The Neon Show
00:53:29 · Siddhartha Ahluwalia, Bhaskar Ghosh
[00:53:29] S0: And, you know, outcome based pricing... [00:53:33] S1: I say one thing about your Whatfix example? So if I remember correctly, what was plagued in the CRM space, which has traditionally been a very important space because it's a top line play. Sales is an investment in the top line of that company. And sales ops leaders are willing to spend an arm and a leg to hit their numbers and go top line. The moment you go into HR, the moment you go into marketing, which are probably cost centers more, the answer is not true. [00:54:06] S0: It leads to a very interesting point, if the industry has to change for pricing, outcome based pricing, what kind of tailwinds are already there and need to be there in the future to support a complete change?
00:54:19 · Bhaskar Ghosh
[00:54:19] S1: Migration? You should be able to migrate your software or your service very easily. Number one. That is not true yet. So when an outcome based AI services provider comes in, they can't just say I'm the best. You have to show how you're to migrate that. Number one. So that is a requirement. Second is, you have to have very measurable proof points around quality. It has to perform things at least at that quality that has to happen. Third is, and this is something that we have not thought about very deeply is, the buyers have to have the right tools to judge things. Which today, on the software side they do have. As I pointed out, remember you were asking me the network question about services? On the services side, there are no automated tools where you can decide a certain BPO or software services outcome is as good. So I think those are some of the those are the requirements. I wouldn't call them headwinds. Tailwind is proof point is happening much faster today. I mean, the innovation happening with Gen AI from the services side of how... We were just talking to a young company which is doing insurance workflow automation for business insurance of certain subset. And we saw the workflow that the workflows are automating. This is not in the guide where... This is not in the source of truth area. These are these are not in the these are in the workflows for say parsing documents and giving answers and stuff. Let's assume that. And we have seen that in our company, Azure also. I mean the speed at which they... This company has done correct workflows so quickly is insane. So I'm saying, when they are going and selling to fairly slow moving insurance industries, and the insurance industries are suddenly waking up and saying, Woah, I can do it this faster. The adoption curve is changing. So that's something for you to consider. That the buyer side is beginning to ask questions because there is secular pressure coming from their board, their investors, their CEO. They are also becoming more aware. So I think we are in a slight bit of a honeymoon phase, but from what little I saw from some of these seed level companies, the quality of the outcomes is insane. It is quite insane, right? That's one.
00:56:42 · Bhaskar Ghosh, Siddhartha Ahluwalia
The other thing that is a slight bit of a detail, but is interesting for you to note is, many of these new AI enabled businesses that you are building, whether it is pure software or services, has a Palantir esque flavor to it. That in delivering this service and in delivering this software, you will need this, you know, the way Palantir built it, what they call Forward Deployed Engineering. We are finding that that is a whole new tier of talent that these new services companies and these new software companies have, which is also enabling adoption much more rapidly. I'll just leave it at that. There's some fundamental new talent movements happening which is making adoption faster. [00:57:27] S0: So Bhaskar, we are coming towards conclusion of a podcast, you know, amazing discussion, lot of food for thought. I might need to listen to this conversation twice to prepare for a third one. Right? But very important. [00:57:41] S1: We should pick a vertical. Hopefully by November, I'll be thinking of some new stuff and we can add to this thing because as I said in AI enabled services, the domains matter. It's okay to construct horizontal thesis, but domains matter. But yes, back to you. [00:57:59] S0: And I would love for you to meet some of our founders who are building hardcore, they're building compliance, [00:58:09] S1: Would love to. [00:58:10] S0: For banking, for example. Enterprise infrastructure domain, right, that's your specialty. [00:58:17] S1: That's how it used to be, true.
00:58:17 · Bhaskar Ghosh, Siddhartha Ahluwalia
[00:58:17] S1: That's how it used to be, true. [00:58:20] S0: Right, so how is that shaping up right now? What are the areas left for founders still to conquer where you think are tailwinds because lot of this model company that we were discussing. I think the biggest challenge for founders today are model companies because they are eating forward revenue, are eating backward revenue. Forward revenue means they're eating application revenue and they are eating backward revenue which is cloud revenue and because these model companies are extremely hungry, I have never seen hungry companies as these right, their aim is to become $10,000,000,000,000 companies, they're not they'll not be satisfied at the pace of growth to become a trillion dollar company, at least my observation is that people might disagree with it, I don't know what your views are. [00:59:05] S1: No, I agree with you. [00:59:06] S0: Yeah, so enterprise infra, what are the areas for founders to tap in, what are you learning from your existing company, you are invested in some of the fantastic companies like Nile.
00:59:16 · Bhaskar Ghosh
[00:59:16] S1: Yep, Nile, Yugabyte, Acryl, which is now called Data Hub. So I would say in enterprise infrastructure, which is a vast stack, includes data, includes security, includes orchestration, includes site ops, SI, many departments, many personas. One particular sub area, which is a legacy area that I am super excited about to see how LLMs will impact them is in all things observability. So the area say that today, Datadog, Splunk, Elasticsearch, all of them play in, the old New Relic and all of them. Massive budget, software that you pay for to operate your business and to know if your site is working okay. Right? And then debug problems. I feel... And the underlying infrastructure is using Mongo in some cases. It's using Clickhouse in some cases. But a lot of it is based on the back end of Datadog. It's based on Elasticsearch on the Elastic Stack. I think this area requires a couple of new companies with LLMs as their backbone. Throwing a challenge out. The interesting thing about this area is, what kind of founders do you want to find there? I think this area is both hard with massive upside, is because the DNA of founders needs to... There has to be a co founder who knows applied AI really well. There has to know... Be a co founder who knows the observability and the whole thing about what we call MET, the nature of data. It's like metrics, events, and logs, and traces, and how Datadog and Splunk are used today. So you need a product expert, a domain expert in product. Third is, you need a founder with systems and data background. Traditionally, it's hard to bring founders like that together. But I feel if you find a founding team like that, in these areas, where now with say if you see open source time series foundation models like TimesFM coming out of Google, I think there is a massive opportunity. So in the observability area, off the site observability. Similarly for security observability, I think there is a massive area at SIEM. I would say these are backward facing categories, which are massive TAM, which will be impacted by LLMs and by using Chen AI. Suddenly, am interested in the forward facing categories that are opening up. I think LLMs will be data hungry. I think unstructured data now coming into blob stores, into the unstructured data warehouses, into what is below a data bricks, into what we call data lakes.
01:02:03 · Bhaskar Ghosh
I think data lakes will see more and more semi and unstructured data, because you can squeeze more enterprise value out of them, which has not been true. So infrastructure around semi and unstructured data lakes, one of our companies, Aaron is working in that domain, I think that's going to be a massive area. The ingestion tier for that data is going to be a massive area. Because now you're not just bringing in site data and sales data, you're bringing in all kinds of other data. Handwritten document data, PDFs, Google Docs, you can do a lot more with them. So all things data observability, around data infrastructure, around storage in the area below LLMs is going to be important. So I can give another example, our company Acryl, which is based on the LinkedIn open source project called Data Hub, has now rebranded itself as Data Hub, is seeing secular tailwinds in AI metadata coming into the So metadata I feel there is a lot of opportunity in all things data, all things extraction. Now coming to LLMs per se, and something you asked me in the prep questions, is observability monitoring feedback loops, compliance, governance around deploying agents and co pilots in the enterprise is going to be a very important area. So there are at least two parts to it. One is companies which are doing orchestration of these tiers itself. In old school, now old school companies like LangChain, LangFuse, LangGraph, next gen companies of that sort which helps you orchestrate agents. Companies like interchange gateways which are called MCP. I think of it as an interoperability play. How will MCP scale, grow, be compliant, be fault tolerant? So these are all backbones of agentic deployments inside the enterprise. And all things monitoring, governance, feedback loop and observability. I feel these are all going to be significant time areas within the enterprise. And the second category that I just pointed out is partly forward facing. And I would say the extraction layer, the observability layer that I pointed out for side part are partly backward facing but will get revolutionized. So very exciting times. Two comments. One is a repeat comment which is the type of founding teams you need for each of these areas are special. Should be a combination of AI infra, of applied AI, of product specific expertise, as well as databases and systems. That's kind of exciting.