of uh doing things uh people are wary to sort of trust in a new way of doing things. So you need to build trust in the system and one way of doing it is to have an AI system running um and giving you recommendations and also have uh the traditional way of doing it and uh compare the two have metrics of success that people can refer to that people can uh look at and then say hey this really does work. Um so you need to build trust in the system. Also you need to have a clear understanding of the capabilities

30:12 as well as the limitations. What can these systems really do well and what can they not do? Um so that then um as an organization you know how to approach it, what it can be used for and what it shouldn't be used for. Strategy integration. So the AI strategy it needs to be integrated into the broader organizational strategy. Essentially AI is a tool that is helping you to reach existing objectives. So you need it needs to be integrated into your organizational strategy not as a standalone separate initiative but very much as a part of what you are trying to

30:58 do as an organization currently and essentially as an aid to help you get to where you are trying to go. So you need to also integrate AI in initiatives into the digital initiatives of the organization. It's very closely tied to how you collect your data, how you manage your data, where you store your data, etc. And you need to focus on AI's effects on your business models and your business goals and your business processes. very often some of the business processes do need to be uh do need to change a bit so that you're able to leverage AI in a meaningful way and finally we come to the AI adoption road map u the re the need for

31:44 the road map is that if you don't have a road map you don't have a clear idea about and an iterative way to build AI solutions um you know you might build a targeted solution today um but then tomorrow you think of another solution and these two cannot work together and they they don't build off of each other they don't complement each other so it's good to have a well-considered well thought out AI adaption roadmap how do we go about it so this is first why and then how which is to say that it is all about your business goals it is all about the current pain points that you have in the organization So first look at why would you want to look at an AI solution? How do you think

32:31 it can help you? Right? And then look at what might be a valid AI solution for this. What kind of models? What kind of deployment and into those technical details. But the first approach point really is your business goals and your business pain points. So understanding why you would want to um uh approach this as an AI problem and then look at how you're going to solve it. There are various dimensions to readiness at an organizational level. First is of course the data readiness. Then there is human resources readiness, business process readiness, uh your budget considerations. Um so we'll talk a little bit about each of these. The

33:17 most important is the data, right? For any AI solution, the most important ingredient is the data. And in terms of data, you need to look at how are you collecting this data? Where are you collecting this data? How are you processing this data? Very often, the data that's collected needs to be transformed in some way before it can be used for training an AI model. So, how are you going to process the data? And then storage. Where would you store this data? Right? Are you going to do a centralized repository? Um what are the security constraints around it? What are the security constraints for your industry um for your clients etc. cataloging that's a very important consideration when you get into big data when you get

34:02 into quantities of data uh that are required for AI models and the reason is um if you catalog your data then you're able to pull data out that is relevant to your use case otherwise you have a sea of data and you and you know nobody knows where to find the relevant data points so it's very important to catalog your data um so that you're able to pull it as you need and you know where to look for it. data security as I mentioned data visualization is a very important tool because when you're talking about big data um you can't visualize it very simply but there are many tools out there that can help you visualize your data and make sense of it and data visualization is very often a

34:47 first step to further processing you know you get a snapshot of your data and you say oh I see a I see something here I'd like to explore it further and that leads to uh the next steps. So in terms of data readiness, these are some of the things that you need to think about and be uh skilled for HR readiness. So your human resources, you do need to um you know get your employees ready for AI. Uh have have them have a clear understanding of AI. Um it would be great to have your resources use off-the-shelf tools for productivity because it's a great starting point. Uh you know you do learn some prompting techniques just by using

35:33 off-the-shelf tools. You get also a a basic understanding of how AI works by engaging with these tools. Um you could have short-term training programs um you know to bring your employees up to speed. uh also have an organizational structure plan. How are you going to structure AI talent in your company? Uh what is the long-term pipeline plan? Right? Very often when you start AI projects, you don't need to uh frontload it much. You don't need to hire heavy right at the start, but you could outsource your development. Have a small team to manage your AI solutions and then build as you go. Um but it's great to have uh and it's uh important to have a well-considered plan for HR right at

36:20 the beginning. Your business processes as I mentioned earlier you do need to re reimagine your business in order to be able to leverage AI um in you know to the maximum that it can give you. Um, and you also need to focus on AI consumption, which is to say you need to understand where would your users use this AI, how would they use it, and you need to deploy in a way that it can be used easily. The harder it is to be used, the less it will be used. So, it's it's important to understand how is how are these recommendations going to be consumed and then deploy in a way that it can be consumed very easily. And you do need to put efficient process in place uh to collect data to process data

37:08 to store it as well. So even in terms of for data collection the easier it is to collect data the better uh it works. You know if someone is going to have to u write down a few things and another person is going to have to key it in. Um the more complicated data collection is uh the harder it is to collect data. budget constraints. So what are some of the things um that will figure in your budgeting for AI? One is the data right u collection, ingestion, transforming, storage, all of this will uh incur some cost. So you do and data often is one of the main um constituents of an AI budget, right? Data takes up a

37:56 lot of budgeting. infrastructure. What is the kind of infrastructure you need? Are you going to use cloud resources or are you going to have resources onsite? Um you know cloud I feel that cloud is a great way to go because it allows you to get get started without a lot of capital investment and then you can scale as your AI initiatives scale. What are the AI models you're going to use? Are you going to build your own models? Are you going to use pre-trained models? So even if you use pre-trained models there may be a per usage charge for example for chat GPT there may be a monthly charge um some charge by tokens um so what are those uh expenses going to be research

38:42 and development so because AI as as a as a field it's not in its infancy but in adoption it's um still early stages so there's a lot of iteration that goes before you have an AI solution that really works well for you. So research and development is very much a part of having uh a good AI solution and you should budget for it. Um very often the first cut of your solution may not be uh what exactly what you're looking for but the ability to iterate uh and to learn from uh you know the various iterations is important and you do need to budget for that you need to consider that monitoring and maintenance. So AI solutions all AI models they are not um

39:28 you know build deploy and forget uh type solutions. You do need to monitor them. You do need to make sure that they are performing uh at the level that you expect them to perform. And what happens also is that the data changes over time. Your consumer behavior might change over time. Your numbers might change over time. So you do need to retrain your models and make sure that they are continuing to perform. So monitoring and maintenance is a very important part of AI solutions. Um and you need to budget for that as well. Human resources as I mentioned um AI talent that you may hire um or upskilling your current workforce um those are considerations as well.

40:15 So once you consider you know you look into your data you look into your processes the infrastructure the AI models what is it going to cost to monitor and maintain these models uh once you consider those things the idea is to have a road map where you have a mix of short-term and long-term projects it's important to have some quick fix uh solutions some quick wins so that you can build trust in AI early on and also have long-term projects that you would tackle. Um, you do need to have a clear path to data readiness, to HR readiness, to business process readiness and budget considerations. So this is what you would have in your implementation road

41:01 map just a an understanding of how you are going to approach AI what are you going to implement and in what order are you going to implement these projects and as I said it's very important to have a few quick win projects you know have a few um projects that you can deploy quickly so that you can uh build trust in the system um and the more people use AI solutions the more they see benefit from it, the more willing they will be to consider uh you know bigger solutions and solutions that require perhaps uh bigger changes in the way they work and I think that's it from me for the first workshop.

41:48 So any questions? If you want to stop sharing your screen, we can go to the chat. And sure, I know Rebecca, you have a question. Do you want to ask it? Sure. Um, very interesting presentation. Thank you. Um, I've been using AI a little bit in in my email application. Um, but my I had my first I have many questions. My first question is um what type of education paths are out there for people who want to pursue a career in AI. So the uh there are a lot of courses

42:35 available online today and I think that is partly uh a byproduct of the pandemic um that a lot of courses good quality courses went online right um so it depends on where in what uh manner you would like to work in AI if you'd like to build models um and train models um it's important to have uh a foundation of the the math and and the data and the statistics the data science underlying uh you know training AI models and then understanding what are the different types of models out there and how you can train it what is the kind of hardware you would need how would you go about it so that is one approach um the other way is if you wanted to use AI um

43:23 to impact your work directly uh you're a working professional and you want to use AI um so depending on your functional area For example, if you are a marketing professional, uh then you would look at what are the marketing tools out there and there are specific training programs online as well for these marketing tools. There are many uh AI tools um companies which then provide training as well to sort of onboard users. So that's another way of uh under learning AI. So really depending on how you would like to engage and work in AI, uh there is uh a lot of content available. It's pretty overwhelming. So were you doing

44:09 something in the data field before you got into AI and what was that? Okay. Um so I've moved around a bit. uh I was working in uh telecommunications earlier uh and not in the data field at all. Um but uh what I started doing after that was um investing in early stage businesses right and then in around 2017 2018 a lot of the companies that uh were approaching us almost every company would have a slide on AI and um we had no in-house understanding of what any of it meant uh and so uh we decided okay we need to look into it so that's how I

44:55 started uh studying about AI but when I started studying about it just was very interesting. Um and so then I went back to the basics um you know studied all the math um and the statistics and the data science and the coding and built up from there and then and then now we you know build solutions for clients. Um we also co-ound companies uh with others and we still continue to invest in AI businesses as well. And what's your company name again? Um so Beyond Z.AI AI is a company where we uh do consulting, we do training. Uh we also build AI solutions. Uh the foundation that Margie mentioned, get AI foundation is where we

45:42 try to bring um you know underprivileged into the AI workforce in to engage in the AI revolution in some way. Um we also have a company called Venture Factory where we co-ound AI ventures. So AI solutions are most um uh effective impactful when uh they're built alongside domain experts, right? People know how their businesses work and then and they know how AI can be used and what is the data that you need to train these AI models. So we try we co-ound um you know AI solutions with domain experts. That's venture factory. It's It is daunting.

46:31 Becca, do you have another question or can I turn it over to Pam? You can. No, go ahead. Thank you. I have a question about um developing and implementing an AI policy in your company. Um this is something that was suggested in our kickoff um presentation for this program and something that we're thinking about doing. Um, and also this week I just saw for the first time one of our clients sending an AI policy for uh work that we or anyone who's supporting their projects um you know guidelines for that. So I wanted to get your insight and just see you know what you're what what you might

47:17 recommend or any kind of insight you see in navigating that because that that's a relatively new thing. um that we're exploring. Yes. Um it's I think it's very important um to have an AI policy in place um especially if you are in uh any kind of uh content generation or you deal with any kind of private or sensitive data. Um and the reason being that it's becoming increasingly difficult um to uh differentiate between AI generated content uh versus otherwise and um not all AI generated content is factual. So I think it's important for a person that's engaging with content that you may put out as an organization to know

48:04 where that content has come from. Um so in terms of AI policy um one is to be very clear about um AI about referencing AI generated content right about mentioning uh when content is AI generated um in terms of an internal policy also I think it's it's it would be great for organizations to have a policy where they can tell their workforce what ways they can use AI and what ways they cannot right what are the areas where you are okay as an organization using AI and what are the areas where you are not um and there are there can be copyright issues um with AI generated content and I think that is a area of a lot of movement and it's not fully um sorted

48:50 yet u so that's important to understand as well when you're putting out content um you may not be able to copyright it if it's uh AI generated right and then um you know of course healthcare um industries like healthcare which you know which deal with very sensitive content. Um you need to be very careful about even how your employees may deal with um that content with an AI tool because any data that you use for AI training uh is can you know it can spit that out out of context. So you just need to be very careful about you know how the data is being used. So in terms of AI policy there's a lot to think about. Um one is usage. um as an organization what can

49:36 and cannot your employees do um in terms of data um you know what will you will will will you use and will you not use in terms of transparency to your external stakeholders how much do you need to let them know and how much do you not need to let them know so a lot to think about and I think it's a very important uh thing to do very important to do thank you are there templates like we could copy for an ai policy say like examples um I don't know of any so there is uh GDPR in the US for healthcare I think uh the EU has an AI act um EU is um the European Union is more stringent uh with the user data currently so they are

50:22 working a lot with uh to understand how they can put a policy in place in terms of regulation um some uh states I mean we spoke at Utah recently they have an office of AI policy which is trying to understand how organizations uh can engage with AI and what are the restrictions around it um so I don't know that there is any um you know offtheshelf policy that you might be able to take off but there a lot of people thinking very deeply about um about this area actually yeah it's interesting I'm going to bring this up to the other chambers as well because I think all chambers need to have a AI policy on their websites It's true. That would be very useful to have. Any other questions for

51:09 Mia? Well, this is part one of a threepart series. So, next Friday at 8:30, workshop two is business strategies for AI adoption. So, we'll take a deeper dive and learn how to transition from traditional analytics to big data analytics and discover how uh what tools can shape smarter business decisions and we're going to talk about skill requirements, critical considerations, two real world case studies showing how AI is driving strategic advantages in the marketplace and then our final workshop is going to be AI for marketing and sales growth. So, um, they all build on one another.

51:54 So, I invite you all for next Friday. And Mia, I want to thank you and your team so much. Um, just so you know, Beyond Z has joined the chamber. So, they are here as a member and, uh, we'll be able to, uh, use them and, uh, you know, just like we do other other members. So, they're a great tool for us. So, thank you very much. And if there are no other questions, I wish you all a good uh weekend. It's like I had to think of what day it was. So, and thank you, Mia. I appreciate you. I know you're you're traveling right now, so we appreciate the time. So, thank you so much. Thank you everybody. See you next

52:39 Friday. Bye.