From dot-com to dot-AE: How can we learn from final technical change (and avoid making similar mistakes)

At the height of the dot-com boom, adding “.com” in the name of the company, was enough to increase its stock price-there was no real customer in the business, the path of revenue or profitability. Today, history is repeating itself. Swap “.com” for “AI”, and the story looks very familiar.

Companies are running to sprinkle “AI” in their pitch decks, product descriptions and domain names, hoping to ride a promotion. As reported Domain name stateRegistration for “.AI” domains increased by about 77.1% year-on-year in 2024, running by startups and incumbents to connect themselves with equally artificial intelligence-whether they have a true AI benefit.

One thing became clear in the late 1990s: using success technology is not enough. Companies who escaped from the dot-com crash were not chasing the publicity-they were solving real problems and scaling with purpose.

AI is not different. This will reopen the industries, but the winner will not be going to slap “AI” on a landing page – they will focus on what more cutters through promotion.

First step? Make small start, find your veg and scale deliberately.

Start small: Find your wedge before scale

One of the most expensive mistakes of the Dot-Com-Com-era was trying to grow up very soon-a lesson AI product builders cannot take the risk of ignoring today.

For example, take eBay. It began as a simple online auction site for collection – begin with something as a niche as a pez dispenser. Early users loved it because it solves a very specific problem: it was associated with hobbys that could not find each other offline. Only after dominating the fact that the initial vertical expanded electronics, fashion and, finally, almost anything you can buy today, such as expanded into broader categories.

Compare it MalevolentAnother dot-com era startup with a very different strategy. Webvan aims to revolutionize grocery purchases with all ordering and rapid home delivery together, all together, all together, together, together. It spent hundreds of crores of dollars constructed on a large scale before the construction of warehouses and complex distribution fleet, before it was a strong customer demand. When development did not take place very fast, the company fell under its weight.

The pattern is clear: Start with a sharp, specific user requirement. Pay attention to a narrow wedge that you may dominate. Extend only when you have proof of strong demand.

For AI product builders, it means resisting the urge to make “AI that does everything”. For example, take a generic AI tool for data analysis. Are you targeting product managers, designers or data scientists? Are you building for those who do not know SQL, with limited experiences or experienced analysts?

Each of those users has very different requirements, workflows and expectations. Like a narrow, well-defined kohort, like a technical project managers (PMS), who requires quick insight to direct product decisions-allow you to understand your user deeply, correct the experience and make something really unavoidable. From there, you can deliberately expand adjacent personality or abilities. In the race for the construction of permanent general AI products, the winners will not be those who try to serve everyone at once – they will be the ones that start small, and serve someone incredibly well.

Duck of your data: Make compounding defensibility early

Starting small helps you find a product-market fit. But once you get traction, your next priority is to create defensiveness – and in the world of General AI, it means that your data is owner.

Companies avoiding the dot-com boom not only could not catch users-they captured the ownership data. For example, Amazon did not stop selling books. They tracked the purchase and product ideas to improve recommendations, then used regional order data to customize the supply. By analyzing the pattern of purchasing in cities and zip codes, he demanded demand, stocking the warehouses smart and well-organized shipping routes-a major profit could not match the competitive match, laying the foundation for a two-day distribution of primal. None of these data is possible without strategy, which is cooked in the product from day one.

Google followed a similar path. Each querry, click and improvement discovery became training data to improve the results – and later, advertisement. They did not build just a search engine; He created a real -time feedback loop, which continuously learned from users, creating a trench that created their results and targeted hard to lose.

The lesson for General AI product builders is clear: Long-term benefits will not only come from access to a powerful model-this will come from the manufacture of proprietary data loops that improve their product over time.

Today, any person with adequate resources can fix an open-source large language model (LLM) or pay to reach API. What is very difficult-and far more valuable-high-component, gathering real-world user interaction data that compounds over time.

If you are building a general AI product, you need to ask important questions:

• What unique data will we capture as interaction with users?
• How can we design feedback loops that continuously refine the product?
• What is the domain-specific data that we can collect (moral and safe) that the contestants will not have?

For example, take Duolingo. With GPT-4, they have gone beyond Original personalizationFeatures like “Explain my answer” and AI roll-play create rich user interactions-only capturing the answers, but how the learners think and understand. Duoolingo combines this data with its own AI to refine the experience, a profitable competitors cannot easily match.

In the General AI era, data should be your compounding benefits. Companies that designing their products to catch and learn from ownership data will survive and lead.

Conclusion: This is a marathon, not sprint

The Dot-Com-com era showed us that the promotion fades rapidly, but tolerates basic things. General Ae Boom is no different. Companies that thrive will not be the pursuit of headlines – they will solve real problems, score with discipline and make real consumption.

The future of AI will be related to the builders who understand that it is a marathon – and patience to run it.

Caliang Fu is an AI product manager in Uber.

Daily insights on business use cases with VB daily

If you want to impress your boss, VB daily has covered you. We give you the scoop inside what companies are doing with generative AI, from regulatory changes to practical deployment, so you can share insight for maximum ROI.

Read our privacy policy

Thanks for membership. See more VB newsletters here.

There was an error.