In recent years, AI has become the ubiquitous topic of conversation. Some fear it will take their jobs, others share tips on how to use it, and some simply enjoy the novelty of chatting with it.

A similar pattern is emerging in the corporate world. Driven by FOMO (Fear of Missing Out) or what is often called “Resume Driven” development, many firms have greenlit projects primarily because they look prestigious on a stakeholder’s CV or make for a flashy presentation at industry conferences. Often, these projects move forward without solid financial analysis or a clear understanding of operational costs. More importantly, many are deployed without a concrete plan for how they will actually deliver value on the ground. By the end of 2025, research revealed that nearly 95% of AI projects failed to deliver a positive ROI, or the gains were negligible compared to the project costs.

Reflecting on my observations as both an IT executive and an end-user, here are a few “inefficient” scenarios from the field:

• The Over-Engineered Chatbot: Seeing massive investments in chatbots that fail to perform as imagined. For instance, telling a grocery app “my active order” when you only have one, yet being asked “Which order?” by a bot that simply regurgitates info already visible on your screen. (As an end-user, I’ll admit I’m quite frustrated with this.)
• The Cost-Benefit Paradox: Automating a task that costs 1 unit of human labor with an AI solution that costs 5 units to maintain. (e.g., spending $10,000/month in token costs to save $2,000 in labor.)
• The “Dead” Investment: Deploying solutions without consulting business units or understanding customer needs, leading to sophisticated tools that find no resonance in the real world.

In an era where every company is racing for efficiency, how do we turn the right technology into the right product?

A Roadmap for an Efficient AI Strategy

• Identify Bottlenecks: Start by pinpointing actual needs and challenges on the ground. Focus on the company’s main bottlenecks and manual, low-value tasks. Analyze whether AI can truly make these processes more efficient.
• Scrutinize References: When evaluating which technologies to use, prioritize suppliers with proven reference projects in similar industries or processes.
• The Power of POC: Every product and process should be supported by a Proof of Concept (POC) involving the actual process owners. There must be a mutual agreement between IT and business units on the final product and its specific use cases.
• Financial Reality Check: Compare the “Total Cost of Ownership” against the projected gains. If the investment isn’t financially rational, don’t move forward.

While these steps apply to all IT projects, they are worth reiterating now, given the recent wave of inefficient or failed AI experiments.

There is also the common anxiety among my peers: “What if I start a project today and a superior technology emerges tomorrow, leaving me behind?” This concern is healthy and normal, but remember that waiting for the “perfect” moment often results in doing nothing at all. Doing our best with current technology is always better than paralysis. Once the right processes are established, the underlying tech can be swapped or upgraded as it evolves. What matters is that the foundation is solid.