How to Fail with Artificial Intelligence: 9 creative ways to make your AI startup fail

This post summarizes nine creative ways to condemn almost any AI startup to bankruptcy. I focus on data science and machine learning startups, but the lessons on what to avoid can easily be applied to other industries.

By Francesco Gadaleta, Abe.

Courtesy of

Most entrepreneurs and data scientists will tell you how to succeed with artificial intelligence (AI) and machine learning, but very few talk about how to fail with such technologies. While the technology is incredibly powerful — and heavily hyped — there is a plethora of ways to fail with AI.

This post summarizes nine creative ways to condemn almost any AI startup to bankruptcy. I focus on data science and machine learning startups, but the lessons on what to avoid can easily be applied to other industries.

#1 Cut R&D spending to save money

Artificial intelligence requires heavy investment in cutting-edge research, experimentation, advanced algorithms, and computing infrastructure. Any AI startup hoping to develop useful AI technologies will have to spend a great deal of money on research and development (R&D) in these areas.

As tempting as it may be to downsize spending in this area, cutting R&D costs will quickly pave the way to failure.

#2 Operate in a technology bubble

Technology is very much bound to the social circumstances in which it is created. Technology is never self-sustaining.

Artificial intelligence has failed several times throughout the history of computer science, not only due to technological issues — that could be and were solved — but also because of a lack of social need and interest at the time.

The lesson to be learned is that artificial intelligence technologies cannot be built in isolation from the social circumstances that make them necessary (like health care, disease research, and finance).

A new technology is only successful when its proponents generate interest in it and obtain resources from the same people who will become their end users or customers. Before engineering the technology itself, visionaries and entrepreneurs should first engineer people, persuading them to suspend their doubts and embrace the novelty and utility of disruptive ideas.

Operating in a bubble and ignoring the current needs of society is a sure path to failure.

#3 Prioritize technology over business strategy

Technology alone is not enough to achieve success, no matter how powerful or transformative it may be. At the end of the day, a tech startup is still a business and needs a solid business strategy to succeed.

Any startup that lacks a strategy for identifying target markets, generating sales, and effectively allocating and spending resources, but gives priority exclusively to their technological assets, is doomed to fail (and quickly).

#4 Work without a clear vision

For any tech company to succeed, it’s crucial to quickly establish a clear vision. This is especially important for AI companies since the technology has applications for many diverse industries (from finance to health care).

This vision should be communicated early and often to employees so that everyone is on the same page about the company mission and roadmap. In addition to this long-term vision, having clear short-term goals and objectives is also essential.

Fragmented and heterogeneous goals are almost always a precursor to failure.

#5 Develop without addressing business needs

AI companies are also software companies. And software that gets written for its own sake, without satisfying any business need, won’t sell.

Useless software, no matter how objectively cheap to develop, is ultimately very expensive to build and maintain because there’s no pay-off. If it provides no value for potential customers, it’s a waste of time and resources to develop and ultimately adds no value to your company.

Developing AI for the sake of AI is a good approach… to failure.

#6 Cultivate a “we’re the best” attitude

While self-confidence is essential for maintaining high development standards, if left unchecked, it could lead to problems down the line. Organizations with over-inflated self-confidence have a tendency to fall into one of the biggest traps of product design: arrogance.

This attitude is common in startups. Assuming that their own products are the best, such organizations tend to develop everything in-house, even tasks that could easily be outsourced. Not only does this waste valuable time and resources, it also prevents teams from focusing on their core business.

Plowing forward with an over-confident attitude will put you on the fast track to failure.

#7 Get caught in a never-ending development loop

This is specific to software companies (and definitely AI companies).

In software design, it’s important to develop fast and release faster, even if the product is imperfect. Why? Because software that’s built behind closed doors doesn’t do any good; it needs to be out in the real world, exposed to real users with real problems. By releasing early and often, you can gather useful data, learn what works and what doesn’t, and iterate to produce the best possible version of your product.

The benefits of moving to market quickly outweigh the drawbacks. Taking the risk of disclosing features to competitors is the price to pay for getting immediate (and frequent) feedback from users.

Staying in a never-ending design-develop-design loop, leads to missed opportunities and messages from the market. This is a sure way to fail.

#8 Assume your customers are like developers

To users, software is an experience. Developers, on the other hand, see software as a tool. Product design should be focused around customers, not developers (unless developers are the customers).

The misconception that customers are just like developers is typical of startup culture and toxic for tech companies.

Developers tend to focus on the technical aspects of a product, prioritizing function over form. Unfortunately, they often ignore design and the user experience. For any AI product to be successful and widely adopted, it must be both functional and well-designed for ease of use.

Developing an AI product that satisfies only its developers’ vision and needs leads to unmet requirements in the real market. In a word, failure.

#9 Assume the AI hype is enough to succeed

This is not the first time that AI is getting a lot of hype. The first time was in the ’80s, when professors claimed that computers would soon replace humans (which obviously didn’t happen). But now AI is being heavily hyped again, and this time appears to be different. Hype that lasts five years without any sign of slowing down is not just hype. It’s a steady trend.

However, hype alone is not enough for a technology to succeed. AI developers have to put their money where their mouth is and build products that actually live up to the hype.

Assuming that an AI startup will succeed simply because of the hype surrounding their core business is the first step towards failure.


To build a successful AI startup, you first have to know how to fail. That means confronting and correcting your erroneous assumptions, mismatched priorities, and unfocused strategy.

Anticipate and avoid the paths to failure to give your company the best chance to succeed with AI.

Bio: Francesco Gadaleta (@worldofpiggy) writes about machine learning and deep learning, and is the CDO at Abe.

Original. Reposted with permission.