How AI Is Democratizing Development for Everyone

00:00:01 Mark Smith
Welcome to the copilot show where I interview Microsoft staff innovating with AI. I hope you will find this podcast educational and inspire you to do more with this great technology. Now let's get on with the show.

00:00:17 Mark Smith
Welcome to the Microsoft Innovation Podcast. Today's guest is from Redmond, WA and she's helping developers build the future with AI's full links for the show. Notes are available. Please welcome Yena vice president at Microsoft. And one of the driving forces behind Microsoft Graph and Azure AI Foundry.

00:00:34 Yina Arenas
Hello, mark. Thank you for having me today in the.

00:00:37 Mark Smith
So Great to have you on. You've had a long history in Microsoft, and before we get into our discussion, particularly around AI foundry in the Community, can you tell me a bit about food, family, and fun? What do they mean to you?

00:00:49 Yina Arenas
Oh, food, family and fun. Oh my gosh. First for me, everything is around family. I am a mother of four, four very active boys. And I enjoy spending time with them and I am extremely proud of who they are and the fun side. I love sailing and spend time sailing with my boys, of course. And my husband and I love traveling as well and in the food side. Oh my gosh, I enjoy. All types of foods and like really like. Learning about new types of cuisine and like getting to like, just share meals with friends and families.

00:01:27 Mark Smith
Nice. Nice. You would love New Zealand. There's a lot of sailing opportunities in New Zealand so.

00:01:33 Yina Arenas
One that I've not been yet to, so definitely on bucket list.

00:01:38 Mark Smith
Nice. Tell me about your journey to your role now in a foundry, because you've obviously had a lot of history in Microsoft Graph, and I think the unsung hero of M365 copilot. Is the graph because it makes everything possible it. Unbelievable level. So tell us about that journey, your experience there and then what's your major focus in FY20 6?

00:02:09 Yina Arenas
Sounds good. Well, I actually just completed 15 years at Microsoft. Wow. And it's been an incredible journey. As part of that, I've had the opportunity to work with amazing talent across the company, amazing teams, great products. One of the incredible. Force multipliers. That happens when you work in a company like this is that like the impact that you get to have across like so many different industries and so many different locales and markets, it's just tremendous. So I'm very proud of that fourteen of those years, I actually spent on. Our productivity suite, the teams that like build product I started with. Point and I worked a couple of years in the SharePoint team was part of the team that was transitioning SharePoint from. Box product that was all sold on a CD every three years to like a cloud service, specifically doing all of the developer platform, moving from full trust solution to like, you know, running that as a multi tenant isolated solutions in the cloud and that like started my journey with. Building for developers as customers and as part of that, during those 14 years that I spent on what Microsoft is the. Enterprise and devices organize. I worked on programmability and extensibility across like just driving that ecosystem and that leads to the birth of Microsoft Graph. It's like how do we make sure that we have a consolidated data model across all of the different products that make Microsoft 365. And yes, of course, Mark, you're right. Like it's a huge differentiator from. Just a plain BLM or like chat based application running outside of the context of like what customers data have and like being able to ground on all of that data, whether it's mail, calendar or teams, all of the SharePoint documents, all of your OneDrive files, all of that content that. Makes you and like your group and your organization so unique, right? So and I had an amazing time driving that as I was wrapping up my time in the Microsoft 365 organization. I worked on driving programmability for our internal platform, our external platform, and copilot. Extensibility as well. And then the opportunity that my door to move over to drive product for Azure AI Foundry and I've been actually as part of like completed 15 years of Microsoft also completed my first year in. Team I joined as a product leader for foundry and as that means I drive lead the team that like drives all of the model offerings. So like how do we enable customers to create, customize and consume models at scale? All of our Agentic platform offers. So like how do we make sure that we have the right set of capabilities and functionalities so customers can.

00:05:10 Yina Arenas
Build single agent multi agents and like the connection with models but like the full life cycle of those applications including connection to knowledge, connection to tools, but also all of the observability. So like, how do you make sure that those agents are behaving as you're expecting, right, both from a quality and a from a security and risk perspective. So that's part of the portfolio that I drive here within the foundry team. It's been a fantastic year. Like there's so many things happening in the AI ecosystem. Every single day there is something new like we are constantly bringing new value to customers, releasing new model. This, for example, is a week of a lot of releases. We're gonna have some that like, I can't like today we released some of the BFL models. Some other models are coming down later in the week. So it is always such a dynamic environment and like there's so many things happening across the industry. It's just fantastic to see. I mean, like, if you like, the pace of innovation right now is just mind blowing. If you think about. That in just a short period of time, AI has evolved from being just a simple rule based system, data-driven machine learning to know having this agentic set of architectures like it's evolved from just being able to have a chat based interface to being able to create agents that. And plan that can reason that can act that can adapt. So it's just mind blowing. It's just like and as you can see and like my excitement about the things that I get to work on and just like jazz about the things that. What to do?

00:06:53 Mark Smith
I love that when you're talking about the speed of innovation, cause I felt for the last two years, Microsoft has been throwing AI into everything and definitely not in a haphazard way. But I feel like it's now really consolidating into some really exciting like you can see, it's like settling into now the groove of really high value delivery. One of the things that often. That ask is what is AI foundry? And the analogy which I it might have even been you that said this so about a year ago, which is imagine a workshop, you got your tool bench, you've got your building materials, you've got and you can go in there and you've got your LM's, you've got your API's, you got everything that you can pull together and create something from. Is that still the analogy you're using or do you have another one?

00:07:36 Yina Arenas
OK. Yeah. Actually, we think about it. As helping customers like from a we use a lot of the factory analogy, right? Like, imagine you have an assembly line where you're going from. Here's a product idea to like, here's the things that we will need to building the set of ingredients or tools that we will need to build it to all the way to take that into production. So Foundry enables that for developers that want to. Infuse AI into their applications. Whether it is that they're building a brand new app or an agentic system, or an automation, or they're like infusing AI into their existing apps, right? Like they can come to foundry and they. Find the models, the agents, the observability, the end to end set of like tools that they need fully integrated with the IDE. So like we have like all of the Visual Studio code experiences, the APIs, the SDK. So like all of the things that you're gonna need to be successful like infusing AI into your applications.

00:08:30 Mark Smith
Another thing I've come across is which AI to choose. You know which one and this is coming. Recently I was involved with the executive meetings around education, university level education and as somebody in the ATS part of the Microsoft organization. Like how do I differentiate why Microsoft is the better tool. And for me, I was like. It's the graph because a WS doesn't have that deep look into an organization's day-to-day activity. You know people live in Outlook, people live in teams, people live in, you know, their document storage environments where a WS doesn't have that front end into everybody's day-to-day live that Microsoft has. Making copilot, particularly a massively compelling story, is there anything else when you talk about differentiation of what Microsoft offering is compared to the other hyperscalers in market that you kind of bring to bear?

00:09:24 Yina Arenas
Absolutely, absolutely, I think. I would say that there is a lot of great product out there when it comes to empowering lots of different set of personas to like leverage AI in their day-to-day workflows. Whether it is IT admins and users, developers and like a lot of different functions. One of the key differentiators. That Microsoft is like that we have that end to end suite of product. Products and that integrate together right? Like if you start with like just all of us as end users as you're saying that spend hours a day on Outlook and teams and like writing documents and collaborating and like just having conversations and all of that information that is flowing in Microsoft 365 is leveraged by copilot. Provide these experiences right? Then like also if you are wanting to like, go beyond what copilot offers from out-of-the-box, you have tools like Copilot Studio to fully be able to create your own agents to create your multi agent systems like without any knowledge about code or any knowledge about how the underlying. Models work. You can be. You can create very complex solutions. With no code, no code. Using a tool like copilot studio, and then if you look at like you know that's on the, I will say employee productivity, the productivity suite. But like then if you're building solutions, right, like Microsoft is a platform company and like we have not only the opportunity to integrate with all of the Microsoft 65 ecosystem, but also like. All of the set of sort of products that we have like think about data for a second, right? AI is great, the models are evolving every single day there is something new that is coming out when it comes to capabilities, functionality, that how fast can they process how cheap it is to process. But like all of that is only going to be as good as the inputs that you give it, right? And like across Microsoft Data State, we have both the.  Unstructured data. The structured data, the public data, all of that can come together in a product like Fabric, right? And they have all of the AI capabilities also embedded into that. So across the stack like we have. All of these different product offers offers for different personas and they are all now like as you were saying, mark being supercharged with AI and like as you see all Microsoft doing that like the platform that they're doing like every single product of Microsoft. That is like infusing AI into their offering is using foundry underneath the covers. And that's the same platform offering that we offer to our. Customers, as they want to develop their own solutions, right, and we see customers doing that across business to consumer scenarios, business to business scenarios, internal productivity scenarios, they typically start more from like multi conversational type of agents building some of these like conversational experience. But then they go more into like business process automation, full autonomous workflows powered by a multi agent orchestration and like you know, it's just you know across the stack whether it is internally at Microsoft, we use foundry to power our products. And then we also offer that to our customers as.

00:12:41 Mark Smith
So tell me about the customer adoption trends and probably from two different lenses, enterprise customers that are really, you know consuming and using foundry at scale, but then also the developer community, their adoption just from a lens perspective, what are developers building that surprise you? What are you seeing from an adoption perspective that kind of delights you, I suppose?

00:13:04 Yina Arenas
Yeah, Yeah. I mean like it's one of the things that I always have loved about working on platform is that I get to see a lot of the different scenarios that our customer. Built and across so many different industries and across so many different types of applications, so right now we have over 75,000 organizations that are using Azure AI foundry and a lot of them are enterprises that are building. As I mentioned, some of these interactive experiences. Others that are building full automation of business processes, others that are like, you know, building a more consumer facing solution. And I think across all of them, there's like examples that are always like, Oh my gosh, is like, the way the things that they're doing is just mind blowing, like, automation perspective. Like for example, the other day we were having a conversation with B&W and B&W like, is looking at like, how do they do manual data transfer from on Prem processing of like? All of the hundreds and thousands of signals that they get from, like their sensors that are in the car, and then they've developed a solution powered by Azure open AI and then to make sure that they can. Like. Greatly expedite the processing time of data. The generation of the insights like the amount of vehicles that they can capture, how much that they can like, you know, process all of that information and really generate insights across their organization. So that's one example. Another example is insurance companies like we have like a customer Providence.

00:14:37 Yina Arenas
Who's like is handling? This is more on the healthcare actually handling an overwhelming volume of patient messages like think about disruptive care, right, like how much time like? You know whether it is like the actually medical professional or all the staff that are supporting that those functions, the amount of time that they just spend on data processing like right. And if that's can completely be taken care of by an ambient AI where then it improves the time that actually the healthcare providers can focus on. Like care for the patient, right? And so that's another just great example of like not just automating but like you know humanizing some of like the applications of a. Why we have customers like just trying to bring examples easy for example is making sure that like all of the tools that they're building internally within the organization are inclusive for neural divergent employees. So like they're code developed inclusive features using AI and they've enhanced accessibility and productivity. Of their solutions, we have customers like Heineken. They developed an internal solution. It's called hoppy and like it's all about how. Do they supercharged productivity of like all of their employees, and as such, like we have lots and lots of customer examples of like, as I mentioned, organizations that are thinking about how do they give their employees more capabilities with AI their they give their customers, there's a lot of like examples on customer service. Right. Like, how do they, like, provide a better customer service? How do they generate new value using AI? It's just like really, really great to see. But I will say the set of examples that like. Get closer to my heart is like when the we get that amplification of like the human capabilities like for example the health one, right. Like you can say like you know now doctors can focus more of their attention and their care on like the patients as opposed to like filling out like a report and doing all of these different things. And like we have examples like that. You know a ton.

00:16:41 Mark Smith
We've seen a lot for developers about is it worth to, you know, to do a STEM degree and get into development? Is development going away and then we're seeing developers that augment what they do with AI, supercharging them, right, allowing them to do so much. More and I don't know if you would see this data, but are you seeing more people come to foundry that weren't what's called the traditional developer that are more in that vibe coding space? They wanna robust, they wanna trust with their platform that Microsoft provides that are now starting to use the tooling a lot more because it's easy to access it. From a I can create something without having you know a data scientist and 50 engineers and things like that. I can actually take my concept into reality quicker.

00:17:28 Yina Arenas
Oh, absolutely. Absolutely. I think one of the things that AI is definitely doing is democratizing access to the technology itself. Right. And I'll give you an example of like how we use it internally on my team like, you know, our product making craft is evolving significantly like in the past, you will have like. A PM and my team maybe like spend time like learning about the market, learning about customers and then like writing a document, a spec that like says like this is our the all of the user journeys and scenarios that we need to support like maybe designing helping design the API. Guys now like the jump is not so much into the document but in the showing. It's like OK, so here's the prototype that we've done of like how the UI would look like and like very much detailed set of like flows. And here's the mockups of the APIs like when you just get start doing them in a way that you can produce that with AI. That like just, you know, health gets us faster to the outcome and the clarity across the team, right? Like so that is definitely I think that you know it's helping a lot a wider audience to get faster to like some of these conceptual ideas and like being able to see them on a more tangible way. Now I also wanna say like it's not like you know taking a solution from idea to code is one thing, but then like take code from to production is another thing and like I still think that there's a lot of like you're either using a software as a service like Studio that helps you like gets you on rails for some. That or like as a developer, you're still wiring up a lot of all of the different services, whereas your app logic gonna run. Where is your data going to be stored like some of these things like you still need to like architect, right? But then like you're spending that time more on that solution, right? Like on architecting the solution and thinking about what are the things that you need?

00:19:13 Mark Smith
Yes.

00:19:22 Yina Arenas
Based on their scenario and not so much on, you know, the more I would say not as fun activities, right?

00:19:28 Mark Smith
Yeah, yeah.

00:19:30 Yina Arenas
You know, validating unit testing and all of these different things, it's like you can use AI to generate a lot of those today. And then like you focus a lot of your time on like what are the new capabilities, what are the some of the things that you want to bring into your solution?

00:19:43 Mark Smith
One of the things I've come across is that there are a lot of enterprise customers. They've got stuck in prototype land or POC's, right? They've got the POC, but there seems to be a disconnect between taking a POC from POC into production and productionized like you're. Saying. Are you saying that? Why do you think it's there? What is that struggle? And do they choose the wrong thing to prototype? In other words, where they're trying to create a shiny new toy rather than something that solved a real business need or something, it was going to add so much value that they wanted to get into production.

00:20:17 Mark Smith
Quickly. What are you seeing?

00:20:18 Yina Arenas
I think that wave, one of AI was definitely a lot of experimentation and like seeing a lot of like trying to understand like, what is it useful for? How do we integrate with it? Like what are some of the use cases that like can be a? Thinkable and companies have been like experimenting like there's a lot of now that is going into production, right? Like there is now going into like, OK, here's how we think about the amount of investment that is needed to take a solution, to build a solution in the space. Then, like, here's how we think about measuring ROI, right, like. And you kind of have to go through those. Needs to like understand this. Like. OK. Are we thinking about it the same way or like how do we need to completely rethink about how do we solve for this problem? And I think that that's a learning curve and a lot of our customers starting on that front where like they're.

00:21:08 Yina Arenas
Like is it just replacing? What part should we rethink the entire system? How do we think about approaching the technology itself and a lot more has gone into like how do we scale, how do we make sure that we can go to production, how do we make sure that we can measure quality, how to make sure that we can have continuous evaluation of the system in production and making sure that it is meeting. Of the expectations. So it's definitely worth that wave 1 and then I mean I will say like it depends because there's like different types of customers. You always have the early adopters, then you have the comments and like all of that. So I will say that like at least for the ones that are in the forefront, that wave of like prototyping like has passed and like we are seeing more and more. Like large scale, high volume solutions. Across our customers, another trend that we've seen interestingly shift is the one around fine tuning. So two years ago it was like, no, you shouldn't think about fine tuning because the models are ordering so fast that like if you fine tune a model and then the model already has like the capabilities.

00:22:02 Mark Smith
Yes.

00:22:13 Yina Arenas
That actually has pivot that there's a lot more customers using fine tuning for their specific datasets, right? Like for their specific domains. And like it's complements, not only the knowledge that the model has already embedded, but also like you know can help a lot on reducing costs. And fine tuning is another approach that has, like you know, picked up. Significantly over the last I will say 6 to 9 months. Awesome.

00:22:42 Mark Smith
My last question is getting you to look at a crystal ball and think of the future around human skills that need to be leaned into. As we continue down this AI journey. If you're advising one of your boys around the skills that you think that they need that are going to give them longevity in their careers, what are you thinking? About.

00:23:06 Yina Arenas
Look, I think that, and I've said it before in in other forums like we are an inflection point in human history and this is definitely a transformation. Technology and it's happening faster than we've ever seen before. Any other technology that has been like such transformational for. Humanity in terms of skill set, I think it's really important for like you know, regardless of your profession like it, regardless of like where you focus on regardless of how technical you are that like you really spend time being curious and learning about what AI is learning about like you know. What it brings like what is a? Chat completion model. What is the reasoning model? What is like the different moralities? You don't have to like go super deep, but like understanding like you know what is it doing underneath the covers to like generate some of the answers like how does it get the knowledge? The fact that like you know there is all of this world knowledge there is these models that transform and things. Like that? Just kind of like. Being a general idea of what's happening and then like give you an understanding of like, OK, this is a tool and like, how are you going to use that tool?

00:24:17 Yina Arenas
Like bring benefit to you like regardless of like the domain that you're working on. And I think that's a baseline. Every single one of us right to just have that understanding. And then like constantly being I think curiosity would be the theme of my guide. My advice is like be very curious about. What's going on and like how you might like? There's no. Set of capabilities, new set of applications. How are you going to use this technology to transform the amount of output that you can generate and that can look into many different things like for example how can you make be more efficient in things that you don't necessarily want? To do.

00:24:59 Mark Smith
Yeah.

00:25:00 Yina Arenas
Or so that you can spend more time on the things that you do want to do. Right. So like, I think that is one that I think is key for. As we look into like next set of like months and years coming in, being curious about what's constantly coming.

00:25:15 Mark Smith
You know, it's been so good to have you on the show. Your enthusiasm is infectious. I love that final bit on curiosity, so powerful and everybody can develop that skill in themselves. Thank you so much for coming on the show.

00:25:29 Yina Arenas
Mark, thank you so much for inviting me and for all of you developers out there. Ai.azure.com, please come and visit us like build with us, build agents, build applications like you'll definitely have all of the tools that you need to be successful in that journey.

00:25:47 Mark Smith
Hey, thanks for listening. I'm your host, Mark Smith, otherwise known as the nz365uy. Is there a guest you would like to see on the show from Microsoft? Please message me on LinkedIn and I'll see what I can. Do. Final question for you, how will you create with copilot today? Kaki day