I showed a business owner how to automate 2 hours of WhatsApp work per day. His response: 'If I automate it, what will that guy do for the rest of his time? I'm paying him a full salary.'

Most “user research” in Indian startups are just people in AC rooms imagining what Bharat wants.

Your AI is too nice for the Indian ecosystem I've lately noticed intense rivalry among AI solutions developed for the Indian market. We have too many founders building too similar products, or even the exact same products - over and over again. Take voice agents; there are 10-15 companies building these for the same industries, targeting the same use cases, and pitching to the same clients. This has created intense competition in a very small market. It's a paradise for potential buyers who get multiple POCs done often without responding because we are not solving a critical need. They have numerous choices, giving them tremendous negotiating power. If you offer a voice call for 10 rupees, they'll ask for 5; if you agree to 5, they'll push for 2; at 2 rupees, they'll demand 1; and at 1 rupee, they'll try to get it for 50 paise. They're constantly undercutting prices. What works in B2B in India is either having strong connections that lead to implementation, or offering services at a much cheaper price-sometimes at cost or even below cost-hoping it's a customer acquisition tactic that will lead to long-term retention. Unlike SaaS products, AI use cases are often complex, requiring workflow mechanisms and drag-and-drop functionality. AI currently needs to be treated as a services layer, not a SaaS layer. When you position yourself as "build your own prompts" or "drag-and-drop builder," making it work properly is challenging, and you end up functioning as a services firm by orchestrating APIs. This makes clients wonder why they should pay you so much when they could potentially build it internally. They might lack the engineering bandwidth or conviction initially, so they'll take your POC and eventually build their own solution. I recently visited a BPO to pitch some products and observed two types of agents. One group was handling loan collections, with agents aggressively confronting borrowers: "You took a loan. Why aren't you repaying? Don't you have any shame?" In another section, agents were selling insurance, sometimes flirting with potential customers to persuade them to buy policies. Though I don't endorse any of those tactics. The AI we're building is too nice for the Indian economy. You need the right language, dialect, tone, and tonality. When we promote AI as a replacement for call agents, it's somewhat misguided because we lack the appropriate training data in the Indian context. Yet we pitch "AI call centers," which isn't realistic. Customers recognize this inefficiency and view AI as a nice-to-have rather than a must-have, which is why they try to undercut costs. You only gain leverage when you're building something essential. The problem is that many of us consider ourselves tech entrepreneurs and avoid the dirty work of research, data collection, and training. Until that changes, we're not solving real problems. We blame Indian customers for not paying enough rather than tackling the fundamental issues. (My observations)

Wisdom has no shortcuts!