Good News For Startups: Enterprise Is Bad At AI (AI column)

In series podcast guys from Y Combinator discussed the studyThe GenAI Divide: State of AI in Business 2025" (see. my analysis)The results of which indicate that 95Percent of corporate AI projects fail. The discussion is based on the reasons for these results and why it is in the hands of AI startups, or more precisely, the guys discuss issues.

• What's really behind the viral statistics about95Percentage of failures of AI projects? Why do internal IT projects, even in large companies, often stall? What makes it difficult to implement complex AI systems in a corporate environment? How do small AI startups win competition from banks and consulting? Why are established companies unable to create a working AI product and what to do about it?

The opinions of the managing partners of Y-Combinator are as follows:

• Throw the phrase "95“Percent Failure” is distorted: it refers to internal decisions in corporations that do not know how to learn from feedback and adapt. In other words, it’s not that AI doesn’t work, it’s that big companies don’t know how to build it. Here, the authors kick Apple for their calendar app, and if Apple makes a crappy decision, then ordinary companies will even have problems. Most corporate software is very mediocre, and attracting expensive consultants (E&Y, Deloitte) It often creates two problems instead of one. In large companies, there are organizational difficulties: an AI project needs to take into account the interests of many teams, overcome political barriers and coordinate requirements between departments. Add legacy legacies and committee design – implementing AI becomes a long and difficult process AI startups win where corporations pass: small teams quickly build solutions for specific tasks, immediately with AI capabilities laid down. (AI-native) And quickly bring the product to the result. In a matter of months, Tactile made AI decision-making systems for banks, while Citi and JP Morgan spent years and millions on non-working analogues. Now is the perfect moment for AI startups: large firms are eager to implement AI and willing to take risks with young teams. Fear of falling behind forces corporations to make decisions faster and buy a ready-made AI solution rather than trying to make their own. The guys from the startup incubator praised the opportunities for startups in every process: the demand for AI is so ahead of the internal capabilities of companies that in every niche there is space for an external solution. If we go back to hypotheses that could explain the failures of implementation, podcasters singled out the human factor: many corporate engineers do not believe in AI, do not use new tools and are even happy to refer to studies like MIT that confirm their skepticism. This culture leads to the failure of projects. (self-fulfilling) and missed opportunities. As a result, businesses where internal teams have given up are ready to call on startups to help.

Anyway, 95Failure to implement AI in corporations is not a reason not to try. On the contrary, you have every opportunity to get into 5% successful. However, partners of one of the largest startup incubators say that large companies are likely to increasingly turn to startups and buy ready-made AI solutions, as internal teams fail:) But I would discount conflict of interest and treat it as an opportunity not only for startups, but also for corporations to increase their internal capabilities to use AI.

#AI #Engineering #Software #Leadership #ML