Barry Russell - The open-source foundation of every AI partnership

In this episode of Partnerships Unraveled, we sit down with Barry Russell, Senior Vice-President of Partnerships at Anaconda. Barry helped build AWS Marketplace from a two-pizza team into a global business, spent time at Microsoft and several data-space startups, and has spent just over a year at Anaconda translating that experience into how AI solutions reach enterprise customers at scale.

Barry opens with the principle he carried directly out of AWS Marketplace: remove as much friction as possible between where the customer is and the outcome they are trying to reach. At Marketplace, that meant letting customers try software quickly before committing to a multi-year contract. In the AI era, he sees the same dynamic: enterprises want to deploy agents fast, prove they deliver the outcome, and move confidently into production. Anaconda ensures that the open-source components developers use to build AI applications are secure by default, so teams can move from experimentation to production without hitting security roadblocks along the way.

What is actually breaking between POC and production comes down to two things: system dependencies that work in a dev environment often do not match what production needs, and CISO teams flagging vulnerability concerns in open-source components. Anaconda addresses both by ensuring the assets a team uses in experimentation carry into production unchanged, with ongoing visibility into vulnerabilities. Barry draws a practical distinction: it is completely fine to discover that an AI application needs a different model or some adjustment. What organizations cannot afford is releasing agents carrying underlying security vulnerabilities. Getting the foundation right from the start is what makes fast, confident iteration possible.

With 95% of the Fortune 500 already using Anaconda for data science and machine learning, the company arrives at the AI moment as a trusted workflow layer, ready to help customers close the gap between experimentation and production at the pace the market demands.

_________________________

Learn more about Channext 👇

https://channext.com/ 

Watch on YouTube  ► 

https://www.youtube.com/@channext 

#channelmarketing   #channelpartners

Transcript

Welcome And The AI Boom

0:05

Welcome back to Partnerships Unraveled, the podcast where we dive deep into the mysteries and the secrets of partnerships and the channel. My name is Michelle. I'm head of marketing at Chanex, and I'll be your host for today. And I'm really happy to chat with Barry Russell, Senior Vice President of Partnerships at Anaconda. Barry, I really appreciate you joining. How's your week been so far? Hey, Michelle. Yeah, I was really looking forward to this conversation. My week has been fun and crazy, as you can imagine, in uh AI startup land. Yeah, yeah. I I went to RSA last week and two things I really noticed. One, there's so much happening in cybersecurity now in AI, it's like almost it almost boggles the mind, right? You really can't even follow so many companies that you've never heard of, and they're doing like billions in revenue, and every single billboard is AI. So I can imagine that it's pretty intense at the moment. Yeah, and well, we had NVIDIA GTC the week before that. And so it was really incredible just to see the number of companies that were there that maybe weren't there the year before, suddenly on the, you know, the show floor uh and just walking around was incredible. Just all the different solutions that that people are out developing and how quickly, how quickly they can form a company and bring something to market. Yeah, I mean, that's another one of the like the baseline value of AI is that good ideas can be executed on quicker, right? I think also the market's really ripe for consolidation. If you look at how many companies are doing similar things, it will really make sense to bundle

Barry’s Career Across Cloud Giants

1:35

them together. But before we dive into uh AI market dynamics, could you tell us a little bit about yourself, your background, and your work at Anaconda? Yeah, sure. I've been at Anaconda for a little over a year now and joined to look after uh building out their partnership business, which really included the work that we do with the public cloud hyperscalers, the partnerships that we have with companies like Snowflake, Databricks, certainly our work with Nvidia, and also uh how we engage the channel, so the traditional VAR, as well as a growing group of SIs. And then before that, I was at a few startups in the data space, which was super interesting as AI was uh becoming more and more the center focus. I spent a good chunk of time at uh Amazon Web Services at AWS, which was an amazing experience. And then before that, I was at Microsoft doing some interesting things there around their cloud and some voice technologies. That's a pretty incredible career trajectory. Uh, just out of my curiosity, when you compare startups and scale-ups to the big guys, like what do you notice is the difference? And I wouldn't say what do you prefer, because I know that like they're totally different worlds, but what did you learn from the differences between the large corporates and the smaller scale-ups and startups? So the large corporates were different. If I focus on AWS, you know, back in the day, back in 2012, you know, they were six years in. Cloud was really starting to take off, but they really operated AWS as a bunch of independent PLs that were formed to deliver customers' value or solve a problem that they knew customers had. And that allowed each of the PL groups, including the one that I was fortunate enough to be part of, uh AWS Marketplace. It allowed us to operate autonomously and move super, super quick. And I mean, you know, making decisions uh, you know, within an hour meeting, not to, you know, not deliberating for days, shipping code incredibly fast, even back then, you know, dropping code production weekly instead of monthly. That autonomy and the way that they operated those different PLs was really the key to them innovating. And that was my experience. And then it mirrors what I've experienced at startups, certainly at Anaconda, where we're going to move as quickly as we can, you know, based on solid customer feedback and where the market's headed. But we have autonomy here to make decisions and move fast, depending on what what business we're looking after, in my case, partnerships. Yeah, I think it's a really smart move from a company perspective to take this kind of agile approach to building good stuff, releasing fast, failing fast, and building on what your customers are basically saying what they need. And I think that's something we'll we'll get to later on in this conversation is really about like outcomes versus features, right? But uh Yeah, I think if you look at your your trajectory within AWS, you did some pretty amazing stuff, right? Building AWS marketplace from something small into basically a global business for AWS. So first of all, congrats on that. That's a quite a great achievement. But secondly, taking your lessons into account, right, when you look at AI ecosystems today, what can you apply from building the AWS marketplace to helping AI solutions actually make it to market?

What Marketplace Taught About Adoption

5:05

Yeah, well, first, um, I have to say thank you also to there was an incredible team that built Marketplace. And in particular, we had an amazing product and engineering team, even though it was small. You know, we were really a two-pizza team, as Amazon likes to call when we started, which means we were really small and proving that it would work. What do I take away? Removing as much friction as possible for the customer to get to the outcome that they're looking for. To me, it's it's that simple. And I think we're in that space right now with customers that are trying to get AI native applications out there. Yeah, I think that makes a lot of sense. And it also echoes some rumors I'm hearing in this space, and it's really all about how services are done, right? That most most software companies have like a six to one services ratio, but the next real AI push is having AI do that services stuff for you. It actually doesn't help people do the job, it does the job. And and I think that makes a lot of sense, right? Really focusing on what the customer actually needs. Yeah, I would tell you like the the one thing that AWS Marketplace did was it it gave customers the ability to try something really quickly to see whether or not it worked without having to be locked into an annual or multi-year contract before they knew the solution could deliver the outcome. You fast forward now to where we are, and I think that's exactly what customers want to know if they're gonna deploy an agent. They want to be able to deploy the agent really fast, make sure it's secure. But just like those solutions from software vendors in marketplace that they could go test to see if it worked, I think they want to know that, yep, it delivers the outcome that they want, and so they can go to production. I think we're in the exact same spot. Yeah, it's it's fascinating. The go-to-market for the entire software space, solution space is just changing. And uh, it's really exciting to see where it

Anaconda’s Role In Secure Python

7:06

goes. I wanted to dive a little bit into how you approach partnerships with Anaconda, but maybe you could give like a little bit of background on Anaconda first, after before we dive into that. Yeah, um, so Anaconda's had a rich history, you know, for many, many years, well over a decade, being the root of data science and then of course machine learning. And the founders formed the company originally to help financial services firms do deep, deep data analysis that hadn't been done before, and do that using that language called Python. And what they realized as they went is that Python needed to be vetted and secured, and that it wasn't good enough to just use an open source variant where the companies, particularly the regulated industries and financial services, didn't know whether they were exposed to vulnerabilities. And so that had to be shored up. And then you fast forward, and Anaconda has literally been part of AI, you know, let's say the last several years, a couple of years before it became the thing everyone was focusing on. They were at the root of helping customers do machine learning. Um, and now we're taking that experience into AI. And I would just mention that, you know, we really do three things for customers. We make sure that whatever they're doing with Python as an open source language is secure by default. We help them build their end-to-end AI workflows in a trusted way at scale. So we're bridging the gap between experimentation and production. And that's really the value that we can deliver, including system dependencies. And then we're going to help developers go super, super fast. So if you put all those things together, you you essentially get a workflow where developers can build these AI native applications and use open source models in a predetermined secure way so that all those checks and balances are already in place and they can ship their agentic applications as fast as they need to. Wow. That's really impressive. Uh, you basically take the perspective of securing the start of the value chain. Right? Yeah. And I uh I think that makes a lot of sense. The start, I know the way I think about it is the start, yes, but all the way through production and the observability of what's happening in the assets, the bill of materials the customer is using from us, so that each time a developer team or line of business want to go build an AI native application, they're not starting from scratch and having to go through all the approval chain. Yeah. And I think that that will rapidly increase compliance procedures speed, like the actual cycle as well, right? I think that's one of the things a lot of companies are running into right now. They don't know what to look for anymore because the threat landscape is so complicated. And you're dealing with so many kind of disparate issues that you have to look at, and being insured that the core of your solution is secure seems to be a huge value add. Yeah, and

Agentic Coding And Supply Chain Risk

9:59

one of the interesting things that's developed, and I think it's just gonna, I don't know what the right factor is, but it's gonna multiply, let's say, times a million is agentic coding tools. So those agen decoding tools and people vibe coding, they also have to source assets, often from open source, in order to build the application that builder is requesting. We have to make sure it's not just humans, we have to make sure the agents are get grabbing the right secure embedded asset as well. And you can imagine how that's gonna explode. That also sounds massively complicated. Well, it it can be complicated unless you are pointing those agents at um secure repository and a bill of materials of assets that you already know are vetted and secure and can be observed. You know, if you do that, you know, and you standardize around that, then it takes away a lot of the risk. Yeah. I think that's a really smart approach. Um I'm really curious to see where where this market heads, right? Especially as we develop more new stuff and more complications arrive and arise and new threats that we've never even imagined arise. I think this is a really fascinating approach. Let's switch

Why Anaconda Goes Deep On Partners

11:12

gears a little bit. Let's talk about go to market. So uh obviously, this this podcast is called Partnerships Unraveled. We talk a lot about partnerships, whether that's in the ecosystem around ISVs, GSIs, all that all the way down to massive numbers of SMB partners. How do you scale, right? Many cybersecurity companies I talk to who have a channel first or 100% channel go-to-market are expanding their channel footprint to kind of maximize that scale. But with Anaconda, you're kind of doing the opposite. You're narrowing down to a handful of high-value strategic partners. So I have two questions. What kicked off this decision and what made you confident that fewer but like much deeper relationships would outperform that massive scale model? Um Yeah, it's a good question. So, first, and not to be cliche, but we're we're always going to go with what the customer wants to do, their preferred VAR, their preferred SI. Um, I think for us, one of the reasons we're focusing right now is to work with a set of partners across the US, UK, parts of Europe, and APGA that are really leaning into building and delivering value for customers. We're always going to maintain a large set of transactional partners because we want to be able to say yes to a customer. But by working with fewer on solution selling and value selling and where we're headed with uh with our product, it enables us to get much deeper with the customer, together with that partner, and really understand and learn what they're trying to build around AI, where they're building it, where they plan to deploy in production, which can be different than where they're building it. Um, and then discover with that partner other groups inside that company that are also looking to replicate or build their AI, their AI native applications. Focus for us right now, you know, we're a startup. Focus for us right now allows us to have more value-based impact uh at customers. Yeah, and if you look at the if you look at GSIs, if you look at ISVs, if you look at Vars, they generally do have more kind of situational awareness in the space, right? So selling that value as opposed to transacting becomes easier if they already know how to package these complicated solutions that really sit at the core of an organization's uh tech stack. Yeah, we're we're we're just I mean, simply we're just trying to go deeper with a set of partners. And part of it's learning too. We're we're I think everybody's learning this in this market now. We're definitely learning what customers are trying to build, where they're blocked, what what security or vulnerability challenges they have, and who are other partners that we need to work with too. Those could be other ISVs, it could be a cloud provider, it could be NVIDIA, etc. So it's really a go fast. I'm sure there'll be some mistakes, you know, which is fine as long as we learn from them, but it's really about going deeper with the customer. It's really that simple. Yeah, and it's it again comes back to that no one partner model fits all organizations, right? Even if you're a an AI, cybersecurity firm, whatever, it it really so depends on how you go to market and and where you fit into that ecosystem. Um and I I think the value of having large amounts of channel partners where you can also dive into the value, obviously, but it is more transactional. When you have a vertical slice like email security, it makes so much more sense to kind of spread that net wide. But when you have a core platform play, it's really important that all your partners are completely engaged and then in the know when it comes to your proposition. And again, how that fits into like really complicated tech ecosystems at big companies. So I think smart product produce. Yeah, and I would agree with you. And uh honestly, I think this AI market is changing weekly. You know, we hear something new each week. And there are different types of partners that are all of a sudden emerging. And then we brought a new product to market. We brought an AI development suite called AI Catalyst to market in November, running on AWS and just before reInvent. And we're learning a lot about how customers are using that. And so it's informing changes we need to make in our partner ecosystem. Yeah, I recently heard a really nice quote that was like you can actually use your VARs, your ISVs, your GSIs as your eyes and ears in the market as well. It's it's kind of a cyclical motion when it comes to information and how you can improve, right? Correct. Yeah, exactly. Exactly right. And um, I think everybody, at least our experience is that we can absolutely help a customer go from experimentation to production, you know, whether they're using an open source model or they're trying to get an AI native application into production, prosperitude services, or manufacturing, life sciences, all the all the highly regulated industries for sure. And again, we're learning real time what customers are building, and I think everybody is. We're not alone in that. And so then that if that informs, okay, for this particular solution that the customer is trying to build, these partners need to be involved. And that could change for the next application that the customer is building. The requirements could change, and I think it's happening that dynamically. So there's not like a static model anymore of, oh, these four or five partners are going to do everything. Yeah, it's really that tailored, personalized approach. I

The POC To Production Breakdown

16:47

really like it. Um you just mentioned that you launched a new AI solution, and apparently that's pretty rare because when we were doing this prep call, we we we talked about this crazy statistic that AWS shared, which is that only about 13% of AI POCs actually make it to production. I think this is also corroborated by other similar research from MIT that went viral. I actually read the report, which was really hard to find, by the way. Everybody was just quoting it, but I was like, I want to read what this actually said. But I think that's a pretty serious failure rate when it comes to POC to production. So, from your perspective, what what's really breaking like between proof of concept and production? Yeah, they mentioned that publicly. It was part of their AI Center of Excellence, uh, where they were calling on partners to essentially come help them and customers get those applications into production. From my perspective, what's breaking? It's the customer is experimenting, whether they're using a model or they are building some type of agentic application, they're experimenting, but then two things are happening. They go to push to production, and the systems dependencies that underlie the POC or the experimentation are suddenly different than what the production application needs to run. And we solve a chunk of that for the customer at Anaconda. Or there are concerns from their CISO organization that the application that is being produced is is rooted in some vulnerabilities because it was built with some open source components, and therefore, you know, maybe it has to be completely rebuilt. Maybe it's a different stack of technology to get it to production. So I think there's some fear around well, what happens if we release all these agents? Where where are the vulnerabilities that exist? Yeah, it aligns so strongly with that gap between like if you're vibe coding, which I I'm not a developer, right? But I I had to try out the different vibe coding apps available right now. And and what I did notice is the gap between what you see, so what you prompt and what comes out, and then actually turning that into a production ready solution, there's still a gap in there that I wasn't able to fill. And that made me think, like, okay, but then then what am I doing, right? What am I missing? Because apparently I'm missing quite a lot of intel here that I need to use to deploy this. And I can imagine that if you're working with like these big partners and big customers, they are also going to be more hesitant to accept these new solutions. So how do you kind of combat that that potential trust gap? So we show them that whenever they're going to experiment and do in a POC, for example, building in the public cloud or building in their sovereign AI environments, we show them that the assets they get from us are going to be the same that go into production. We show them that they're going to have visibility into vulnerabilities that might arise and how we remediate those quickly. And then we show them that some of the systems dependencies that might change from local development on a laptop or a much more localized environment pushing to a larger production one. Maybe they're using somebody else's infrastructure. We show them how those systems dependencies remain consistent and they don't have to rebuild the application. Yeah, I think that again, it comes back to expectation versus reality. And um, I think there's two interesting points here, and that's almost that both software vendors and partners become trusted advisors in a different way than only having real relationships between each other, but also becoming a trusted advisor when it comes to the security of the tech stack. So I think that's also going to require a mindset shift in the entire kind of partnering space and the software or software space. You know, the other thing is the only way the customer is really going to know if their AI native applications are doing what they wanted to do, they've got to get them into production. That might seem obvious, but um, you know, there there are gonna be failures, but they need to go quicker. I think we've all seen how quick this market is moving and people coming up with new ideas and solutions. They've got to get those agentic applications into production. Um, they've got to make those agentic applications smarter over time as they learn in order to find out whether the thing works or not. Yeah. So it's really about failing fast, focusing on outcomes and making sure that your entire value chain is secure. Yeah, fail fast, but without exposing your company to underlying vulnerabilities in the model or the agentic application that you've pushed out the door. It's fine to fail on the outcome. Gosh, this thing just didn't deliver the rate of return or the credit card approval process improvement that we thought it was going to have. Maybe it needs some tweaks. Maybe it needs a different type of model attached to it. What's not okay is failing on security and vulnerability. Yeah. It's something that you have to do right, you can't get it wrong. I totally agree. And it's it it's really interesting to me that that what you're describing now reminds me so much of ESG. I don't know if you know, like Environmental, social governance. And that that concept came far after companies had their entire like supply chains, value chains built out. And one of the things that I found fascinating in ESG was something that was called scope three. And it's basically assessing the impact that your suppliers have, and that that becomes a part of your score, basically. And what I'm hearing now is how Anaconda approaches this is basically preemptively doing that. And I think that's a really, really important part of this process is where so many companies are just building. They're building and they're not taking into account these dependencies, these potential flaws when it comes to pulling assets from areas you're not even aware of, that doing this right from the start, down the line, we're going to figure out where everything is going wrong, right? So doing this right at the start is such a huge value driver. Yeah. And the cool thing about Anaconda is we've always been there. So what I mean by that is 95% of the Fortune 500, as an example, already utilize us. And they utilized us for their data science and machine learning, which has now moved into AI. And so we're already a foundational layer there. And so when we think about working with partners as AWS or Microsoft or Databricks or NVIDIA companies like that, and they're working with customers to utilize models or build their AI native applications, we're already a workflow layer in place. And it's really easy for us to be added to that solution or the workflow that the customer is trying to do on that platform. And so I think that's what is really making it powerful for us to engage deeply with this AI partner ecosystem and explain to them, you know, our heritage of machine learning. And when we do that, the you know, proverbial light bulb definitely clicks. We've already established trust with the customer. Exactly. You've already delivered outcomes in the space, you're already a part of their stack. You're building on that. And uh yeah, that's a really smart approach. Yeah, exactly. Exactly. So this has been fascinating.

Next Guest Nomination And Closing

24:07

There's one thing we always do in this podcast, and that's that we ask our guests to nominate the next guest on the podcast. So from your perspective and your storied career, who should we have next? There are, gosh, I'm gonna make maybe one person smile, and the rest are gonna be like, how come you didn't say my name? So this is like this is a really tough question. You know, there's one person in my mind that stands out, and she has such a rich history in the partner ecosystem, and in particular, um, around open source, which is just a critical component to customers building AI. I mean, we heard all about open source, you know, at NVIDIA GTC two weeks ago. That person is Christine Puccio. She's currently at SUSE. I would absolutely have her on because I think she's got a unique perspective on working with the partner ecosystem and what's happening around AI too. I love it, especially having that tech focus again. So I like splitting my time right on this podcast between channel sales, channel marketing, and tech. And one thing that these conversations about ecosystems and like deep tech do is they keep me sharp, right? They make me remember that this is all still like really complicated and super cool. So I love that. No, I'll definitely uh send her an invite. Before we wrap up, do you have any final insights, words of wisdom, tips or tricks that you'd like to share with the audience? Yeah, I mean, I think partnering, you know, people talk about go to market and co-cell and marketplaces, which are near and dear to my heart, and all those are great buzzwords. But at the end of the day, you better understand the outcome the customer is trying to achieve. Um, and once you do that, then you can engage a partner or partners in the right way, particularly as a startup. You know, it's different if you're on the other side and you're a huge platform like AWS. It's different if you're a startup and you know, trying to do things like Co-Sell or work with a customer. That's one and two. Um, I think it's early days, and I'll speak for myself, still learning what customers are really trying to build with their, you know, with their AI initiatives. So I have a lot to learn. So focus on outcomes and keep learning. Keep learning, know the outcome. I think this is applicable to anybody who's listening. So I love that. Um, Barry, thank you so much for sharing your thoughts and taking the time to speak with me. And uh, for you, dear listeners, thanks for tuning in and see you in the next episode. Thanks, Barry. Thank you.