Enterprise AI and Developer Productivity with Microsoft's Melissa Durbin

Kate Bevan: Hello, and welcome to this episode of the Infosys Knowledge Institute's podcast about enterprise AI, the AI Interrogator. I'm Kate Bevan of the Knowledge Institute. My guest today is Melissa Durbin, who is the America time zone lead for developer productivity at Microsoft.

Melissa, thank you very much for joining us today. It's great to have you.

Melissa Durbin: It's great to be here, Kate. Thank you for having me.

Kate Bevan: Oh, obviously we're delighted. Why don't you tell us a little bit about what you're doing now at Microsoft, and how you got there? Because it's quite the story, isn't it?

Melissa Durbin: Yeah. It's been quite the journey, for sure. I'm actually coming up on 10 years at Microsoft. I have always been in the software development lifecycle by trade. That was where I got my degree in, and have worked in various parts of application and software development. Starting with back in the day of creating synchronization applications with Blackberry and PalmPilot devices, to more recently I was leading a team in our manufacturing vertical here at Microsoft for the past three years, and focused purely on digital and app innovation.

What does that mean? The north star for digital and app innovation on the Microsoft Cloud is enabling and empowering customers to modernize their existing applications, either internal or external, to be cloud native and cloud-based. My team would focus on modernizing, and then infusing AI into those applications. Really, what gets us excited, really driving eager net new revenue for our customers, mitigating risk for our customers, or some operational efficiency.

With that, we have been spending a lot of time over the past year with AI. Within our manufacturing and operating unit, we had a lot of success with a technology called GitHub Copilot, the original Copilot that Microsoft has launched.

Kate Bevan: Do you want to just explain to the listeners who maybe aren't familiar with the GitHub Copilot project exactly what that does, or roughly what that does even?

Melissa Durbin: Microsoft has come out with all these different Copilots. If you think of what that word is, a copilot is somebody that sits along with you and helps you in your daily work. Yeah. GitHub Copilot is a Copilot that just does natural language search. What we're seeing is a huge increase, anywhere from 20 to 40 percent increase in developer satisfaction because of the tasks they spend their time on, in documenting the code, putting in simple syntax and phrases. Or even, rewriting applications.

When you think about that from a security perspective, and you're automatically scanning the code using AI to determine what the vulnerabilities might be, or issues with that code, and then being able to quickly remediate what those vulnerabilities are. It's helping to reduce the amount of time that it takes to find the vulnerabilities, and does a more thorough job of that analysis is number one. Then also, providing some auto-generation around what might help remediate those vulnerabilities.

I was very interested in that. That's how I ended up moving from our manufacturing operating unit, to now leading a team. We call them Global Black Belts. They're essentially the best technical experts that we have in this specific technology area. Now I'm leading that team across North and South America.

Kate Bevan: How did you get hold of that black belt? Because one thing we've discovered from our research is that getting hold of the right talent for AI in the enterprise is really hard right now.

Melissa Durbin: Yes. I'm actually in the process of hiring some people right now. Definitely, this is an area of investment for Microsoft, which is exciting. I think specifically in this technology, it's really like anything new. It's curiosity. It's having a passion for the technology to solve business problems. And bringing together folks that understand how to have top-down C-suite level conversations so they can understand really what's driving this business and this change.

Because one of the biggest challenges with AI is really change management. We're having to change the way people develop and deploy software.

Kate Bevan: Which sounds really quotidian, doesn't it? You think of all the challenges might be, of getting the coding superstars, the computer scientists. It's like, no, it's change management.

Melissa Durbin: It really is. I think it's definitely, on our team, it's a balance of those that are experts in GitHub, Azure DevOps, and the cloud technologies. Then also, being able to create a journey around platform engineering, and understanding where is the customer today, and how can we meet them where they are so that we can see two-week sprints, and develop POCs and MEPs, so that they can execute quickly to start seeing value out of the technology as quickly as possible.

Kate Bevan: I think that's a really interesting point, to actually getting value from the technology. Because at the Knowledge Institute, we have done quite a lot of research on where AI is being used in companies, how ready companies are. It seems to me that we're at this mess around and find out stage. And where the next stage to move to the actually finding out where do you get value from it.

Your approach, is that finding and generating value already, do you think?

Melissa Durbin: 100%. We recently did a use case with a chip manufacturer, where they were getting several points of percentage increase in how quickly they were able to develop and deploy the chips themselves. Netflix is another big one, where we were working with them, and enabled six to eight thousand developers to be able to release updates to their software faster. Which in turn gives more features to the customers, and that directly impacts the satisfaction of how they're serving their customers.

I think when organizations are starting to think about how to embrace AI, we have created a framework for specifically that whole use ideation. Really, doing design thinking, or a way to analyze different parts of the business, and create a funnel and a backlog of use cases that will go into different categories of value that can be created for the organization. If it doesn't create value, or an efficiency, or a mitigation or risk, then what's the point?

I think another exciting example that our team is working on is around the security aspect of it. With our technology around advanced security, we are looking at ways to help mitigate vulnerabilities that the line of code level, so as organizations are developing their software, and being able to understand what those vulnerabilities are, and fixing them at that level is going to be so much more cost-effective than deploying a piece of software, and risking some vulnerability being out there once it gets deployed. That's a big focus for us is security.

Kate Bevan: I'm going to play devil's advocate here. How do you know that you using Copilot, using generative AI to do that with code isn't actually creating more vulnerabilities?

Melissa Durbin: That's an interesting question. I think where we see organizations really thrive with something like an AI technology like GitHub Copilot is with mid to senior level developers, those mid to senior developers that have the experience to understand what to look for, insert. It's a copilot. We can't do AI without the knowledge of those workers being experts in what they do.

We talk about this all the time. Is AI going to replace jobs? No. What's it's going to do is make people more efficient in how they're doing their jobs. It's going to help automate things that are remedial tasks that need to be accomplished. But then, the AI models are only as good as we teach them. With those senior developers, they're able to leverage their own skills to be able to leverage the technology in the best way.

Kate Bevan: Would you say that part of the change management piece then is matching the AI capabilities to the level of the humans?

Melissa Durbin: I think doing an assessment of people, process, and technology is critical when you're looking at deploying AI technologies. So understanding the skill level and going back to the use case. What are the actual use cases that we want our organization to use this technology for, and what makes the most sense?

Kate Bevan: I want to come back to sustainability, which you mentioned very briefly. Because one of the things that really tickled me when we spoke earlier was you said you live with data centers literally in your backyard in your part of the US. We've seen some alarming news stories coming out of how Google for example was failing to hit its net-zero goals at the moment because of the emissions for the compute for AI.

How do we make AI sustainable? This surely must be an issue for all the big technology companies, such as Microsoft.

Melissa Durbin: I think what I said was, "Do you know what's unique about Virginia?" In Virginia, we have more data centers in the US than any other state. Most recently, in my 40-some years living here, I got a notification that I had to change my watering schedule on my lawn because of the water reserves here locally where I live. They're just building data centers all around us. The amount of water, and electrical, and power that it takes to cool these data centers, especially now with AI technology, is just massive.

Kate Bevan: Virginia is not exactly a desert with a water shortage, is it? Being told you have to rethink your own water consumption as a consumer is quite scary.

Melissa Durbin: Yeah. It was definitely eye-opening to me. I think there was an article recently that came out about the movement of data centers here in Virginia. It's definitely something that's top of mind for me. It's something that I live with and I'm seeing on a daily basis, just driving around where I live here.

I think it's important, Microsoft talked about our carbon footprint, and what we're doing to make AI sustainable in really different, innovative approaches. They talked a little bit about the analog optical computing, which researchers at Microsoft are developing an analog optical computer that has the potential to accelerate AI inference and optimization workloads by 100X. It's a new kind of computer that uses physics and physical systems for computation, making it significantly more efficient compared to traditional GPUs. That's one area.

Another area that Microsoft is researching is our brain-inspired design. It's inspired by the efficiency of the human brain, where our researchers are working on neural networks that simulate brain processing. These aid to enhance the accuracy and efficiency of predictive models to improve AI's proficiency in the natural language processing and pattern prediction.

Then, sustainability by design. With that, sustainability by design, they're advancing the sustainability of AI by integrating continuous innovation into the technology. That will include efforts to decarbonize and accelerate both sustainability journeys and business growth.

There's a lot of research and a lot of initiatives, not only in the investment of these new data centers, but also at the same time, into our sustainability practices. We have an entire sustainability team here at Microsoft that not only works with our customers on being sustainability, but holding ourselves accountable, and how we're going to be able to leverage that to be more sustainable in the future.

Kate Bevan: Well, I think this is one area where people who don't live and breath AI, like you and I do, really worry, and I think have a legitimate opposition to it. There are other oppositions to AI. Creators are worried about it, we worry about jobs. But I particularly worry about sustainability.

Do you think we'll reach a point where the value organizations and enterprises can get from AI is greater than the hit to sustainability? Therefore, it is objectively better to do it.

Melissa Durbin: Yeah. I think that's a great question. I don't know if we have all the answers to that yet, because we are so early in it. But I think in just the initial use cases of what we're seeing organizations leverage AI for, it's creating safer products. It's creating more efficiencies operationally. And it's certainly helping organizations mitigate risk. Then it comes to, well, what is the ROI of that for that organization? It really just depends on what those particular use cases are.

I think in healthcare, there's a lot of human error. In a lot of industries, there's a lot of human error. What can we do and how can we leverage AI to eliminate simple and easy mistakes to help us be more efficient, to create safer and healthier environments? I think really, we do have a great opportunity to do that. It just comes down to what is going to work best for an organization and how they're going to innovate.

But there's one thing for sure. That if you don't have an AI strategy, you're not thinking about creating ways to use this technology, you'll probably be left behind because most of our organizations that we're seeing start to come up with ways to adopt AI technologies, and enhancing that into their daily life, into their modern work, into their line of business applications, there is a significant ROI that is happening with our customers.

Kate Bevan: Do you think that's different to the blockchain hype and the Metaverse hype? Is AI different to those?

Melissa Durbin: I would say yes, because I'm literally living it every day. I remember when blockchain and Meta came out. I was just like, "I'm sorry, I don't know what this is. I can't explain this to my mom." I've been in technology my whole life. I still don't get it. Maybe I'm simple-minded. But no one's really doing that any more today.

Just in a year's time, the amount of adoption that we've seen, and I think you can see that based on the earnings calls that we're having, what we did just from a development perspective, what took us five years to do, we've done in one year's time. We're seeing the results at Microsoft ourselves. We're a case study for that. Then we're seeing those case studies and examples for our customers.

I do think this is different. I think this is a moment in time that, it's like the Industrial Revolution, and all these different points in time where technology was established. This is definitely a movement that is here to stay.

Kate Bevan: I tend to think it's a genuinely transformative technology, unlike some of the stuff that's gone before. Well, that brings me to my final question. The question I always ask is do you think AI is going to kill us?

Melissa Durbin: With any technology, there are opportunities for good and opportunities for evil. We're already seeing people use this technology in ways that are not good. Large AI is watching all data and video, and then people who want to control society or use that information to control citizens or shape things, that potential is very abusive and negative. Those outcomes can be very severe. I think that would be the worst possible outcome that could happen is too much governing and too much control. I think we need to permit freedom, permit privacy, and prevent that from forming in the first place.

The AI, and right now the language models, are only as good as how we train them. I do believe that we'll see what happens in the future with robotics and things of that nature. But I don't think we're there yet.

Kate Bevan: That's reassuring. Melissa Durbin from Microsoft, thank you very much, indeed. It's been a great pleasure having you.

Melissa Durbin: It's been fantastic to be here, Kate. Thank you.

Kate Bevan: The AI Interrogator is an Infosys Knowledge Institute production, in collaboration with Infosys Topaz. Be sure to follow us wherever you get your podcasts, or visit us on infosys.com/iki.

The podcast was produced by Yulia De Bari and Christine Calhoun. Dode Bigley is our audio engineer. I'm Kate Bevan of the Infosys Knowledge Institute. Keep learning, keep sharing.