Sonar’s Harry Wang on AI Software Development: “Trust and Verify”

hi I'm Jam deguire here at AWS reinvent show there's a lot going on at the show this year today I'm joined by Harry Wang vice president of growth and new Ventures at sonar Harry really good to talk with you today thank you for having me Tim all right so I want to talk about artificial intelligence is software development how it can create better code and also how important it is to have a verification system to ensure that code is really high quality before we get into that please give a naal description what a sonar is how how do serve clients absolutely uh sonar is best known among the developer in the customer Community as for our product son Cu uh we've been around for 15 years we ultimately in in the business of helping developers as well as the organizations to build a better software by writing better code and we do so by supplying the tools and techologies throughout the software development process all the way from the moment the developer writing code in the to the the SSD workflow and to the delivery production when they do the testing and in production they find the issues how they address the back in code so we very much in the philosop of Shifting not only shifting left but of starting from the left and trying to identify issues and also help developer address the issues up FR early in process okay and today at uh you knows reinvent in Vegas I'm very proud that yesterday EV has announced SAR as The Rising Star Technology P of the Year well congratulations so great I mean obviously the company's far from New but it it's still great to get the award and get that recognition yeah yeah um so let's talk about Ai and software development I I wonder where are we with it now is is it closer to beginning is it closer to Min adoption um I read the statistic by Gartner they said that by the year 2027 70% of professional developers will be using AI based software development tools I would think it's earlier than that I think by next year most developers will be using a AI tools what is your s of where we are with AI and and software development yeah I agree in terms of adoption of the software development tools powered by AI is probably actually a high percentage will get to 70% much sooner um in general in terms of AI being a dominant you know force in sofware development I would say we in the early stage in this stage if I me to use the autom driving as analogy we're not at the level five a to driving like a vehicles wayo running around on the street we're more likely between somewhere between first control and maybe level three at uhuh okay so it's it's it's there it's playing a leaning role in the industry but it's not quite all the way adopted yet yeah it's really around if you think about productivity it's very helpful for developers to leverage the tool to stay productive right instead of writing every single line of code and being able to actually just r view what's being generated by the AI what's missing at this moment that in other words why we can't ditch the steering wheel is yet is as an organization as a team there's a lack of actually standards on how you want to ensure quality how you want to ensure security as an organization and how they enforce that as an organization the ultimate goal is actually building application that's TR fory to the user and to the to the ACT customers you know what I think companies are wondering is you know how do we get the most from artificial intelligence in our code process I mean it's it's still a somewhat new technology in terms of you code development um what are some best practices you recommend I would say trust and verify right so there are um many many different tools and many many different code generation models out there in general in my personal experience when I trying out the different one it's actually pretty powerful I came from engine mag myself I haven't coded for real for production for a long time I found the very helpful for me to just use those tools and start actually you know um building hobby applications and uh on on the flight they they work in general it's um um so that's the trust part right but the verify is really actually there's a higher bar as the organization when you build professional business Enterprise ready type of vacations whether it's about actually security standards or you want to make sure the application you build is actually predictable and trustworthy that elements is really something you still want to have human being in that human can be the developer them for chunk of cod actually modit writing but also could be the organization that's the cesos that's the cios or the vpf engineer need to set a policy so um trust and verify right now is at this stage for the adoption s dep is that it's best practice so really all code needs to be reviewed it makes perfect sense it needs to be created either by AI or by human everything must be reviewed before it goes out the door until we have a higher degree of confidence on the code and the application being gener application I would say yes the they actually get to the heart of accountability who is the accountable for what's being written and what's deliverable uh right now it's a it's actually human we have actually have a you know some reporting from customers about the outages the global outages for major organization and that when they trace try discover what's going on it was for one line of code checked in by a particular developer but when the customer actually you know start inspecting what's going on the developer say oh that's not my code I generate so there is a confusion about who's accountable and you know software Engineers special jior Engineers may let their guard down given that the increasing higher quality that AI is actually generating um so that accountability right now is actually I think one of the H button issues within oranization as they start rolling out from you know AI tools from just a pro concept to the ACT much wi production I I think on that note there is a sense among some people like oh AI is going to create it must be perfect that's really not true not yet yeah I'm off miss this I think that we will get it one day just actually it will be a journey yeah well let's talk about the sonar products uh there's an on-premise product there's a s product it's an ID product what's important for people to know we meet the where our users are and uh we have a lot of customers in banking sectors in the regulated space and you know the governments Federal governments the on Prim deployment is the most assured deployment model in terms making for the code stay on premise within the Water Garden air gap in many cases and then for developers individual developers they work in the IDE all day long right so being able to deliver the Sonic Cube capability within the IDE is uh you know it's the most natural interface and then we have Enterprise customers who prefer to they they have a cloud first strategy as a organization and they want to get started with analyzing the code and detecting issues right out there with a you know within minutes and seeing this integration with the GitHub and all the other you know devops platforms so we'll meeting we providing the bestas insurance regardless where you were so tools integrate with other environments absolutely we integrate with the most popular IDs out there V code intelligy Etc but also most of the CSD work well let's look to the future I think people will get you know excited about that they want to know where it's going and and how they can get ready now when you when you look to the future of AI and software development what do you see and and do you see any new products from sonar perhaps yeah I um I will stay to the company line um but I can I do can I can share you some of the recent announcements um and also give a general direction to where we're heading the um I think while I'm speaking about that stage where we are in terms of AI in software development we are in that trust and verif verified stage what it really means is as organization as a team they like to start really Define the standards as a team and as organization but also enforcing that standards you know across the development process whether it's the uh like you are relying on a human gener code or AI gener code Sonic cube is the tool that really facilitate and make it easy to you know orchestrate this work collaboration right so we are now soner to a code Insurance capability a few weeks back and that's something available right now and we have a customers and Enterprises are actually deploying that um we also at the same time want to help our and user developer user to stay productive you think about the you know uh well Rich collection of tools they can use today you know it's all about how do we make AI capability right in the interface right in the product surface they using we have 7 million developers out there using S Cube and we're making AI cix suggestion capability recently in Cube so right at the moment where we surface the issue to the developers with a clicko button they can actually see the suggestion how fix it they don't need to Pivot to another interface they don't need pivot they can pivot that from the son Cube son Cloud right but but also they can actually stay in the IDE as well right so um so these are the recent capability we announced going forward as you know we as a community move from the you know early stage of AI adoption to maybe the automet driving adoption we are very much excited about this agend AI world so I I think as a company we adapting to how software is build and there will be a growing population of developer out there um some of them may actually coming traditionally from more of business user perspective and we want to be rela than stay um the true s compation to help them you know the AG think of the ipace is really interesting I would think in terms of writing the code for that it's going to be far more sophisticated code because the agent AI applications can do more just a simple bot right that seems like it UPS the importance of verification oversight because that code is going to be some complicated code so you really have to make sure it's verified yeah absolutely I think it's there there are two layer to that a gen the supper development world one is really there um there a need for centralized gatekeeping in terms of the testing the the and applications from performance testing from functionality test perspective and there's also a degree of how the different agent potentially from different vendors from different you know sources should talk to each other and collaborate each other right so we as industry don't yet have a framework or a platform or standard and protocols about how the agent talk to each other right yeah so I I think it's an open question for for everyone participating in the space you know what I should look like and how um you given the AWS event we have a lot of Partners and we have a conversation that how we all each other see that future and how we can work together towards that future the the other thing I find really interesting about sonar is that it's it's actually overseeing human code and AI created code so it's like it's it's really policing or or or verifying two different code bodies I can you talk about that a little what is what does that mean for the system is it particularly challenging or is this part of what it does or is like is is there more mistakes found or what I don't want to put one down necessarily like but is is is there one that's more difficult not necessarily I think the um the fundamentally a generate code also came from the same you know uh the source code that's Avail written by humans and Point train through the models right and represented and predicted through thei model right um so from uh the in terms of detecting the issue in terms of the nature of issue being surfaced in AI or by the human developer we find it's actually very similar it's there's volume is certainly picking up with all the AI agents out there right um but in terms of really just detecting the issues the same technology we have been tried into is very B the what is really interesting is really this notion the accountability we talked about earlier and and how you really start helping the individual stakeholders within the team within organization start orchestrating and start developing a common standards and putting that in place has been a challenge right so we traditionally as soon CU as person C product has been widely adopted at a project level at A Team level and then with the Gen AI with the geni in software development we seeing a more and more corporate wide adoption this speak to the need for company want to Define common standards and we were happy to be in that Center of that you know that that Journey great here I think you said it it's a lot of good stuff and I hope we get a chance to speak again sometime my pleasure and looking forward to it thank you yeah thank you

This transcript was generated automatically from the video's captions and may contain errors.