Stellaris Venture Partners, an early backer in firms such as Whatfix, Mamaearth, Propelld, Kiwi, Zouk, Dashtoon, Slintel and Nestasia, among others, is looking to invest in artificial intelligence-as-a-service (AIaaS) or AI-led services startups. AI-led services is a new category where AI handles the core delivery work, and humans in the loop focus on judgment, quality, and client relationships.
“This model often requires augmenting AI with humans to guarantee accuracy, what we call a full-stack AI,” Alok Goyal, partner at Stellaris Venture Partners, who leads investments in AI, told FE.
The firm, founded in 2016, has so far backed this thesis with its investment in insurtech startup Pibit ai, accounting startup Atlas, and others. It is equally excited about AI on the consumer side. Here, it is looking for firms that democratise AI access for the next 100-200 million Indians – delivering personalised, high-quality services in financial services, healthcare, education and content at price points and quality levels previously out of reach. Other notable AI startups that the firm has invested in include Arrowhead, Drizz, Truva, Dashverse, Lumio and Orbitshift.
AI-native enterprise software is also a segment that the firm is actively evaluating. The firm believes this entire space is ripe for disruption, and incumbents will struggle both with their business models and their inability to attract top talent.
Disruption in Enterprise Software
“Within the enterprise world, there is approximately $500 billion in IT spend specifically for downstream services for large enterprise software like SAP, Oracle and Salesforce. Similarly, there is still a lot of legacy code, written in languages like COBOL, sitting inside large enterprises – particularly in telecom, utilities and financial services. Much of this is served either by IT Services firms on an FTE-based model or staffed in-house,” he said.
Physical AI has also caught the firm’s fancy. Goyal believes that the large opportunity in text, video, voice, and images is largely behind us, and the next big frontier is training general-purpose robotic or physical AI models. “These are fundamentally more complex, with multi-model inputs and multidimensional actions, and training them will require modelling physics at a granular level, demanding an enormous amount of real-world data,” Goyal said.
In FY26, Stellaris backed 11 startups across vertical AI applications, full-stack AI, voice AI, vertical quick commerce and services, and consumer tech. It hopes to continue maintaining its steady pace of 10-12 new investments a year, mostly entering at the idea stage, even before the company is incorporated. “First checks are just the beginning; we keep healthy reserves for follow-on investments in our performing companies. The real conviction shows up in how aggressively we double down on our early belief in the founders and what they are building,” he said. The firm is currently deploying through its $300 million Fund 3.
Shifting Mathematics
The most exciting trend that Goyal has observed in AI this year is vertical AI going after the compensation budget, not just the software budget. “In the past, we would evaluate a vertical by its software spend, typically 1% of industry revenue, which made most micro-verticals too small to build a meaningful business. That math has completely changed,” he said.
He added that the US GDP is $30 trillion, 75% of which is services. And, roughly two-thirds of that is OPEX, and over half of that is labour, which is close to a $10 trillion labour pool – people on phones, sending emails, doing classical workflows – that AI can now handle.
Goyal also said that agent infrastructure that provides enterprise-level reliability and security, consumer AI and physical AI are some of the emerging sectors that require more attention.