2025 marked the transition of AI from an assistive tool to an autonomous teammate.

We’ve come far from the early days of 2025, when models like GPT-4.1 and Gemini 2.0 Flash worked mostly in autocomplete mode, like an intern in need of constant hand-holding. Fast forward to today, we have Gemini 3 Deep Think, Claude Opus 4.5, and GPT-5.2. These advanced models are trained to execute end-to-end workflows, make decisions in milliseconds, and operate 24/7 across languages, channels, and contexts. This is a categorical shift in AI capabilities.

What makes this transition particularly significant for India is the timing.

The Indian startup ecosystem is evolving in lockstep with this shift, adopting AI as both software and a strategic lever. We're seeing founders integrate AI agents into customer support operations that autonomously resolve queries, deploy voice AI to handle complex conversations across multiple Indian languages, and build tools that automate knowledge work at unprecedented scale.

Poorvi Vijay, Investor at Elevation Capital, believes this presents a fantastic opportunity for the Indian ecosystem to take center stage.

From an India-first lens, this is an incredibly exciting moment. Unlike the SaaS movement, where India lagged by nearly two decades, AI companies in India are competing on par with global peers. We have the talent, the scale to test at, and a pragmatic approach that global markets are only beginning to appreciate. This is India's moment to lead, not follow.

The Agentic Force Multiplier

For years, AI has been a faithful assistant. Alexa will play all the right songs, Google Home will dim the lights, and Siri will call whoever you want. Useful, but fundamentally reactive.

2025 marked the inflection point where AI moved from the sidelines of a supporting role to take complete ownership of workflows. This shift was driven by three technological unlocks:

• LLMs can now parse messy human language with near-human accuracy
• Open-source frameworks enable real-time tool orchestration
• Continuous feedback loops allow agents to quickly

Together, these advances crossed a critical threshold where AI agents deliver finished outcomes.

The impact has been immediate and profound. Across customer support, sales, and marketing, AI agents are handling millions of interactions that previously required human judgment. We've witnessed this firsthand through our portfolio companies. At Meesho and Swiggy, significant portions of customer service operations now run autonomously while maintaining quality that customers find indistinguishable from human support.

Within this broader shift toward autonomy, we’re particularly excited about voice AI as India's big unlock for the next billion users. Indian users are inherently comfortable with conversational interfaces. This puts founders in a strategic position to build voice-first products that global markets will eventually adopt, much like how we defined the playbook for mobile-first internet experiences a decade ago.

The applications emerging across India validate this thesis.

Beyond common use cases like voice search and assistants, founders are tackling unique problems: AI companions for the Bharat audience (like Ira by Rumik), language learning tools helping millions improve their English speaking skills (like SpeakX), and voice-enabled financial services reaching users who've never filled out a form online.

With India's AI agents market expected to grow at a 57.4% CAGR through 2033, we've been investing heavily across the stack, including GreyLabs AI to build voice agents for financial services across sales, support, and collections, and Synthio Labs to create clinical-grade voice AI for life sciences. These investments reflect our conviction that the infrastructure layer for autonomous AI is as critical as the models themselves.

But we also realize that building production-grade agents has been deceptively complex. Most product teams spend 6-9 months just to make an internal pilot. This is why we invested in Adopt AI, a platform designed to conveniently build agentic experiences on top of static interfaces in less than 24 hours. We believe this solution allows more and more companies to tap into the power of agentic AI and deliver a seamless customer experience.

How India is Driving AI Adoption

2025 made one thing abundantly clear: AI adoption in India is driven by a cost-benefit analysis. Instead of chasing moonshot projects, Indian operators are solving concrete ground-level problems with measurable ROI. And they're doing it in environments where cost and speed matter more than cutting-edge experimentation.

We're seeing a clear pattern of this pragmatism in action. AI adoption in Indian startups is fastest where workflows are clear, data is abundant, and revenue impact is direct.

• GTM leads because sales and marketing teams can plug AI into existing tools like automated outbound campaigns and AI SDRs.
• Engineering teams move almost as fast because the feedback loop is immediate. One founder shared how his team went from one engineer using Copilot to all 100 adopting it in under two months.
• Operations and back office is where India's ops-heavy businesses see disproportionate gains. Fintech, logistics, and edtech companies with enough data to train models, enough volume to justify investment, and enough cost pressure to demand results.

But speed of adoption doesn't guarantee success.

Most founders hit a wall with data infrastructure. The overwhelming majority underestimate how fragmented their data actually is. With unstructured or siloed data, meaningful AI implementation becomes nearly impossible. Success starts with strengthening data foundations and treating AI adoption as an organizational change problem.

And even with the right infrastructure, we're seeing a more fundamental shift reshaping what it takes to survive in AI-first markets. The bar has moved. Adaptability is your biggest differentiator in a landscape that shifts too quickly. Distribution strategies change and product plans pivot, sometimes in weeks, not quarters.

The winners will be those who can build exceptional products while staying light on their feet, ready to pivot the moment the market signals change. Krishna Mehra, AI Partner at Elevation, frames it this way:

Every six months, you have to survive and then thrive, then survive again. You have to keep re-earning PMF because things change that fast. Adaptability and agility are the only things that keep founders on top.

Cost discipline buys you the runway to experiment. Speed of execution buys you the chance to learn. But adaptability is what ultimately separates companies from experiments that couldn't keep pace.

Organizational Maturity is the Differentiator

We spent the better part of 2025 studying how hundreds of startups approach AI adoption. The patterns tell a stark story. Roughly 70% of companies remain trapped in "experimentation mode" with scattered pilot projects, ad hoc budgets, every department building in isolation, and optimizing for demos rather than deployment. The remaining 30% have crossed into systematic adoption, embedding AI across product, GTM, and operations as core infrastructure rather than experimental feature sets.

Teams in the 30% bracket build institutional knowledge, establish feedback loops, and improve continuously with each deployment. Meanwhile, the 70% restart from zero with each new initiative.

What separates the two groups is:

• Disciplined problem selection: The 30% of founders aren’t trying to "do AI everywhere." They're hunting for use cases that are repetitive enough to automate, data-rich enough to be accurate, and valuable enough to justify ongoing investment.
• Clear, cross-functional ownership: The teams that make progress have somebody responsible for driving AI adoption. This role evangelizes, measures, pilots, and maintains tight feedback loops.
• Strong data infrastructure: Most founders don't realize how fragmented their data actually is until they try to build something on top of it. Information lives in siloes. The leaders spend months (or quarters) establishing clean data foundations.

The final differentiator is mindset. The companies capturing real value don’t look at layering AI onto existing workflows. They redesign workflows from scratch. One of our portfolio companies replaced tier-one customer support entirely with autonomous agents. These AI agents now handle 95% of inquiries end-to-end, cutting resolution time by 60% and cost per ticket by 70%.

New Partnerships

We partnered with Adopt AI by leading their Seed round. Adopt AI builds a layer of dynamic agentic experiences on top of static app interfaces. These agents understand intent and act on behalf of users. Read our investment memo here and watch our Day One podcast with founder Deepak Anchala here.

We also led the Series A round for GreyLabs AI. GreyLabs are building deeply verticalized, AI-driven voice agents tailored specifically for India’s highly regulated banking and financial services sector. We believe they solve real operational pain points in contact centers with measurable enterprise outcomes. Hear our conversation with founder Aman Goel in the Day One podcast here.

In the voice AI space, we also backed Synthio Labs. They're creating a clinical-grade voice AI operating system for the pharmaceutical sector, addressing a $30 billion annual market that remains largely manual. We led their Seed round.

We’re also proud to partner with SpeakX, an AI-powered English learning for middle India with over 1 million monthly learners and 200,000 paid subscribers. In our conversation with Arpit Mittal on the Day One podcast, we discuss how the platform leverages AI to empower users for whom English represents confidence, career, and dignity.

Building the Playbook: Enabling India's AI-First Transformation

Throughout 2025, we rallied around a single goal: helping Indian founders move from curiosity to real AI execution. We knew that founders don’t need more theory around what AI can do. They need exposure to what’s actually working, the confidence to build hands-on, and access to peers already shipping AI in production.

With that in mind, we started by putting founders in the builder’s seat. Our AI Hackathon in Bengaluru brought together 60+ operators from across our portfolio—product managers, ops leaders, and business teams who deeply understood their problems but had never written code. In one high-productivity evening, they arrived with real pain points and left with functional solutions.

We also worked on shortening the distance between global AI innovation and the Indian ecosystem. To that end, we hosted deep, technical sessions with teams from OpenAI, Anthropic, Google, and ElevenLabs.

OpenAI’s Thomas Jeng spoke candidly about platform direction and India-specific priorities. Anthropic’s Daniel Delaney and Chloe Ho walked through production-grade multi-agent architectures. Google’s Generative Media team showcased how their creative stack unlocks new possibilities for content-heavy businesses. ElevenLabs gave us front-row seats to their latest capabilities and brought together multiple teams to showcase their voice AI use cases.

But the strongest learning came from peers in the trenches. Our AI Show & Tell roundtable series brought operators together to share their on-ground playbooks—what they’d shipped, what broke, and what they learned. Fintech players discussed deploying AI under regulatory constraints. Consumer brands shared how they scaled personalization and support. One founder’s eval framework became another’s starting point; one agent design saved others months of experimentation.

We also hosted multiple experts and builders to share tactical insights with our community.

Ethan Smith from Graphite shared his tried-and-tested playbook for Answer Engine Optimization (AEO) and tips to double down on Reddit. A vibe-coding workshop with Abhijeet Jha from Lovable showed 70 operators how to ship live MVPs in 90 minutes. Khilan Haria from Razorpay shared how his product org moved from scattered pilots to systemic AI adoption. And Vikash Rungta from Meta focused on a simple truth: most AI products fail not because models are weak, but because systems engineering is misunderstood.

And we even took learning beyond conference rooms. A walk in the iconic Cubbon Park with Chiefs of Staff revealed a critical pattern: AI adoption is deeply personal, learning spreads through osmosis, and generalists who span functions are often best positioned to spot transformational workflows.

Across all of this, something bigger emerged. The conversation shifted from questions about the potential of AI to hands-on approaches and experiments with AI. In 2025, we built a community of operators willing and eager to learn in public. In 2026, we’ll double down on shared learning initiatives to narrow the gap between experimentation and execution for the Indian ecosystem.

Looking Ahead: The Shift From Helping to Doing

We’re already seeing how experimentation is giving way to systematization where teams work with standardized guidelines and processes.

As we enter 2026, the stage is set for deeper automation and AI integration in how businesses operate.

We're grateful to all our founders, partners, and the broader ecosystem for an incredible 2025. If you're building in AI—infrastructure, applications, vertical solutions, services-product hybrids—we'd love to hear from you at ai@elevationcapital.com.

Here's to building what comes next.