3 Reasons Why Businesses Should NOT Use AI

Things SMBs should consider before making an AI investment

5 min readJun 21, 2024

There are many good reasons for businesses to start using AI. It’s a power technology, customers are demanding it, and competitors are using it. However, these points tend to overshadow the other side of the coin — why businesses should not use AI.

Over the past year, I’ve consulted 65 small to medium-sized businesses (SMBs) on how they can leverage AI. Through these conversations, I’ve seen a few key problems that hold companies back from getting a return on their AI investment. To (hopefully) help founders and operators avoid these problems, here I discuss 3 reasons why SMBs should NOT use AI.

Reason 1: Lack of AI Infrastructure

The most common limiting factor for businesses is lack of AI infrastructure. More specifically, the lack of two main things: data infrastructure and technical talent.

Data infrastructure makes it easy to access and use data relevant to your business. This is important because it significantly reduces the barrier to entry (i.e., lowers cost) for most AI projects. Common examples of this are building a RAG system or fine-tuning a model for a specific business context.

Additionally, an existing data infrastructure provides several opportunities for value creation. Namely, AI can enhance current data workflows through new ways of capturing and structuring data. For example, capturing unstructured data from customer reviews, company documents, and public webpages to augment existing datasets and decision-making.

The other side of AI infrastructure is technical talent. Although AI is easier now than ever, building custom projects is still technically demanding. This requires having the right people to support its development in your business.

Generally speaking, there are two options for this.

Option 1: Internal AI team — hiring and onboarding technical contributors. Pros: cheapest option in the long run, and the AI team can better integrate with other parts of the business. Cons: teams take time to develop, thus this is a long-term investment.
Option 2: External AI team — hiring a vendor. Pros: team can get started quickly with low initial investment. Cons: expensive in the long run and risks vendor lock-in.

Reason 2: Unnecessary Costs and Complexity

The second thing to consider is that AI often introduces unnecessary costs and complexity. In business, value is generated through solving problems. The trouble with using AI to solve a particular problem is that it often comes with significant overhead (i.e. data, talent, compute).

However, even if the data infrastructure and teams are in place, some AI solutions might be overkill. For instance, one of the most common projects clients approach me with is building a custom chatbot that can retrieve information from a knowledge base.

While this is a new and powerful way to solve the information search problem, it is significantly more complex to engineer than a simple search interface. And the marginal value generated may not justify this additional complexity.

Reason 3: AI Risk

The final reason to consider is AI risk, which is something that can easily be eclipsed by all the AI excitement these days.

When people talk about AI risk, they most often discuss (i.e., debate) existential risks—that is, AI risks that can harm humanity on a global scale. However, there are many clear and present AI risks relevant to businesses.

Risk is often a statement of ignorance, and there is still a lot we don’t understand about how AI systems like ChatGPT work under the hood. While this gives rise to many potential risks, here I am going to focus on two: lack of explainability and hallucinations.

More powerful AI systems usually come at the cost of less explainability, i.e., we don’t know why a system generated a particular output. While this might be no big deal if trying to classify cat memes, it can pose significant risks when system outputs influence business decision-making and customer experience.

The other risk is hallucinations. This is when generative AI systems create false outputs that are potentially harmful. A recent example of this was an Air Canada chatbot that provided false information about the company’s bereavement policy, leading the customer to sue and win $812.02 [1]. While this is a negligible amount for Air Canada, it’s not clear what the impact of future hallucinations might be, i.e., the cost and scale of damages.

Recommendations for SMBs

I don’t want to give the impression that businesses should give up on AI because of all these potential challenges. In fact, I believe the opportunity for SMBs is even greater than for larger companies because they are still flexible enough to explore new ways of doing business that can help them replace the incumbents.

AI is shuffling the deck, which means the “little guys” have the most to gain from it. Toward that end, I want to share 3 recommendations for how SMBs can approach using AI.

Rec 1: Start with the Problem (not technology)

There’s a quirk of human psychology called the “Hammer Problem,” which says that when you have a really nice hammer (like AI), everything looks like a nail. This leads us to go around and wack every problem with it. While you might hit a few nails, you might break a few things, too.

That’s why it's important to think critically about which problems you apply AI to. Ask yourself, why do I want to use AI to solve this? Is there a simpler solution?

Rec 2: Start with ChatGPT (and friends)

Often, much of the value of AI use cases can be realized completely for free using ChatGPT and similar tools. This is a great low-stakes way to experiment with and validate use case ideas.

In other words, if you can’t realize the value of using ChatGPT, then it may be a red flag for the feasibility of a more sophisticated solution. However, if ChatGPT does provide a good solution to the problem, that’s a good indicator of a custom AI solution’s potential.

Rec 3: Keep it Low Stakes

Beyond using ChatGPT to experiment with ideas for free, it’s important to limit the costs and risks of new AI applications. This can be achieved in many ways. A few suggestions are given below.

Validate ideas with quick and low-cost POCs
Keep a human in the loop of the AI system
Apply it to areas where the cost of being wrong is not much greater than the benefits of being right.

Conclusion

Although AI presents tremendous business potential, weighing its potential upside and downside is important. Here, I shared 3 things SMBs should consider before making AI bets, along with 3 recommendations.

I’m always curious about how businesses are using AI to create value, so if you’ve found a successful application in your business, please consider sharing it in the comments 😁