Стартапы используют ИИ как «сооснователя», но опасаются зависимости и роста затрат

It was much like a job aspirant using ChatGPT in the application to make it to the interview stage, but then stumbling.   Take the example of a marketing graduate from Rajasthan who was there to pitch the app he had built through AI. The app tracks an employee’s sales targets and recommends better job opportunities to her once she has registered on it, thereby saving her the trouble of actively looking for a job switch.   “I am a product guy,” said the founder who did not wish to be named. “The idea was mine but executed mostly by AI.” He used the AI tools during their trial stages, so the cost was close to nil. Going ahead, he reckoned he might have to use paid versions and, perhaps, even employ someone with deeper tech knowhow. “The cost might bite for some time,” he said.  The startups using AI as a “cofounder” are varied. From a reel-like app for reading books or an online relationship therapist to a software that builds stronger parental control on the internet, the range is vast. What’s common, though, is the “.AI” in the domain name.    “The dominant view is that putting AI at the core of your product can boost its valuation by 2x-3x,” said Juhi Bhatnagar, a Gurugram-based angel investor and an AI expert. However, that is not how the market works, she added. As an investor, she said, it did not matter how and where the AI technology was deployed as long as it was profitable, and the founders knew where they were heading with it. “Every use should be justifiable,” she added, which is not always the case.   According to the latest Global Startup Ecosystem Report (GSER) 2026, released in June, Bengaluru ranks second in Asia for AI innovation (Beijing is the first). Since the 2021 report, the city’s AI ecosystem has recorded a 190 per cent growth in value, outpacing the global average of 149 per cent and Asia average of 66 per cent.   “Every layer of the startup will have it (AI), but the challenge is that we have not questioned the technology enough,” said Aaqib Jameel, senior manager, Incubation & Investments at IKP EDEN, a Bengaluru-based startup incubator. “We are advancing, but are still less paranoid than those in the West.”   A number of startup founders at the meetup were fresh graduates from some of India’s leading technology institutions. Many of these institutes have started to discourage students from using AI from the beginning to the end to build prototypes. According to a PhD scholar at the Indian Institute of Technology Delhi, whose research is focused on AI, the initial samples created by students are sometimes full of bugs. “The concern is that they don’t know how to fix it because AI has done most of the job,” he said.   Santosh Patra, founder of Bengaluru-based AI chatbot solution Bolofy, said overreliance on artificial intelligence can pose serious problems. “We are an AI product, so yes, that is our core. But sometimes, the backend team uses it so much that my app starts to show glitches,” he said. He has, therefore, had to reassess his cost strategy.  According to Jameel, managing technology cost is an important differentiator for startups at a later stage. “At that stage, AI can only be an enabler, not a partner,” he said.  There are already concerns over businesses burning their AI budgets. Uber, for instance, reportedly used its entire AI budget for 2026 in the first four months alone. “Unlike larger corporations that can splash money on AI, startups are more cautious about where and why they’re using it,” said Iesh Dixit, CEO and founder of Powerplay, a Bengaluru-based construction management software-as-a-service (SaaS) startup. Even though AI’s cost efficiency hasn’t been fully mapped, the output gains are visible, he added.  With AI emerging as the most efficient enabler for startups, the role of founders has become more important than ever. “AI is arguably the most capable execution layer we’ve ever had. It can help build, test, and analyse at a speed that wasn’t possible before,” said Dixit. “But the answer to why we are building this still comes from a human brain, and until AI creates a genuine intent of its own, it can only be a tool and not a partner,” he added.   Scaling, said Jameel, becomes much easier if founders understand the product’s architecture. “You don’t know coding, that’s okay. AI can do it for you. But the logic has to be understood,” he added.