Fausto Fleites, vice-president of data intelligence at gardening specialist ScottsMiracle-Gro, is a different type of digital leader. “I’m very hands-on,” he says. “I’ve developed a lot of the machine-learning models that we use myself.”
This tactile approach means Fleites continues to hone his on-the-ground data experiences. While promotion to senior executive roles can mean some digital leaders become detached from the day-to-day data challenges that confound businesses, his hands-on approach means he has a sharp awareness of how technology can provide a solution to organisational issues – and that focus is crucial in an age of automation.
“I have a clear understanding of what we can do with AI and how to use it,” he says. “Most of the companies that are unclear about AI believe the marketing materials or are listening to someone who is inflating the power of AI. The key to success is understanding what AI can and cannot do, and how to apply AI in a way that’s a copilot, as opposed to something that does everything, end to end.”
Formerly in senior digital leadership positions with Sears and Accenture, Fleites started working with Scotts in February 2023. The 150-year-old company was eager to exploit its data more effectively, and he was keen to develop a strategy for transformation.
“They used to make a lot of decisions with Google Sheets, but the company had the market share to get all the juice out of advanced analytics, and there was an opportunity to do a lot with their information,” says Fleites. “So, to me, it was the perfect company to do something of real value with advanced analytics and everything that entails.”
Aiming for benefits realisation
As part of his initial data strategy, Fleites created a modern data lake in the Amazon Web Services (AWS) cloud and developed the first use cases for machine learning. These examples demonstrated how data could inform decision-making processes, such as understanding the impact of sales promotions and the effects of macroeconomic factors, and powering simulations to test operational plans.
“Through all that work, I showed that we can provide insights that change the way we do business, and I was promoted to vice-president of data intelligence,” he says, referring to his move to his current role in May 2024. “I now think about how we govern our data and AI assets more efficiently and how we do business transformation in the world of AI.”
Fleites says the shift in his role is due to broader changes in the technology environment. He recognises that a major change occurred with the release of ChatGPT by OpenAI in late 2022. Today, he says most companies want to use AI because they believe it’s the future of business. However, that’s not an end in itself.
“The question is, doing that work correctly in a way that provides return on investment to the business,” he says. “At Scotts, the demand for AI created a movement in the executive leadership to consider how we might use emerging technology to advance our digital transformation. And based on my experience of AI systems and machine-learning algorithms, I was suited to guiding the implementation of how to scale those efforts appropriately.”
Fleites says the direction of digital transformation at Scotts is a more general three-stage move towards benefits realisation. He suggests the foundational stage of digital transformation is where companies create a strategy for data-led change and secure executive buy-in. The second stage, implementation, focuses on digitising processes, deploying new technologies and supporting data-driven decisions.
The final stage of the digital transformation journey is realisation, which is where the business benefits come from the strategy that has been enacted. Fleites says Scotts is well advanced in its implementation phase and is starting to reap benefits as the organisation moves into the realisation stage.
“We’ve made significant progress in our digitisation efforts. We have many years of practice in process automation through RPA, we are in the midst of modernising our digital assets with AI and personalisation, and we’re deploying production use cases, both in machine learning and AI, that have changed how the business operates,” he says. “So, we are advanced in that implementation phase, and we are already reaping some of the benefits of the realisation in some of the use cases that we have implemented.”
Embracing machine learning
Fleites says one of his major accomplishments during his time with the company is establishing the organisation’s firm focus on machine learning. He says this approach allows senior executives to benefit from timely insights for decision-making processes.
“The way we use machine learning in the company is definitely a key achievement. We’ve not only deployed the technology in production forecasting tools, but we’ve also operationalised and applied explainable AI to break down that prediction into a why,” he says.
“Every week, business leaders in our sales teams, in supply chains and brands look at this insight to understand what happened, and that approach changes how we adjust our marketing investment for the next two weeks. So, data is having a meaningful impact on the way we operate.”
Fleites says the company’s machine-learning models are underpinned by its AWS data lake. His team uses Airflow to orchestrate and retrain its models. The type of model the company uses depends on the requirement.
For short-term predictions, the team uses XGBoost or LightGBM models. For store or regional-level predictions, which can involve thousands of product types, the team uses Google’s deep learning models.
He says the resulting insight-driven approach is crucial because many big companies concentrate on using machine learning to create operational benefits for end customers. “They use it for replenishment in the supply chain and things like that,” he says. “But being able to have the fine detail of data at the level of senior leadership is unusual.”
Exploiting generative AI
Fleites is also leading the release of innovative AI-enabled features for customers. Using a retrieval augmented generation (RAG) application in Google Vertex AI, customers can use natural language to search the company’s catalogue to receive answers to their questions.
The company is also using AI to improve the quality of conversations in its web-based chat agent. This technology uses data insights to address customer concerns, and they can be transferred to a live agent for detailed queries.
“I now think about how we govern our data and AI assets more efficiently and how we do business transformation in the world of AI”
Fausto Fleites, ScottsMiracle-Gro
The chat experience is powered by Sierra, a technology company that provides personalised AI agents for customer service, co-founded by ex-Salesforce co-CEO Bret Taylor. Fleites says his team feeds product catalogue and question information to Sierra through APIs that run on the Google Cloud platform.
“Those two features are just the first step in our AI journey,” says Fleites, referring to AI explorations in search and chat. “However, they are important, because now it’s just a matter of adding new journeys, enhancing our content, and going beyond that stage.”
While Scotts continues to make progress in terms of its AI-enabled front end for customer services, Fleites says the company’s AI strategy has another component focused on back-office elements. AI-powered copilots are being adopted to reimagine internal processes, so employees can dedicate their time to higher-value tasks.
Fleites says his organisation developed one use case in production about a year ago, known as Email Rewrite. The customer services team traditionally pieced together responses to emails using knowledge articles in Salesforce. This time-intensive process produced emails that felt incoherent and lacked brand voice.
Fleites’ team has developed an agentic tool that extracts text from the Salesforce system and creates cohesive emails in under 30 seconds with the right brand voice. He says the automation has also improved the quality of email responses by up to 10 times. Email Rewrite was the company’s first use case for back-office automation. Now, Fleites is looking for other opportunities through an internal process called X-ray.
“We want to understand the process from a technology point of view,” he says. “We’re thinking about how we operate with AI in the middle to prepare for the future of work. If I think about five to 10 years from now, as the technology around AI agents matures, this approach will become crucial to the back-office area of major companies.”
Delivering proactive personalisation
Fleites reflects on the changes he’s enacted so far and paints a picture of the data-led business he’d like to be supporting 24 months from now. Across customer services and internal operations, the aim is to move hard into benefits realisation, the final area of his digital transformation. He turns first to customer-focused services.
“In two years, our digital platform will be a different animal,” he says. “We are working on releasing the new version of our platform, where AI, personalisation and dynamic content are centre stage from the beginning, and we’re going to leverage our predictive capabilities to anticipate consumer needs. The way we engage with our consumers, in terms of e-commerce and digital assets, is going to revolutionise completely.”
Fleites says customers will be able to take a picture and ask questions about their lawns or plants via expert advice through generative AI. More generally, his organisation will use machine learning to anticipate customer requirements.
“For example, it could be that in July, we know from the weather forecast that humidity is going to be significantly high next week, and there’s a chance of lawn diseases,” he says. “So, in that case, we can proactively tell you what to do about it.”
In the back office, Fleites expects to deploy at least a dozen use cases of agentic automation throughout the company during the next 24 months. “That success would solidify the approach that we take and set the stage for the next phase of scalability into back-office automation,” he says.
