He believes that most business people do not understand what they can actually achieve with these technologies.
– We're not talking about 5 percent, 10 percent. or even 15 percent efficiency increase. We're talking about exponential gains over time – says Bennett in response to Business Insider Polska's questions about the existence of the AI bubble.
In his opinion, understanding the opportunities offered by AI is the fundamental difference between a reasonable approach to technology and an empty desire to join the trend. The real paradox is that while the capital ecosystem believes in the enormous transformative power of artificial intelligence, most implementations in real companies end in failure.
Read also in BUSINESS INSIDER
Start-up dreams are indeed big, but Bennett suggests that the problem is not too high expectations – more in that they are of the wrong kind.
The rest of the article below the video:
AI efficiency: who actually makes money?
According to an MIT study 95 percent AI pilot programs in enterprises do not achieve their intended goals, achieving revenues of only 5%. start-ups. These are data that can shock any rational investor.
Bennett doesn't deny the reality of the problem. In his understanding, the point is that companies draw too far-reaching conclusions too quickly. – You need to consider where you are in the cycle – he emphasizes.
AI bubble? Bennett: Nobody disputes the impact
When asked directly about the artificial intelligence bubble, Bennett replies: – When people talk about bubbles, they usually mean a situation in which the scale of investments significantly exceeds the return on these investments. I think it's really hard to say that the impact of generative AI won't have an extremely large impact on every aspect of our lives.
He estimated that the discussion about the bubble usually focuses on how much capital was invested in companies. Bennett proposes to shift the emphasis to whether this capital is actually reflected in real products and services.
Here, in his opinion, there is an important signal contradicting the bubble thesis: GitHub Copilot, an AI-based coding tool created by Microsoft, has been implemented in less than 18 months and already shows that developers can work at more than half the speed. Lovable, an AI-powered UI development application, offers similarly significant reductions in time to implementation.
– If we are really talking about profits of 30-45%. in customer service and 20-45 percent in software engineering, the question is: is it a bubble or a revolution? asks Bennett.
As the next one an example comes from Lila Scienceswhich attempts to “apply the agent-based scientific method” to health care. “If their thesis is confirmed, the impact on the entire medical ecosystem will be enormous – from research to the average patient's experience,” argues Bennett, adding that the AI-based applications, products and services visible today are only “scratching the surface.”
Where real business intelligence comes into play
However, Bennett states that most AI startups don't ask themselves the right questions. They focus on models, algorithms, GPU, data architecture. Indeed, they should focus on something completely different: how to build a profitable and sustainable business in the long term. It's not that spectacular, but it's effective, he argues.
“For a lot of the startups that I would say are scaling successfully and doing it at a really high level, the focus is more on how to build long-term, healthy, vibrant companies that can stay consistent with their mission and culture,” says Bennett.
Bennett suggests that actual winning startups are thinking about something entirely different: building an organizational culture that can survive scaling. Instead of asking, “Is this AI?”, they ask, “Is this value to my customers that I can deliver consistently?“.
Capital restrictions
Bennett emphasizes that AWS boasts 300,000 start-ups they support, operating on the company's platforms and tools. But there is also another important piece of information behind it: Support does not mean equal access to successu. AWS poured over $7 billion. in loans and investments for start-ups, but suggests that the real value lies in human resources – teams of founders, CTOs, people with business experience who can guide young companies through the dangerous first years.
This also explains why Silicon Valley and Seattle still dominate, even though Bennett talks about “democratizing” product creation. Access to capital and knowledge is now more globally dispersed, but access to some type of consulting is — business intelligence, not just technical intelligence — remains uneven.
Agents change the rules of the game
Bennett's final point is perhaps the most economically relevant. He thinks it's agent stuff AI solutions are changing the financing structure of start-ups. Some companies can now work with smaller seed rounds because they can do much more with less. Others may take longer to scale because their teams are more productive with AI tools.
What started out as simple – everyone wants to be an AI – is turning into a more complex picture. New economies are emerging. AI startups (robotics, autonomous systems) have a completely different time-to-revenue profile than agent companies. Their capital implications are very different, emphasizes Bennett.
When 95 percent AI pilots fail, it is difficult to talk about a spectacular success. But when the 5% that succeed generate profits that can amount to tens of percent, the bubble discussion begins to seem peripheral.
Grzegorz Kowalczyk, journalist of Business Insider Polska
