The Unexpected Power Of Failed AI Projects, And How They Help Diagnose Operational Weaknesses

In the rush to embrace AI, the most valuable lessons might come from its failures. While most leaders dread a failed AI project as a costly setback, this viewpoint overlooks a crucial insight. What if a stalled AI initiative isn't a problem to be fixed, but a precision diagnostic, designed to expose the foundational weaknesses in your business that no other tool could reveal?

That's the mindset Gabriel Lozano brings to his team. A data-driven marketing professional with over 15 years of experience in digital strategy, CRM, and automation, he currently serves as the CRM Lead at Athena Home Loans. His time as an AI Business Solutions Manager at e-commerce giant Temple & Webster gave him firsthand insight into the impact of AI projects and the lessons experimentation can bring.

Lozano sees a common trend where companies are rushing to join the AI bandwagon without first doing the unglamorous work of re-engineering their foundational systems. He advises that before leaders chase the latest tech, they should first ask if the underlying business process is solid enough to support any kind of improvement, otherwise they risk optimizing for superficial metrics at the expense of the actual customer experience. "You have to ask if your current process is solid and robust enough to be improved with AI, or if it simply doesn't work and you have to start from scratch," he says. "You must think from a business perspective: Is my data correct? Is my infrastructure supporting that? What is the value generated for our customers and what is the ROI?"

• Roadblock to revenue: While working in e-commerce, Lozano's C-suite pushed to automate product returns, eyeing a big win for the P&L. But the project barely got off the ground before hitting a wall. "We faced a major issue: the process we had to manage stock inventory and spare parts wasn't robust enough to support the business case," he explains.

• Uncovering the real problem: But the stalled project was a gift. It forced the team to move beyond superficial fixes and ask about its own quality control problem. "The failure forced us to use a framework like the 'Five Whys' to get to the root problem. We began to question why we were getting so many returns and what was happening with our product quality assurance process. Only then could we identify what we truly needed to fix first," he says.

A superficial approach often masks deeper inefficiencies, leading to a false sense of progress and ultimately undermining true customer satisfaction. "I've seen businesses fall for a vanity metric. The time to resolve a ticket decreases because an LLM can answer twenty at once, but the actual outcome the customer wanted is not achieved. Then the brand experience is damaged and repurchase rates suffer."

Without deep understanding, any attempt at a solution, AI-driven or otherwise, risks addressing symptoms rather than the underlying issues that truly impact business outcomes. Lozano’s approach provides a roadmap to get unstuck, redesigning workflows for greater business impact with generative AI. While he encourages starting small, leaders should get expert guidance to establish a privacy and control environment, with solutions like clean rooms, to protect customer data.

• Permission to experiment: An iterative approach minimizes the risk of costly, large-scale failures and instead cultivates a culture of continuous learning and adaptation, where every attempt, successful or not, yields valuable insights. "Just embrace it. Try it and test it in small increments. There will be bumps on the road, but that doesn't mean it's not progress. I think not using it and not implementing it in your business hurts you more than using it, because you miss out on the learning."

• The pivot principle: A learning mindset transforms perceived setbacks into critical data points, guiding future iterations and ensuring that resources are continually redirected toward more effective solutions. "When you try something and it doesn't work, the question is not, 'Why didn't it work?' but 'What did we learn and how can we improve it?'"

Ultimately, the goal isn’t just to implement a single technology, but to focus on the outcome. "Embrace innovation, whether it's using AI, robotic process automation, or just the standard automation that has been available for twenty years," Lozano concludes. "Think about the outcome, how you're innovating the process, and what is the right thing to do for your customers, your stakeholders, your business, and your employees. Think from the human perspective and just experiment, learn, and grow."