6 Tactics to Generate Real Business Value from AI
▼ Summary
– There is growing concern that AI is a bubble, with MIT reporting only 5% of AI projects deliver value, but leaders like Fausto Fleites are addressing these fears through tactical approaches.
– Fleites emphasizes iterating quickly, learning from mistakes, and avoiding multi-year projects to adapt to AI’s rapid evolution and ensure clear business value.
– He advises creating a strategy with measurable KPIs, forming cross-functional teams, and investing in foundational technologies to make data accurate and accessible for AI applications.
– At Scotts, AI has improved consumer-facing features like search and chatbots, enabling natural language queries and personalized product recommendations without hallucinations.
– The company is also using AI to automate back-office processes, such as rewriting customer service emails quickly and cohesively, freeing staff for higher-value work and building trust in AI use cases.
While many businesses struggle to demonstrate tangible returns from artificial intelligence investments, a strategic approach focused on specific use cases can deliver measurable results. Fausto Fleites, Vice President of Data Intelligence at Scotts Miracle-Gro, has successfully implemented AI initiatives that generate real business value through careful planning and execution. His experience demonstrates that organizations can achieve significant returns by aligning AI projects with clear business objectives rather than chasing technology for its own sake.
Fleites joined the 150-year-old gardening company in early 2023 after holding senior digital leadership roles at Sears and Accenture. He immediately began building the necessary technological infrastructure using platforms from AWS and Google, establishing a foundation for applying deep-learning models to enterprise data. This groundwork enabled the creation of actionable insights for executive decision-making while supporting the development of generative and agentic AI applications.
Over the past year, Fleites has cultivated interest in AI across the organization through tactical implementations that demonstrate clear value. “We’ve been very deliberate about selecting use cases that deliver quick wins,” he explains. “This success has naturally sparked interest from other departments we’re targeting for future implementations.”
His comprehensive AI strategy encompasses six key principles that other digital leaders can apply:
Embrace rapid iteration and learning from failures. “We developed our product recommendation prototype extremely quickly using tools like ChatGPT and Gemini,” Fleites notes. “This allowed us to quickly identify limitations specific to our business and pivot accordingly. Creating a culture that values rapid experimentation and learning from mistakes proves essential for success.”
Start with manageable projects and leverage specialized expertise. “Avoid multi-year ‘big bang’ initiatives that lack clear business value, especially with AI evolving so rapidly,” he advises. “Be skeptical of anyone claiming decades of generative AI experience. Instead, focus on adapting as the technology matures.”
Develop a coherent strategy with measurable key performance indicators. “Many companies implement AI simply because they’ve read about it online,” Fleites observes. “They deploy technology for technology’s sake and inevitably fail. Begin by connecting AI use cases to business strategy with clearly defined, quantifiable KPIs.”
Foster cross-functional collaboration and inclusive participation. “We’ve established cross-functional teams that operate like startups,” he says. “This structure enables us to address short-term use cases effectively. Organizations must undergo cultural transformation to become more agile, collaborative, and data-driven.”
Strengthen foundational platforms and refine methodologies. “Invest in technologies that improve data accuracy and accessibility,” Fleites recommends. “For our AI chat functions, we leverage our 150 years of product knowledge. However, we’ve had to reformat that information in various ways to optimize AI efficiency.”
Prioritize change management and demonstrate concrete benefits. “We began developing use cases nearly eighteen months ago and continue presenting potential applications to the company,” he explains. “Early successes in consumer services have created momentum that allows us to replicate those achievements in other areas.”
Customer-facing AI implementations have yielded particularly strong results in search functionality and chatbot services. The AI search application, built on Google Vertex AI, enables customers to use natural language queries rather than requiring specific product terminology. “Previously, customers needed to search using exact terms like ‘fertilizer’ to receive relevant results,” Fleites notes. “Now they can phrase questions naturally and still obtain accurate answers.”
The company’s web-based chat agent, powered by Sierra’s personalized AI technology, handles product recommendations without generating inaccurate information while addressing common troubleshooting issues like grass seed germination. The system connects to live agents for more complex inquiries. “The chat function provides sophisticated product recommendations by recognizing customer intent and asking probing questions,” Fleites explains. “If someone requests lawn fertilizer, the AI engages in multiple levels of questioning before recommending appropriate products with location-specific restrictions.”
Internally, Scotts utilizes AI as a copilot for repetitive tasks, freeing employees to concentrate on higher-value strategic work. “While many companies focus exclusively on customer-facing chat applications, the real return on investment for organizations like ours comes from automating back-office processes,” Fleites emphasizes.
The company has already implemented an Email Rewrite service that transforms text from internal Salesforce knowledge articles into coherent, brand-appropriate responses. “This agentic tool rewrites emails in under thirty seconds while maintaining our brand voice and ensuring textual cohesion,” Fleites reports. “Staff can experiment with different brand tones, and we’ve achieved significant efficiency gains while improving email quality dramatically.”
Through an analytical process called X-Ray, Fleites’ team continues identifying additional backend automation opportunities. “We’re examining how agentic automation can function as assistants to enhance job performance,” he says. “We’re actively addressing concerns to build employee trust in these use cases, recognizing that successful implementation requires both technological capability and organizational acceptance.”
(Source: ZDNET)
