Harnessing Codeless AI: Gwenael Bego's Journey from Hong Kong to Paris and Revolutionizing Workflows with Microsoft AI Builder and Power Automate

Mark Smith: 0:33
Welcome to the Power 365 show. We're an interview staff at Microsoft across the Power Platform and Dynamics 365 Technology Stack. I hope you'll find this podcast educational and inspire you to do more with this great technology. Now let's get on with the show. Today's guest is from France. He works at Microsoft as a Group Product Manager for AI Builder and Power Automate. He is working on democratizing access to AI technology with no code. You can find links to his bio and socials in the show notes for this episode. As always, Welcome to the show, Gwenael.

Gwenael Bego: 1:08
Thank you, hi. Mark, thank you for having me. Did I pronounce your name right? Yes, Gwenael, or Gwen for short, is fine as well.

Mark Smith: 1:16
Awesome, well, good to have you on the show. What part of France are you from? I am from Paris, paris, france. Awesome, you've lived there all your life.

Gwenael Bego: 1:23
No, I have a pretty multicultural background, so I was born in Hong Kong, actually, and then I grew up in the United States, in New York. I've been in France for quite a while now, about 20 years. So back in France, but I am French originally.

Mark Smith: 1:39
Would you call yourself a Parisian?

Gwenael Bego: 1:40
Yes, multicultural Parisian citizen of the world.

Mark Smith: 1:45
Nice. Now when you said Hong Kong that interests me because I've done a lot of work in Hong Kong in my career what part of Hong Kong we are in and why were you born in Hong Kong?

Gwenael Bego: 1:55
I don't actually know. Remember I left when I was three, so my family went abroad, lived in Asia for many years and so, yeah, both myself and my sister are born over in Asia. Yeah, traveled a bit Hong Kong, singapore, australia but yeah, I was very small so I remember not much of it.

Mark Smith: 2:14
Yes. Have you ever been back to Hong Kong?

Gwenael Bego: 2:16
I did, I did, I did. I went back a few years ago, interesting to kind of go back on the motherland or your birth land.

Mark Smith: 2:22
So yeah, that was quite something. Yeah, I love Hong Kong. I love it. I love spending time there. It's such a great city and I'm talking about really the island, being out on the island rather than the mainland part of Hong Kong. That's where I've spent most of my time, but, yeah, definitely love it. Tell me about food, family and fun. When you are not working, when you're not focused on your Xbox core or something like that, what are you up to?

Gwenael Bego: 2:47
So I have a family which keeps me quite busy with the three young children and fun. I love music. So yeah, still play a bit of guitar in bands kind of here and there, but yeah, I think progressively over the years, work and family have supplanted any of my efforts in terms of music. But yeah, this is what I enjoy for fun, and then sports as well. Enjoy a lot of basketball and running as well.

Mark Smith: 3:14
OK.

Gwenael Bego: 3:15
OK.

Mark Smith: 3:16
Tell me top five things to do in Paris that are not the typical. You know, Louvre, Eiffel Tower, Moulin Rouge, Meltmar or Notre Dame, or Cruise on the Sin. Don't give me any of those. I've done all those. Tell me the five hidden secrets of Paris.

Gwenael Bego: 3:36
I would say the Microsoft office is maybe a hidden secret, but I don't know if that would be. No, that's good, so much enjoyable. Yeah, I mean, you have to go up to kind of the you know where the kind of the real Parisians go to hang out, maybe a little bit up north, in the north of Paris, areas around maybe Republic Oberkamp, some places maybe in the 18th. So yeah, there's a very vibrant nightlife in Paris. So, yeah, there's definitely lots of things to do there. And yeah, there's, I mean, the Louvre, obviously famous museum, but there's lots of super interesting museums. There's, for example, fondation Louis Vuitton. I was there very recently to catch Super interesting expo around works of Andy Warhol and basket. There's a very, I mean super vibrant in terms of of culture. So I would say, yeah, those are definitely some things to do. You definitely have lots of culture around theater as well, the arts. So you know, maybe catch a play at the Comédie Francaise, some classic French theater, and yeah, there's definitely some interesting areas to kind of go. There's lots of rooftop places, rooftop terraces that are starting up in Paris, so nice yeah, that'd be interesting to catch.

Mark Smith: 4:53
Excellent. I'll be listening to this back to note all those, and On the next time I'm in France, there's so much more of France I want to explore. I've been a lot in the south of France, yeah, you know, stayed there for some time, and you know, I haven't been to champagne, which I feel is is shocking, being that my wife and I both love our champions, and there's a couple of UNESCO World Heritage sites, I think up around that region as well that are a must see at some point. And then, of course, the cheese is. Oh, of course, yeah, so I'm so cheese, you know cheese is.

Gwenael Bego: 5:26
Champagne is amazing. When I went there, I felt like every occasion they just drink champagne it's instead of having a bottle of water, hey, champagne. Yeah, yeah it's wonderful there.

Mark Smith: 5:36
I was in a wine bar in London and we'd been drinking wine you know, all evening bottles. I was with a friend that was very much into his French wines. We wanted to switch gears and so we needed a palate cleanser. And we asked their host what do you recommend as a palate cleanser? And he was like oh, champagne, no, I know. And my friend was like, oh, no, no, no, let's extreme. You know, I'm like, no, let's do it. So he had an expensive bottle of champagne and then switch gears onto another variety of wines. So it was superb. Gaming. Are you a gamer?

Gwenael Bego: 6:15
Ah, I used to be, but I'm not so much of a gamer. I used to work in the Xbox division actually. Yeah, for many years we were working on a Media service for Xbox, and so we had lots of great perks there of, you know, getting access to the Xbox and a ton of games. So that's maybe the period where I was mostly a gamer, but I've lessened my gaming activity quite drastically.

Mark Smith: 6:40
I see Sachi has just did a big promotion of Starfield. Yeah, been released and I've spent probably five minutes before being called away by children to in it and I'm looking for a Window, you know, to get some good game time in, because that just looks like an epic new game that's out now. I'm not on Xbox. I've never really taken of owned Xbox consoles of one them in Microsoft competitions and sold them pretty quickly, because I'm more on a PC is my my preference when it comes to gaming.

Gwenael Bego: 7:10
So the mouse and keyboard.

Mark Smith: 7:14
Yeah, yeah, rather than a controller. Okay, let's get on to the meat and potatoes of the show, which is around your area of expertise. So just tell us a bit about Before we get there. How did you get to where you are at Microsoft? What was your it tech journey outside of Microsoft, then inside of Microsoft, and and bring us right up to date about the area that you focus on inside the tech stack.

Gwenael Bego: 7:38
Yeah, so I mean I studied Computer science, I think, as probably a lot of folks at Microsoft. I started my career in startups Paris another thing about Paris super vibrant tech, startup ecosystem, awesome. So I spent about kind of five years in that ecosystem and we were working on a software distribution, actually for game companies and software publishers so not very far away from the realm of Microsoft. And then, yeah, I kind of wanted to see a bigger structure, for better or for worse. I think one of the things that I loved about startup was everything is going super fast but at the same time it's kind of crazy with no processes at all, and I just wanted to see kind of the other side of the house to be able to grow kind of personally. So, on a different route, I went to Accenture for a little while and then Microsoft so kind of saw lots of different flavors of the IT, software and services industry and I started as a PM, as a product manager, individual contributor in the Xbox division, yeah, working on media. So, yeah, historically there's a team of research and development for IT development in Paris, officers for Microsoft, the Microsoft Engineering Center, and we focused over the years on different types of products and A couple years back I think there was a pretty strong sense to put a ton of Resources within Microsoft to focus on this kind of vibrant new Ecosystem around power, platform and dynamics and creating, you know, innovative business solutions. And you know our team was brought into the mix and welcome that change. It felt like very core to Microsoft's kind of core business, which maybe Media music and video is a little bit further away from that core. So I'd welcome that change and have been there for about, I think, probably four or five years now, and so I've done, I think, different things aside from, I see, a product manager over the years worked into being a product lead. I went to engineering division for a little while as well, managing engineering managers, and I went back to PM because I feel like that's where I can bring the most value. Defining products and talking to customers and Solving problems for customers is what I get a kick out of the most.

Mark Smith: 9:56
That is epic, and you know what an observation I've made about Microsoft and I've been in the Zaps since 2003. So 20 years now. Is I noticed in more recent times as laser focus on customer within the team and I think it's creating great results. You know being able to pivot and accommodate and set direction based on customer need. Tell me about the specific area that you're working on at the moment, as in so what are you responsible for? What's your tech focus within Microsoft?

Gwenael Bego: 10:30
Sure. So within the Power Platform, as you probably know, we have different types of products, and then we also have shared capabilities that are distributed across different product surfaces, like Dataverse being one of them, the kind of foundational data layer for a lot of the different apps connectors as well. And then AI Builder is another one, and so I am responsible for AI Builder. Ai Builder is conceptually kind of the AI platform that other products can use within the Power Platform, and it gives the citizen developers access to a ton of AI capabilities that are mostly powered by Microsoft's Azure's own AI models that have been developed by Microsoft Research, and we give makers access to those models in a low code, no code form factor, so it's very easy for them to leverage these capabilities, to insert them, for example, in the applications that they're building, in the workflows that they're building, essentially to make those solutions more powerful than the standard authoring. Of course, now you know, as things are evolving, ai is kind of infused everywhere and integral part of all products, but we still have that kind of AI customization, ai building, type of focus in our product layer.

Mark Smith: 11:44
So you obviously support. When we look at the five apps that wrap the Power Platform, the five technologies, we've got Power Apps. We've got Power Automate, power BI, power Virtual Agents and Power Pages. Which ones does AI Builder specifically support?

Gwenael Bego: 12:00
So right now we're specifically integrated in Power Automate and Power Apps and we have a lot of focus on Power Automate a short term I think that's where a lot of users get the most tangible value out of using AI is also in the context of automation, and so we have lots of different AI capabilities that people infuse as part of their workflows to be able to automate really in an end to end workflow, an end to end working process, to be able to automate tasks that they didn't think were automatable before, for example, you know, reading a document and understanding the context of the document, classifying the document, extracting data from a document things that are now feasible with AI, that before people had to do as part of a manual task. So those two are the primary products in which we're in and, I think, lots of focus on Power Automate specifically.

Mark Smith: 12:55
Yeah, tell me about the models that are there. Just refresh me on the ones. I'm familiar with, a few of them, about four of them. But what are the models that are there straight away, and, if you like, to use cases that are supported by AI Builder?

Gwenael Bego: 13:11
Right. So there's document processing. I think is a very popular use case, so that's using Azure Document Intelligence in the background, but there's several pre-built model capabilities that allow you to automatically extract data from things like an invoice, a receipt, an identity document, for example. Then, if you want to process a document type that is a very specific format or a different, unique, maybe company format, you can build a custom model. So, essentially, there's a training interface that allows you to submit your dataset with examples of the types of documents that you want to extract and you label the different elements on the page and once you do that for a couple of documents, you end up having an AI model that's trained and that you can use to automatically extract content from those custom documents. So documents is one area. There's another area around text processing, and so the typical scenario here is going to be, for example, email processing. You have an email comes in and you have a very unstructured document format, so not well-defined grid or tables where items are all sorted with a label, a totally free-flowing, unstructured text, and you still can train an AI model to understand that, extract data from that, classify that and so typically in a workflow, when an email comes in. Instead of having somebody spend a lot of time analyzing the email, figuring out what it's about, who it's destined for, you could do that automatically parse the email, extract key points, forward it to the right person and then, even now with GPT, go even further, maybe draft even a response that somebody can have a look at and iterate on and use as kind of a building block to streamline customer interactions. So text processing and now evolving rapidly towards GPT, which is a new capability that's been available, and then still more areas one around object detection, so essentially image processing is kind of the key target scenario there. Lots of manufacturing industry scenarios around analyzing the content in an image, and then structured data scenarios to predict outcomes based on historical data. So those are kind of the four broad categories I think.

Mark Smith: 15:25
Do you have sentiment analysis?

Gwenael Bego: 15:26
Yes, yes, so that is one of the text processing capabilities. Yes, so yes.

Mark Smith: 15:31
So you mentioned document part and recently looked closely at Microsoft Syntex, which is Microsoft's tool around AI in the context of documents, in the context with you. Combine it with SharePoint and we're talking about, for example, legal firms, insurance companies, etc. A large document type led organizations, maybe in the medical industry. Is there any interconnection between the Syntex products and AI builder and what you can do there? Yeah, absolutely, is it a non-ramp to that product, is it? You know what are your thoughts.

Gwenael Bego: 16:11
Yeah, no. So we have a partnership with the Syntex team and so they leverage a lot of the same kind of out of the box AI capabilities that we give Power Platform customers access to, except they give those you know, the users that are in M365 and SharePoint access to those capabilities. Yes, when it comes to building out custom models, to custom training a model, they actually embed an AI builder model creation experience. So we've partnered with them essentially to embed our AI custom model interface and so users can go a step further if they have a unique document that they want to process, they can use an AI builder model in the context of their SharePoint library. Yes, so, yeah, we have a very close partnership with them. That's working great. I think the scenarios are very similar. I would say maybe the scenarios SharePoint, syntex are a little bit closer, in my mind at least, to knowledge management, information sharing.

Mark Smith: 17:05
Yes, yeah, 100% yeah.

Gwenael Bego: 17:08
Ours is more of kind of an end to end workflow, where you're taking data from one end and kind of posting it in another system. So they're very similar and linked in the interface, but the workflows are a little bit different.

Mark Smith: 17:19
Yeah, interesting. You know now that Generative AI has come into the game and so let's take of pre-December 2022 AI builder to post December AI 2022. And I'm using that date loosely but the explosion of with what OpenAI did and bringing chat GPT to the public, how has that changed your thinking and application inside of AI builder?

Gwenael Bego: 17:53
Yeah, I think it's changed everything OpenAI did. Which was incredible on top of the actual power of the model is they built a very unique user interface in my mind that made it super accessible to go interact with these models, because the technology had already been there, you know, for a couple of months before, but they really created that sort of immersive chat experience where anybody can use the product, and so that really boosted up, I think, the importance given to AI models and completely shifted, I think, how we think about AI. You know it's not no longer a kind of a custom scenario or kind of an advanced scenario. It's an inherent part of everything we do in the power platform and at Microsoft obviously. So I think that has changed also some of the customer perspective around that for sure, and I think from AI builder specific kind of point of view, it's changed how we think about the model training experience, where I think in the past we had a lot of fit for purpose models that were good at doing one specific thing and if you wanted to build a custom version of that you would have to feed it your data, set, label items, spend time training the model, keeping the model fresh, and now with GPT it's kind of a very different ballpark. You have very generative general purpose model that can do a wide range of things, in some cases even better than some of the fit for purpose models, and so I think we're probably seeing a shift towards that model creation experience where hopefully in the future, you know, we will be able to use GPT to do much more than text processing, tax, but also maybe extracting content for documents and that will, you know, simplify the model tree creation process for makers. I think we're not quite there yet. A lot of these fit for purpose models. Obviously I think they run probably faster and at a more efficient cost than GPT, but perhaps we'll get there over time.

Mark Smith: 19:48
So, as of today, there's nothing. Correct me if I'm wrong. There's nothing in private preview or GAID yet in AI builder that is using generative AI, but oh, so there is so we've created a GPT action and the associated prompt building experience in AI builder.

Gwenael Bego: 20:11
So essentially you can go create a custom GPT prompt and you can save that as a function of sorts that you can reuse across the different elements of the platform, so in power, automate and power apps, and that is available in preview form today for US customers. We are currently in the process of expanding Gios as we get, obviously you know, more GPUs available to run these powerful models and have those be available locally across the globe, and we are working on making that publicly available and getting towards GA states. So that will be should be shortly, later this year.

Mark Smith: 20:52
Awesome. Where does AI co-pilot? Where does co-pilot come into the mix in the context of AI builder, or doesn't it?

Gwenael Bego: 21:01
Yeah, so co-pilot, I mean I think obviously folks have seen the news. It is very much a kind of a central theme that we are rallying towards and that customers see a ton of value in. Already you know it is an assistant that can help you do multitude of tasks. I think we're already seeing in early signs that people who use co-pilots are more productive, they get work done faster and they're happier with the products, and so that is definitely a strong part, I think, of our AI story. Within the power platform is to have a build a world where makers and users can have co-pilots all across the different types of business solutions that they develop to help them do things you know better and quicker. And so AI Builder is, I think, not yet part of the co-pilot ecosystem, but we're certainly working towards that goal to give users the ability to customize some of the underlying co-pilot behaviors with AI Builder.

Mark Smith: 22:09
Excellent. Tell me about data privacy and security. I know it's a big area, but just really your lens on it. And yeah, I mean Sachi recently announced, of course, the whole copyright side of things at Microsoft is taking responsibility, which of course sets a lot of customers at ease. They don't have to be concerned about that. It wasn't Sachi, it was Brad Smith that announced that, it might have been or Justin, I can't remember. But tell me about data privacy and security from the way you're thinking of AI, its growth, its proliferation, and what are you thinking in that context?

Gwenael Bego: 22:47
Yeah, for sure. So I think it's definitely a massive part about what we do. We want to give access to customers to these amazing capabilities, but for sure we want to be able to do it responsibly, I think, for one, to prevent any misuse of the capability. So we want the content to be accurate and just, and so there's a certain amount of responsible AI regulations that we're putting forward in the product to encourage human review. But in terms of security and maybe privacy more specifically, I would say probably the key part is around the data that you send to these co-pilots and to these AI systems. One of the key principles I think that we have at Microsoft is that we don't use the data that you send to the model to retrain and improve the model. I think that's a key customer point that comes up in very many customer interactions is what happens to my data? Will my data end up in some other kind of customers cloud? Will that improve the model? Will other customers know about my own data? And absolutely not. Your data is your data. We keep that private. We don't use your data to retrain or improve your model, and the underlying models are part of your Microsoft cloud. They are your intellectual property, and we don't use that for anything else.

Mark Smith: 24:07
So in that training side, I just want to clarify something there. You're not using it to retrain the generative AI model, the one that might come from Meta or her open AI, et cetera, but there would be a training or a grounding in your own data right that it does understand your data, so how your application of it would get smarter over time in the context of your own organization, your own data. What's not being done is that used to train back the master models that have been born onto the platform. Is that?

Gwenael Bego: 24:37
right, yeah, correct, at least not in the capabilities that are in product now. But yeah, that's the general idea is you're sending an instruction to the model through the form of the prompt and then you're giving data, essentially to the model to be able to ground that information to a specific context or with a specific output, but you're not actually custom training the model or fine tuning the model in itself.

Mark Smith: 25:01
Yeah. Final question, chat GPT. What are your thoughts? How are you applying it in your own life, how you personally, outside of your tech career in Microsoft? What are your observations about the future of AI? Are you excited, are you worried, about a dystopian future? Are you excited about a utopian future? What are your thoughts? What's your research? What are you thinking?

Gwenael Bego: 25:27
Yeah, I mean I'm very excited about Chat, gpt and the future of generative AI. I think it is a tool, and so you can use a tool for good, for bad, and so that's why it's important that you have the right guard rails to prevent malicious use. But the realm of the things that you can do with this and the speed at which it's evolving is quite mind-blowing. I use it a lot for content generation, I think, as a starting point to sometimes get ideas running or get the flow going. Obviously also the new sort of Bing with GPT integration I find as a much more valuable search tool than the traditional web search. So I use that a lot as a means to kind of tailor the output, and that is, I find, really a very great time saver and a great learning tool as well. I think I feel like I learned faster when I have the content that's already produced the way that suits the information that I was looking for, instead of having me to go through, you know, parse different sources. And yeah, I use a lot for those specific purposes. And yeah, I mean, I think there's lots of different applications, obviously, and it's a little bit too early to know how this will go. I think the spectrum of using AI to kind of solve humanity's most pressing problems around research in the medical field, I think are the sort of the best spectrum of where it can go. I think there's very interesting discussions also with Sam Altman, where he's kind of portraying this and it is a tool and so it can be used for harm and misuse as well. I'm confident in, if we have a kind of a common legislation and the right sort of safeguards, that the tool can be used for good and that we can use it to augment, you know, human capabilities rather than the opposite.

Mark Smith: 27:17
Well Gwen, it's been a pleasure to have you on the show. Thanks for coming on.

Gwenael Bego: 27:21
Thank you, Mark, appreciate it, thank you. Thank you very much. It's been a pleasure as well.

Mark Smith: 27:25
Hey, thanks for listening. I'm your host business application MVP, mark Smith, otherwise known as the NZ365 guy. If there's a guest you'd like to see on the show from Microsoft, please message me on LinkedIn. If you want to be a supporter of the show, please check out buymeacoffeecom. Ford slash. Nz365 guy. How will you create on the power platform today? Ciao.