The single, most common reason why most AI projects fail is not technical.

Having spent almost 15 years in the AI & data services space, I can confidently say that the primary cause of failure for AI initiatives happens at their inception: the misalignment between technological capabilities and genuine business needs.

Contrary to what many frustrated decision-makers might believe, it's not a lack of data or computing power. It's not that Generative AI (GenAI) is expensive at scale. It's not even a shortage of skilled talent, budget overruns, or poor execution. This silent threat emerges before a single line of code is written.

Unveiling the silent threat

Unlike a technical error that might throw an exception or a budget overrun that shows up in financial reports, misalignment doesn't announce itself. It frequently goes unnoticed until much later in the project lifecycle, and its effects are often attributed to other causes.

This problem has reached epidemic proportions in the current GenAI gold rush. Goldman Sachs recently published a report titled Gen AI: Too Much Spend, Too Little Benefit? in which they argue that even with trillions of dollars being spent on GenAI, the payoff will come much slower than previously thought.

Those who know me have probably heard me emphasize this over the past year, but it is important to understand that this issue is not a fault of the technology. The fuel for this threat lies with executives and decision-makers.

If we consider the report published by McKinsey in June 2023, which estimated that GenAI could add between $2.6 and $4.4 trillion annually to the global economy across various use cases, it becomes evident that many companies, driven by FOMO, are implementing AI solutions without a clear understanding of how these technologies will benefit their business.

Eager to jump on the bandwagon, leaders often skip crucial planning steps. Instead of asking, "What business problems do we need to solve? How could AI help?" they are asking, "How can we use ChatGPT?" It is as if "AI" has become a magic word, expected to solve all problems overnight. This approach undoubtedly leads to disaster.

How to ensure your AI project is set up for success

Picture a Venn diagram: on one side, your business needs; on the other the expanding capabilities of AI. The sweet spot for AI implementation is where these circles overlap. As AI technology evolves, this intersection shifts. Stay informed about AI advancements, but keep a laser focus on your business objectives. That's where the magic happens.

Define your problem

One crucial distinction often overlooked is whether the problem you're trying to solve with AI is technical or adaptive. In their article "The Work of Leadership," published in the Harvard Business Review, Heifetz and Laurie explore this distinction in depth and discuss its implications for effective leadership.

Technical problems

Technical problems are well-defined, easy to identify, and typically have clear solutions that can be addressed with knowledge, authority, or subject-matter expertise.

For instance, tasks like “improving image recognition accuracy in low-light conditions” or “adapting an algorithm to run on a specific embedded device” are technical challenges. In such cases, AI and the surrounding engineering efforts can often provide direct and effective solutions. Technical solutions often face very little resistance from stakeholders in organizations.

Adaptive problems

Adaptive problems are often harder to identify because they are complex and involve changes in beliefs, attitudes, and behaviors. Examples include “reducing customer churn,” “training the workforce to understand AI,” or “becoming more data-driven in decision-making.”

For adaptive challenges, AI might be part of the solution, but it is rarely the complete answer. Recognizing this distinction is crucial for setting realistic expectations and selecting the appropriate approach.

Some key points to combat the silent threat to AI initiatives:

1. Misalignment with business strategy or unclear problem definition

Projects kick off without a clear understanding of their motivation, the problem being solved, and its significance to the business. AI initiatives must support broader business goals; otherwise, they become isolated experiments with little impact.

2. Failure to identify impact metrics

Every production-grade AI project should have a way to quantify its impact on the business. Note that "production-grade" is emphasized: research or exploratory projects are fine without a metric, but you should never expect these to reach production. The output of a research or exploratory project will either be the knowledge that something doesn't work or a proof of concept demonstrating the feasibility of turning it into a production-grade project with a quantifiable impact on the business.

3. Impact metrics but no Return on Investment (ROI)

Sometimes, the cost of AI implementation can be too high relative to its impact. This may change in the future as the development of specialized products simplifies the implementation of certain AI projects. Moreover, as technology improves, the likelihood of achieving a positive ROI is increased.

4. Lack of stakeholder buy-in

If key decision-makers like the CEO/CTO/CIO or even end-users do not understand or support the project's objectives, it is doomed from the start. Ensuring that these stakeholders are on board is crucial for the success of AI initiatives. Their buy-in helps align the project with business goals, secures necessary resources, and fosters a supportive environment for implementation.

5. Overemphasis on technology

Many projects focus excessively on the AI technology itself rather than the business value it can deliver. The technology is a means to an end, not the end itself—a common bias among technical teams. For your business, AI should be a feature, not the product.

Final thoughts

At Tryolabs, we've developed a multi-stage framework to help organizations navigate AI implementation and ensure alignment. This framework is essential for setting up AI initiatives for success and keeping stakeholder expectations realistic.

Regardless of the specific methodology, it’s important to step back, clearly define your objectives, and ensure your AI initiatives align with your business needs and value propositions, this is crucial for your business success.

If you embark on an exploratory phase, be prepared for the possibility of it not working out. While it may not be as flashy as following the latest AI trend, this approach is the surest path to long-term success.

In the rush to adopt AI, don't let FOMO drive your strategy. The silent threat will have nowhere to hide.