AI Code Can be Messy. Now, VCs are Backing Startups That Can Fix Flaws

Until now, AI investments have largely flowed to companies developing foundation models and compute infrastructure, but venture capitalists are now betting on the next wave of value creation.

For Rahul Agarwalla, Managing Partner at SenseAI Ventures, the future lies in startups that help enterprises manage the complexity AI itself creates. The firm led a recent ₹7.5 crore investment in Sherlocks AI, which helps engineering teams investigate software incidents and identify root causes. 

Rather than relying solely on frontier intelligence, investors are increasingly backing tools that make AI-driven systems more reliable, scalable, and useful in production environments.

"As more and more code is produced, the problems it creates need to be handled," Agarwalla tells AIM. "Code is never perfect, whether written by humans or AI."

The widespread adoption of AI-assisted software development is fundamentally changing how enterprises build and operate technology. As AI tools generate larger volumes of code, organisations inevitably face greater complexity, more interactions between systems, and a growing number of potential failure points.

"If you're going to have a lot more code, that means by definition you will also have a lot more incidents," he adds.

The Emerging Market for AI-Era Infrastructure

The Sherlocks AI investment highlights a category that is attracting increasing investor attention: tools designed to support AI-native software development and operations.

According to Agarwalla, the company stood out because of both the market opportunity and the founders' technical understanding of the problem. He points to the team's operational experience, domain expertise, and ability to build sophisticated technology for complex enterprise environments.

"This is a deeply technical problem, so technical depth is absolutely necessary," he stresses.

However, technology alone is not enough to secure venture funding. SenseAI's evaluation focused equally on how the product integrates into enterprise workflows and whether it can build trust among engineering teams responsible for mission-critical systems.

"Trust is a massive factor now," Agarwalla highlights. "These are production systems."

The emphasis on reliability reflects a wider evolution in enterprise AI. While early adoption centred on experimentation and productivity gains, organisations are increasingly deploying AI tools in environments where downtime, errors, and operational failures carry significant financial consequences.

Investors are now paying closer attention.

AI Investment Playbook has Changed

The rapid pace of AI development has also transformed how venture capital firms assess opportunities. Agarwalla says the standards expected of founders today are significantly higher than they were even two or three years ago.

In the past, building an impressive prototype or product demonstration was often enough to attract investor interest. Today, such capabilities are largely expected.

"Every team is doing that. It all looks good. It all works," he points. "You have to peel the layers much more carefully now than you had to do five years ago or even two years ago."

The rise of foundation model providers such as OpenAI and Anthropic has added another layer of scrutiny. Investors are increasingly asking whether startups possess defensible advantages that cannot be replicated by a future model release or a larger technology company entering the space.

"The threat of the large language models is clear and present," Agarwalla says. This has pushed investors to focus on factors such as proprietary data, workflow integration, customer adoption, and domain expertise rather than purely technical novelty.

Competition for capital has intensified accordingly. SenseAI reviewed approximately 1,200 companies last year and invested in only six.

"The benchmark for the team has to be 99-plus percentile in every criterion," he highlights. A recurring mistake founders make, according to Agarwalla, is focusing too much on AI itself rather than on the customer's problem.

"Nobody cares that it's AI. Customer doesn't care it's AI," he adds. "A startup is still about solving a problem and solving it much, much better than it's ever been done before."

AI Funding Follows Specialisation

Despite the growing number of AI startups, funding continues to flow disproportionately towards a relatively small group of companies. As per data provided by Tracxn, while the overall funding surged between Q4 2025 and Q1 2026, the number of rounds fell from 28 to 14, with larger cheque sizes dominating the narrative.

Agarwalla argues this reflects the stage of development the industry is currently experiencing rather than a structural weakness in the ecosystem. He firmly believes that "we are still in the very beginning of the first innings."

Much of today's investment is directed towards foundational layers of the AI stack, including compute infrastructure, chips, and models. He compares the phenomenon to the early stages of economic development, when capital is first deployed into roads, ports, and other enabling infrastructure before broader commercial activity emerges.

"Today, all the money is going towards tokens," he adds. However, he expects the market to evolve as enterprises seek practical applications that generate measurable business value.

Investor behaviour is also influenced by familiarity. Many traditional venture firms have only recently begun investing in AI startups and are still developing frameworks to evaluate them effectively.

Policy Support and India's Competitive Position

India’s AI ecosystem has comprehensive government backing, with the IndiaAI Mission democratising access to GPUs and the AIKosh dataset platform boosting model accuracy. These initiatives are also fuelling investor confidence, although Agarwalla believes they do not directly determine investment decisions.

Instead, policy support creates a favourable environment for capital deployment by signalling long-term commitment to emerging technologies.

"The government encouraging these investments and having a positive policy momentum helps startups gain investor confidence," he says.

When discussing India's competitiveness, Agarwalla rejects the notion that talent is a constraint. "We have the talent pool. Both the density and quality of talent are great in this country," he stresses.

The challenge, he argues, is ensuring startups build with global markets in mind from the outset. While India's domestic enterprise market can help founders develop core products and gain primary customers, they need to remain globally competitive in order to scale.

Next Opportunities

Looking ahead, Agarwalla believes some of the most promising AI opportunities remain underexplored.

Consumer AI products built specifically for Indian users have yet to attract significant attention. More importantly, industries with historically low levels of digitisation, such as manufacturing, construction, and agriculture, could become major beneficiaries of AI adoption.

These sectors often rely heavily on manual processes and have struggled to adopt traditional enterprise software. AI, he argues, could allow them to leapfrog directly into more advanced digital workflows.

"This is much like India jumping from not having landlines to everybody having a smartphone in their hand," he remarks.

SenseAI’s investment thesis remains clear. The greatest opportunities are not necessarily in the companies producing intelligence, but in those applying it to solve meaningful problems.

"We only invest in the application layer," Agarwalla remarks. "That's where the rubber meets the road. That's where the value is created."