The Vanguard of AI and Data Strategies for Competitive Edge with Ryan Wexler

Prepare to unlock the secrets of successful data infrastructure investment with the guidance of Ryan Wexler, VP at Unusual Ventures. Transitioning from the meticulous realm of data engineering right into the heart of venture capitalism, Ryan offers an unparalleled perspective on pinpointing the most promising data companies. This episode is a treasure trove of insights, where we uncover the critical ingredients that elevate a startup from a mere niche player to a scalable powerhouse in the competitive data sector, all thanks to the strategic support Unusual Ventures provides.

As we navigate the intricate evolution of the modern data stack, it becomes clear that while data warehouses once lured enterprises with their cost-effectiveness, burgeoning scales of operation have led to some sleepless nights over soaring expenses. This is where our discussion takes a turn into the groundbreaking realm of data lakes and independent storage solutions – the silent disruptors offering a respite by decoupling storage from compute costs. Listen in to understand how businesses are strategizing to harness these technologies for optimized data management, marking a seismic shift in the tech landscape.

And then there's the undeniable surge of AI – a tidal wave of innovation that's transforming the face of industry after industry. This episode peeks behind the curtain of AI integration, highlighting how trailblazers like Druva and ThoughtSpot are embedding AI to revolutionize their offerings. As we dissect the proliferation of AI tools, our dialogue serves as a compass for startups and enterprises alike, emphasizing the importance of a laser focus on ROI and the wisdom of keeping those burn rates low amidst an ever-changing economic backdrop. Join us for a journey that not only demystifies the complexities of data ROI but also navigates the myriad choices in the expanding universe of AI adoption.

What's New In Data is a data thought leadership series hosted by John Kutay who leads data and products at Striim. What's New In Data hosts industry practitioners to discuss latest trends, common patterns for real world data patterns, and analytics success stories.

Transcript

0:16

Hi everybody. Thank you for tuning in to today's episode of What's New in Data. Super excited about our guests today. We have Ryan Wexler, Vice President and Investor at Unusual Ventures. Ryan, how are you doing today? I'm doing well, John. Thanks for having me. Excited to be here. Absolutely. We've been wanting to do this episode for a while. I've, I've wanted to pick your brain for a long time. You have great insights on the market as a, as a venture capitalist and someone who's been analyzing the space. For, for a long time. So, you know, we'll get into a few of these topics here. There's a lot going on in the market and you have a great read on whether things are going first, I just want you to tell the listeners a bit about yourself. Yeah. It's funny you mentioned, like we've been trying to do this. I think we spoke about doing this, like. A year ago, something like that. And I've known you since I joined last firm since like, what, 2018. So like five, six years at this point. But yeah, excited to be here. So quick background on myself. Like you mentioned, I'm a vice president here at unusual ventures. Unusual Ventures is a seed Series A firm based in the Bay Area, where we invest in across all things enterprise infrastructure and enterprise SaaS. We were founded by two people, John Vronis, who has a long time experience as a venture capitalist, as well as Jody Banzal, who is the founder and CEO of a company called AppDynamics, which we sold to Cisco, and is now CEO of two different companies, Harness and Traceable. Unusual Ventures is set up pretty differently from other VC firms that exist. We go pretty deep with the founders that we work with at the seed stage. And we have a lot of resources that most VC firms don't, for example, we have a internal head of sales. We have an internal head of recruiting. We have an internal designer, which I haven't seen any other VC firm have head of marketing and things like that, that we offer up to our portfolio companies, if, and when it makes sense for them to have it. So I joined the team here about three months ago. So I'm one of the newest members on the team. To cover data infrastructure, AI, ML, cyber security, and developer tooling. And before this, I was actually at another VC funds where we're investors in Striim called Dell technologies capital. I was there for six years doing most of our data infrastructure work, worked with a bunch of really great companies, got to invest in some really smart people. And before that I was a data engineer myself. So I worked at a hedge fund called Magnetar Capital on the data engineering side. Being able to buy a lot of third party data assets and this is the weird time between like HDFS and Snowflake where I was kind of like all in on Redshift. I mean, like the first separation of storage and compute. Yeah, and that kind of leads me to where I am today. Yeah, that's why I love your perspective because you know, you were in the trenches building the pipelines and now you're applying that same expertise, but from a very high level analyzing the market and still having a good sense of, you know, What's what's fact versus fiction when it comes to claims of startups and vendors and some of the trends in the market. That being said, you know you know, you look at a lot of data companies. What makes, From a venture capital and startup perspective. What makes a data company investable for you? Yeah, it's a great question. And I really like answering this one because We have seen over the last few years an explosion of data tooling that exists that I wish was available to me when I was useful and out of VC. We've seen it just how every wave comes. We saw it after the big snowflake IPO and all the VCs saw that there's a lot of money to be made in data infrastructure. And so try to invest in every single different category that exists. We are entering a period where that's kind of, I don't want to use the word deflating, but let's just say slowing a lot of data teams. I speak with are cutting costs right now. Just like most teams on the planet are reducing. And so what we look for when we speak to start up specifically right now is, you It used to be enough where if you had a unique insight in a very specific subcategory where you could try to, you know, sell to the hot Bay area companies and build out an open source community and get massive adoption and try to follow the likes of some really big success stories like dbt and FiveTran. I'm not sure that's. As possible as it used to be, we're seeing a lot of companies focus on selling to the enterprise on day one. And really focusing on the revenue and customer facing problems that exist today. What makes a company investable going back to your question right now is a clear line of sight towards building a large sustainable business. And not just becoming another point solution that kind of gets muddied within the waters that is the modern data stack and data infrastructure today. So what I like to tell companies is it's fine to have a starting point, but make sure that starting point is in a highly strategic area where you can continue to grow and expand your product coverage. To build the next snowflake or the next Databricks or the next whatever it may be, as opposed to getting stuck in that, you know, modern data stack, I fit in this little subcategory of a subcategory, I'll own that space, get to like 10 million ARR, and then it kind of stalls from there, I really try to find like, what are those strategic areas where you can build really large standing, differentiable businesses, Which is easier said than done, but at least that's what we're looking for. Absolutely, and that's a, that's a great point that, you know, there are a lot of, you know, you can call it point solutions, things that are very, maybe niche in the, in the modern data stack, and like you said, it's a, it's a Point solution for a subcategory of a category in the stack and yeah, you can get to 10 million really quickly by, you know, tuning your internal growth metrics and kind of growth hacking your way there. But you know, like you said, it sort of hits a wall. So and maybe at a time that was, that was a, you know, a good proposition for, for early stage VCs to get into those types of companies, maybe, you know, in 2018, 2019, the peak zero interest rate phenomenon era, but you're saying now you're really looking for those companies that have. a real path to becoming the next Databricks or Snowflake in terms of revenue, performance, market penetration. Would you say that's a good summarization of your view on that? Yeah, I think a lot of investors in a lot of startups have, I don't want to say, overestimated the TAM of the categories that they're working in. But they tap out pretty quickly on selling to their friends and their friends of friends to get to that, like I said, five to 10 million ARR range. And don't really have a stronger story of how they can grow past that point. And so I'm not saying that it's a bad idea to like, start with like a single point solution or cause you have to start somewhere, but you need that strategic story of, Hey, we're going to start in this single category by owning this. Specific workflow or owning this specific subcategory, we have a story to build that larger standing business. I'll give an example. The clear one that's been talked about over the last couple of years is dbt. They were ones that try to own the transformation gap that exists. Once data sits inside of a data warehouse. And if you are able to own that transformation layer where all the data is actually being transmodified and transformed so it can be pushed into a dashboard, you get a lot of really interesting insights and metrics that can lead to a larger company. They've gone into like semantic understanding of how metrics are being created. They're probably at one point going to go into actually creating the ingestion points themselves into the warehouse. There's a lot of really interesting ways that you can go about a business like that. If you start with that highly strategic area, one of those areas that's really important to own in the modern data set continuum is the transformation layer. And because of that, that's, we decided to invest in lead the seed in a really early stage company called Tobiko data. We invested in Tobiko, which is the team behind SQL mesh and SQL glot, because they had a really unique insight in how a data transformation tool should be created given that their experience being data engineers and data architects themselves. And we believe that they created the correct way to do data transformations at massive scale for enterprises. SQL mesh does some really interesting things. So it's able to lower compute costs for large enterprises. As opposed to doing larger table scans, like dbt does, they're able to actually look, do things like caching and materialized views. Where you can really lower those transformation bills. Yeah, data transformation is a really interesting area. Obviously, DBT has this huge community. They started as Fishtown Analytics. You know, it looks like the overnight became a big sensation around, you know, thousands of people in the Slack community exchanging ideas. But when it comes to, Productionizing and going, going to production and operationalizing a lot of these transformation workloads, there isn't a strong consensus on how to do it. And there's new concepts coming in that the warehouses are bringing to the market, such as, you know, lake formats, dynamic tables, incremental materializations. And at this point, we're seeing a lot of innovation that teams can do on top of the compute that's being offered with lake houses, warehouses, et cetera. So when it comes to transformation you mentioned a great portfolio company of yours that's, that's bringing a new perspective on, on how to do it. What are some of the best things that data teams can prioritize going forward? Yeah, it's a really great question. I think every single data team that I speak to right now. Is really interested in cost reduction and getting the most out of what they have to utilize today. We, as you mentioned before, we're entering an era where interest rates are no longer zero and CFOs have to now do buy in for any new product or any new back hire that's being filled today. And so most data teams right now I speak with our focus almost entirely on besides AI, they're focused entirely on cost reduction. There's a lot of ways that they can do that. They're focused a lot on like tool, tool consolidation, but not just that. They're looking at building out their infrastructure stacks in differentiated ways than have existed before. Really focused on lowering their compute and consumption bills for things like Snowflake and things like dbt and are exploring new product offerings that can really lower those bills for them. Definitely. And, you know, Snowflake, BigQuery, Databricks, Redshift, you name it. They're all doing work to help customers optimize their costs. And, you know, in the early days of adopting the modern data stack, you know, cost was one of those things that was, you know, a lot of these data warehouses came off the shelf. You could start with a 20, 30 K budget and, you know, instantiate all your workloads there, start your migration, et cetera. But now that it's reaching scale these data warehouses are great because, you know, you can store a lot of data. And it's not going to be particularly expensive to store all that data. And it's getting even cheaper with these data lake formats. You know, you have Delta, you have iceberg, you have things like Hootie all coming into the picture. Now in terms of optimizing the cost further, transformation is a big part of that, right? Because as new data comes in and is being fed to your lake house, you don't want to recompute and, you know, rebuild all your tables from scratch, right? You want to just change the things that have incrementally arrived to your, to your warehouse. So do you see that as one of the value adds that your, your portfolio company adds? Yeah, definitely. And you said something really interesting about the separation of storage and compute. The rise of Snowflake and Databricks and BigQuery and Redshift was because of this. But businesses are starting to realize that just because the storage itself is super cheap and they're able to increase the level of the amount of data that they're storing today, the cost of consumption and the cost of that compute have not gone down in a similar fashion. So, yes, you can store so much more data and so much more information, but once you're storing all that information, your tables get so much larger. The level of computation and the level of spend it requires to do anything interesting with that data is actually increasing at such a fast rate that businesses are really struggling with how much their snowflake bills have risen because of the level of adoption. That this huge amount, this explosion of data that they've really been doing for storage has happened in the enterprise. Going back to what you said about like Iceberg and Hoodie and Delta Lake. Look, a lot of enterprises are adopting these technologies because they don't want to be locked down into a single solution into a single data warehouse. The reasons that they would adopt an independent table format as well as an independent storage solution like S3 or like MinIO in an independent query engine like Trino or Presto or Starburst or Bodo or the like. One other company, AHANA, which IBM acquired, and a few others like, DuckDB being like the hot one today, like smaller tables, is businesses are realizing that they don't want to be locked down into these single warehouse providers, and are really focused on how can they lower their The overall spend for that compute bill. So they believe that bringing a lot of these technologies in house, as opposed to being stuck with one of these providers that gets all their money from this computational bills, is a really smart way to do that. Yeah, definitely. And the, the ideal topology that a lot of people are talking about, I don't know if it's proven yet. Is yeah, you have this open lake in your cloud with super cheap storage. And AWS is announcement this week around express zone was a step in that direction where, you know, you could do really fast lookups and inserts into S3. Not everything's thought through yet because it's really, it's, it's, it's just about, you know, having kind of. Compute that's local to the storage and things like cross region replication aren't there because obviously you do that. That's going to slow things down. However, the division is still, yeah, you have this, this open cheap data lake format. Yes, you have your, your, your, your high end warehouses, but you know, you're not reliant on them for all your compute and reporting on the data and then you can still have these. You know, data science point compute workloads, things that are just processing the data that's sitting in the lake, you know, all at commoditized prices. So I think the, the, the, the market is certainly going there. So it's, it's great to hear that, you know, you're seeing that from, from your end as well. I don't think the technology is quite there yet. Do you have any thoughts on, entrants that are helping to make this better? Yeah. What I'll say is, yes, these are definitely newer technologies. But they've been adopted at scale at some pretty large enterprises. Like I'll just like, I know for Iceberg specifically, it's being utilized in businesses like Airbnb and Apple and LinkedIn and. I believe Salesforce also has like a pretty large instance of it nowadays. So while it hasn't really gotten to the level of maturity where it's like, Hey, you can just like pick it up and utilize it for like beta teams that don't have hundreds of people just yet. It's been proven. For these larger enterprises that have reached the level of scale and complexity that requires them to bring a lot of these technologies internally, as opposed to just utilizing third party data warehouse. Your question was around, like, are there any, like, businesses that have been created to help utilize these? Definitely. What the creator of Apache Iceberg, Ryan Blue started a company called Tabular, which is focused on helping bring Apache Iceberg to the masses. He's working a lot with companies that are still in the Hadoop and HDFS world and bringing them on to these new age technologies like Iceberg and S3 and MinIO and like the separation of the table format and storage solutions. For a hoodie yeah, the team behind hoodies started a company called OneHouse, which is trying to be like the all in one solution for the data lake. And then obviously Databricks has their own offering Delta Lake, which I believe is fully open source at this point. They've open sourced it slowly or like in parts or something like that. Yeah. And I'm sure that they saw the rise of Iceberg and Hoodie and kind of had to open source more and more of it over time. Yeah, and all those things are, have, have really interesting adoption in their, in their own rights respectively. One point I'll make about Delta is, yes, it's, you know, Databricks made it very mainstream. Even Microsoft is adopting it in a big way. For instance, their, their, their new announcement of Fabric. Their, their key data warehouse data lake compute workloads are also using delta as the open storage format. So, you know, if you use fabric as a lake house, yeah, you are using delta to serialize and store the data, which is. You know, good to see that there's some interoperability across these tools and all these formats. It's going to be very interesting to see, you know, how you know, whether you're a startup or or enterprise, you know, which, which table format you choose, because. It would be, I could imagine it being complex, like, Hey, these workloads all run in, in Delta and these ones are in you know, iceberg and, you know, some are in Hootie and yeah, you know, there's, yeah, you can theoretically convert the data. I think they all have ways to do, like, you know like one function to convert, like a set of files from whatever parquet to Delta or Delta to iceberg. But, you know, you don't want to constantly be doing that type of, you know, disk IO and work on these, you know. Theoretically large scale data sets. Right? Believe Microsoft had that realization also. So they have, I believe they call it OneLake where they wanted like a single solution as part of fabric where you could query data, whether it's on iceberg or hoodie or delta lake or whatever new table format arises in the next few years from whatever large enterprise. And Microsoft Fabric itself, going back to what you said, although it's still very early, Microsoft recognized this trend, which is people don't want these point solutions anymore. They want like a single centralized solution. That's really able to focus on lower costs for both storage and compute. I haven't ran into too many enterprises utilizing it today. So pretty new product, but it has a lot of excitement from some of the enterprise I speak with that already are huge Azure shops. Oh, yeah, absolutely. And the great thing is, you know, I work with a lot of companies through, through Striim, our, our data ingestion, data ingestion and data streaming product that we're already, you know, using us to move data into Azure, whatever it is, like it could be event hubs, could be synapse, could be you name it. And I'm asking them like, Oh, are you, are you looking at evaluating this for fabric? And they said, Oh, we already are using, you guys don't just say it in a fabric. We didn't have to do anything different, which is a good, good thing to hear from customers that, you know, they're actually able to, realize Microsoft's vision of one lake where, yeah, if you ingest into any one of these Microsoft products that the data is going to be unified and centralized and, you know, not siloed. So I love hearing that from, from, from real world. Enterprise customers were doing this at at scale. And you know, it is exciting to see what Microsoft is doing. All the all the cloud providers are actually doing an amazing job you know, responding to to to market feedback. So the other thing that, you know, is certainly interesting is, you know, we're we're in this. Macro environment where, you know, it's not the same zero interest rate phenomenon that we had in the, you know, 2020 time period. And I would say 2022 is the big kind of vicious pullback, in early 2023 as well. Things are sort of stabilizing now. How do you see data startups handling this current macro environment? Yeah it's a great question. So there's definitely been that pullback in 2022 and 2023, like you mentioned, where people are trying to do more with less. I do agree from like everyone I've spoken to. It seems things have stabilized and we've reached whatever a new normal is. The way I would tell data startups to handle this new environment is how I tell every startup to handle this new environment. Really focus on reducing your burn in any way possible in doing actions and activities that have very, very clear R O. I. For data specifically, there have been a lot of, let's say, events and conferences and new emerging trends where the ROI for those types of things, wasn't super clear. I won't like say anything specific, but I've been to a bunch of data events where there turned out to be not that many customers, and it was really just data startups selling to other data startups. And I think those types of events, although there are a ton of fun, it's really difficult to show a clear ROI when you have to go back to the board and go to the CFO of your company and say, we spent X amount of money and we didn't get that much out of it. So just do the things and cut back on like how every startup really is nowadays. It's not necessarily. Differentiated for data teams, but everything you do going forward, does it have to has to have that very clear ROI story? We'll see what that means for the data world, like in general. But I don't think data is going away anytime soon, and I'm still a believer, so things will be fine in the long term. Absolutely. Absolutely. And I'm glad you gave a shout out to some of the the data events where, yes, it did feel like kind of a vacuum of startups selling to startups and, you know, everyone kind of exchanging bottles of Kool Aid. I love community building. I love building out brand recognition, all that stuff, but When times are tougher, there has to be that very, very clear ROI story for anything you're doing when expenses are, you know, non negligible. Yeah, certainly, certainly. And, you know, what are some of the, you know, we talked about transformation, we talked about data lake formats, you know, what are some of the other trends in data that you're excited about right now? Yeah. So we talked about data transformations and how that's a very strategic angle. We talked about the disaggregation in the warehouse and some of the table formats and query engines that are emerging. I think one of the super interesting areas that. I know that you and I said, like, we want to talk too much about AI, but I feel like we just have to bring it into the conversation. Yeah. Is the level of interest I'm seeing with unstructured data and unstructured data management. So. The modern data stack and all these data warehouses have been so focused on structured and tabular data for so long. And it's obvious because there's very clear use cases for those types of things. You build a dashboard, you build a BI report, you can build an executive report that you bring up to the board or to your CEO, whatever it may be. But unstructured data has always been one of those things that's like, You know that picture of like the iceberg where it's like 10 percent of the iceberg is actually showing above the water and then 90 percent of its underneath unstructured data has always been that like 90 percent where you're hoarding all of this information, but you just have no idea what to do with it. Unstructured data can be everything from like, you know, like image data, audio data, video data. Anything that doesn't sit clearly inside like a tabular CSV report. And we're, I'm starting, it's very, very early, but we're starting to see some companies actually utilize this vast amount of unstructured data to enrich their product experiences. A lot of that, like I said, is due to the AI and large language model hype cycle, which is bringing a lot of these. Let's say I don't want to say forgotten about categories, but let's say not as often discuss things back to the real world. It's very, very exciting to see, like, you know, 90 percent of the data stored actually being used for, like. Real and interesting use cases. Yes, and that's one of the kind of side effects of the, you know, the new storage formats is, yeah, you can use, you can start indexing unstructured data and using these sort of cosine similarity queries to see if there's any relation, right? It doesn't need to be like a exact SQL query. With you know an exact results, you know, where you're matching two strings literally. Right. So it is that that's one of the nice things. And, you know, some document stores had had part of this part of these capabilities. Like if you look at, you know, elastic search slash open search and solar and Lucene and things that were indexing documents. But now with, you know, vector extensions, it's, it's a lot more powerful. It's not always 100 percent accurate, but as long as you have a human in the loop it's, it's, it's pretty close. And it gives you a lot of power and control and flexibility around, you know, what to do with things like documents or you know, poorly structured data, PDFs, you know, because, because companies want to be able to leverage those insights as well, right? I mean, like you said, document stores are not new. MongoDB was founded over a decade ago, I think in like 2007, 2008. And they've been talking about this trend for a long time. I think what is new nowadays, though, is the ease of adoption for some of these new technologies, given the rise of language models and given the rise of chat GPT. People are able to build internal use cases and POCs. Much, much faster than was ever really possible before. And that's kicked off like, again, like you mentioned vector databases. Our team is an investor in an open source vector database called Quadrant, which is being utilized by open AI directly, as well as large businesses like Twitter to power their new Grok model. We're starting to see these like bottoms up developer motion businesses being built. Because that ease of utilization and that level of complexity dropping so much that it's now possible to do these interesting things that people have talked about for a long time. But we're always, you know, one step away. It always required an extra month to figure out. It always required that one extra tool. So it's super exciting to see that level of complexity really dropped pretty tremendously. Yeah. It sounds like a very exciting company and that one that's being leveraged by. Yeah, the heavy hitters doesn't much get much bigger than open AI and, and, and the work that X is doing with Grok. And, you know, I think we'll see a lot of adoption of those types of technologies. You know, speaking of that from, you do a lot of research on the market in the state of enterprise budgets, you know, where do you see budgets shifting and being allocated going forward in the data and AI landscape? Yeah, every chief data officer I speak to is all in on Gen AI and discovering use cases right now. It's like a CEO top down initiative for these businesses to explore at least how they can be utilizing foundation models and large language models to improve their businesses. We're still in that, like, let's call it early hacky days where teams are given some budget to build out those internal POCs and play around with all the various open source models and vector database offerings and everything that's really out there. But I do believe that just given the widespread interest, that something interesting is going to come out of all this. You asked about like where data budget, where budgets are moving for data specifically, it's definitely stalled. I think the last couple of years were like the years of people actually building out their data teams and rejiggering everything given the rise of the data warehouse in the specifically like the cloud data warehouse and that's separation of storage and compute we talked about. Look, the data world is still growing. We're still in early days of maturity. It's not like everyone has figured out their data strategy going forward. But I do believe that we've kind of hit peak data startup, I'll say, where if you're just building another point solution, it's really hard to get that new budget, at least for the next, like, you know, like let's call it 12 to 24 months, a lot of data teams that speaking with I'm speaking with are really tightening their belts. And so you have to be very, very, very focused on driving those highly successful, strong ROI stories. AI is a whole different thing right now, like I said, but for data teams, it's not the easiest going. Yeah, that's that's tremendous insight. So it seems like on the data side, it's, you know, budgets are, are, are stalling. Cost optimization is, is, is critical on the other, on the other hand, gen AI is this big top down initiative, like you said, from CEO visibility, everyone wants to get AI. into their customer facing experiences and into their, their back end internal operations to, to, to further optimize them. And, you know, it, by it, and it really always does come down to either making money or saving money within the, within the business. So there's a lot of opportunities there for sure. Do you feel like it's a gen AI has reached the prime time where it's able to have that type of impact for, for enterprises? I mean, I'll tell you like in the last, like Call it six to nine months. Every single enterprise customer I speak with brings up AI and large language models and chat GPT in some way, shape, or form. Even if the conversation wasn't supposed to be about that, it's just top of mind for like everyone. I speak with, your question was, have we reached prime time? Just yet? Chat GPT itself is literally only a year old. I think it's a year old yesterday. We're recording this in like, December 1st, 2023. So I don't want to say that we've hit prime time just yet. But we are seeing a lot of really interesting use cases that just weren't possible. I spoke recently with the head of generative AI at Druva, which launched their internal chatbot called Dru. Which a customer is able to interact now with the entire Druva platform utilizing a chatbot that just wasn't possible before. I've spoken with companies like ThoughtSpot that have utilized generative AI across their entire product offering. And they're heads down really working on text to SQL and making it like the best it can possibly be. I've spoken with teams at Notion that are using Gen AI for Text summarization and autocomplete and things that really were, let's say, possible before, but were much, much more difficult to actually implement. I think it's definitely real, but we're still in that early POC beta stages outside of those. Obvious success stories that people are talking about. Yeah, really appreciate your feedback there. I'm definitely seeing similar signs from when I work with enterprises and you know, what I've seen here at AWS reinvent where it looks like, yeah, there was a lot of flashy. demos and people are talking about the art of the possible, which is amazing. And then you have some, some pioneering products, obviously like chat, GPT, and, that, that, that certainly have, you know, market recognition and, and are generating revenue you know. The thing that's really blowing me away right now is just how much choice there is. And, you know, I'm walking around the expo hall of AWS reInvent. It's not like it's just new, shiny products talking, you know, pitching AI. It's, you know, legacy companies that have been around for over 50 years. Talking about how like they're the solution for AI. And then of course, the, the cloud providers, it's not like, oh, hey, we have this one AI product. It's like, no, like all our products are part of your AI. Strategy. And, you know, even if you want to build your own internal chat GPT, we have like six products that you can build that on. Right. And it's, and you ask them, okay, which one should I use? And even the people working at those companies are not exactly clear on it. And, you know, how would you say a data practitioner or data executive shouldn't navigate all this choice for adopting generative AI? I mean, it's just funny you mentioned that I was just like driving to the office today and like I 280 in the Bay Area and every single billboard is talking about AI in some way, shape, or form, like, whether it's like a data company or an enterprise in for a company, or like a fintech SAS app, whatever it is, they're always like, Your AI copilot or your AI partner, whatever it may be, for how I'd recommend data teams and AI developers to look through all the choices out there, I would say, don't start at the info level because like you said, there's a lot of different options that are out there today, but start out on figuring out what those projects and you really want to do, and figuring out, like I said, it all goes back to that clear ROI.. I'm meeting with a lot of teams that are testing and playing around with all the info providers that there are out there today without having like a clear reason to utilize all these AI tooling. Once you figure out what project you really want to go after, whether it's an internal chat bot, like you mentioned, or improving your product in some way, shape or form, or completely modifying like the user experience to be like chat based. Like mentioned, I would say tr do your best to go with the tried and true partners that you've been working with before. Look, no one likes the process of onboarding a new platform, a newer provider. I invest in startups, so it's what I do day in and day out. So I'm always gonna say, go with the startups and go with our portfolio companies. But a lot of businesses I speak with are going with. Like you mentioned, you're at AWS reInvent right now. AWS Bedrock is pretty good. They're partnering with a lot of third party providers, like for Vector Databases, they partner with MongoDB. They have their own offering. I don't think they've partnered with any, they've partnered with a couple of startups. Nothing that I can say just now, but stay tuned. A lot of business I speak with are just trying to wrap their heads around the entire market right now and what use cases they can really adopt. So don't get too sucked into the infra vortex that is everyone offering AI tooling and just figure out what makes most sense for your business. Yeah, great advice. You know, work with the partners that you've proven success with, and, you know, there's going to be a lot of options, but, you know, ultimately, you don't want to do anything that introduces too much risk into your business. I do, you know, my, in my sense is that a lot of, Adoption around AI is going to be more of an evolution within companies rather than a revolution. And you know, there's people who disagree with me on this. And, but I, I really do feel like teams need to figure out how to first Incrementally improve their internal operations and experiences with AI with the incremental add ons that their existing partners are adding, like if you like, like you mentioned, like, if you're already an AWS shop, you know, it's going to be a very low barrier to entry to adopt their, their AI products, right? You don't have to rock the boat and migrate all your workloads to, you know, another service just to adopt AI. So. That's some of the benefits of all the flexibility and choice in the market. And I would also say like, if you are a newer entrant, like a startup you just have to lower the level of complexity requires to deploy you into production. So like, I'll give a quick plug again to the quadrant team, which we're investors in, they started out as the only open source vector database. And I believe they're now the most popular with like. Over 10, 000 get hub stars or whatever it may be goes up every single day. By being an open source first company and having a pretty permissive license, they're very easy to adopt for these enterprises, whether it's a large business or the weekend hackathoner, or, so if you do want to build a company in these types of categories where. The incumbents are not asleep at the wheel. AWS is doing interesting things. Microsoft's doing interesting things. Google is announcing stuff as well. You have to really lower that level of complexity and that ease of accessibility to break into these types of businesses. Whether it's open source, whether it's bottoms up community adoption, you have to start somewhere. So make it as easy as possible for your customers to utilize you. Absolutely. Yeah. Lowering friction is amazing. At the same time, you know, like you were mentioning in the beginning of the podcast. To make a company investable you need that clear path to revenue. And you know, the, and that's, that's the, the, the, the tight rope that a lot of these open source companies need to balance is like, you're not giving away too much of your IP, in the open source product. You want to make it as useful and add value as much as possible, but you know, a real clear value add that people are willing to pay for. And you can scale your business on with your With your enterprise or subscription models. And of course, that's been proven out very well with companies like Databricks and Confluent and many others. So, Ryan Wexler, Vice President and an investor at Unusual Ventures. It was great having you on the podcast. Where can people follow along with your work? Yeah, definitely. Reach out to me directly. Feel free to email me at any time at rwexler@unusual.vc I'm pretty active on like LinkedIn and Twitter. So just search me. You should find me pretty quickly and the unusual ventures team here We run a podcast and a blog that's very active Just go to the unusual ventures website and you can see our startup playbook and a bunch of content We put out about like you were talking about How you should navigate that open source versus like revenue market. We put out a lot of content, so reach out and hopefully use as much of it as possible. Excellent. Super valuable stuff. Thank you, Ryan. I'll, we'll have links to that down in the description for those who are listening. Ryan, thank you for joining and thank you to everyone for tuning into today's episode. Thanks for having me, John. And thank you to all the listeners.