How a semiconductor veteran turned over a century of horticultural knowledge into AI-led aggressive benefit
For many years, a ritual performed out throughout ScottsMiracle-Gro’s media amenities. Each few weeks, employees walked acres of towering compost and wooden chip piles with nothing greater than measuring sticks. They wrapped rulers round every mound, estimated top, and did what firm President Nate Baxter now describes as “sixth-grade geometry to determine quantity.”
In the present day, drones glide over those self same crops with mechanical precision. Imaginative and prescient methods calculate volumes in actual time. The transfer from measuring sticks to synthetic intelligence indicators greater than effectivity. It’s the seen proof of one in all company America’s very unlikely know-how tales.
The AI revolution finds an surprising chief
Enterprise AI has been led by predictable gamers. Software program corporations with cloud-native architectures. Monetary providers companies with huge knowledge lakes. Retailers with wealthy digital touchpoints. Client packaged items corporations that deal with bodily merchandise like fertilizer and soil weren’t anticipated to steer.
But ScottsMiracle-Gro has realized greater than half of a focused $150 million in provide chain financial savings. It experiences a 90 p.c enchancment in customer support response instances. Its predictive fashions allow weekly reallocation of promoting sources throughout regional markets.
A Silicon Valley veteran bets on soil science
Baxter’s path to ScottsMiracle-Gro (SMG) reads like a calculated pivot, not a company rescue. After twenty years in semiconductor manufacturing at Intel and Tokyo Electron, he knew how you can apply superior know-how to complicated operations.
“I type of initially stated, ‘Why would I do that? I’m working a tech firm. It’s an trade I’ve been in for 25 years,’” Baxter recollects of his response when ScottsMiracle-Gro CEO Jim Hagedorn approached him in 2023. The corporate was reeling from a collapsed $1.2 billion hydroponics funding and dealing with what he describes as “strain from a leverage standpoint.”
His spouse challenged him with a direct immediate. If you’re not studying or placing your self in uncomfortable conditions, you must change that.
Baxter noticed clear parallels between semiconductor manufacturing and SMG’s operations. Each require precision, high quality management, and the optimization of complicated methods. He additionally noticed untapped potential in SMG’s area data. 100 fifty years of horticultural experience, regulatory know-how, and buyer perception had by no means been absolutely digitized.
“It turned obvious to me whether or not it was on the backend with knowledge analytics, enterprise course of transformation, and clearly now with AI being entrance and middle of the buyer expertise, numerous alternatives are there,” he explains.
The declaration that modified every little thing
The pivot started at an all-hands assembly. “I simply stated, you realize, guys, we’re a tech firm. You simply don’t understand it but,” Baxter recollects. “There’s a lot alternative right here to drive this firm to the place it must go.”
The primary problem was organizational. SMG had developed into purposeful silos. IT, provide chain, and model groups ran impartial methods with little coordination. Drawing on his expertise with complicated know-how organizations, Baxter restructured the buyer enterprise into three enterprise items. Normal managers turned accountable not only for monetary outcomes but in addition for know-how implementation inside their domains.
“I got here in and stated, we’re going to create new enterprise items,” he explains. “The buck stops with you and I’m holding you accountable not just for the enterprise outcomes, for the standard of the artistic and advertising, however for the implementation of know-how.”
To assist the brand new construction, SMG arrange facilities of excellence for digital capabilities, insights and analytics, and artistic features. The hybrid design positioned centralized experience behind distributed accountability.
Mining company reminiscence for AI gold
Turning legacy data into machine-ready intelligence required what Fausto Fleites, VP of Information Intelligence, calls “archaeological work.” The staff excavated many years of enterprise logic embedded in legacy SAP methods and transformed submitting cupboards of analysis into AI-ready datasets. Fleites, a Cuban immigrant with a doctorate from FIU who led Florida’s public hurricane loss mannequin earlier than roles at Sears and Cemex, understood the stakes.
“The pricey a part of the migration was the enterprise reporting layer we now have in SAP Enterprise Warehouse,” Fleites explains. “You could uncover enterprise logic created in lots of circumstances over many years.”
SMG selected Databricks as its unified knowledge platform. The staff had Apache Spark experience. Databricks provided sturdy SAP integration and aligned with a choice for open-source applied sciences that decrease vendor lock-in.
The breakthrough got here by systematic data administration. SMG constructed an AI bot utilizing Google’s Gemini massive language mannequin to catalog and clear inner repositories. The system recognized duplicates, grouped content material by matter, and restructured data for AI consumption. The trouble diminished data articles by 30 p.c whereas rising their utility.
“We used Gemini LLMs to truly categorize them into subjects, discover comparable paperwork,” Fleites explains. A hybrid strategy that mixed trendy AI with methods like cosine similarity turned the inspiration for later functions.
Constructing AI methods that really perceive fertilizer
Early trials with off-the-shelf AI uncovered an actual danger. Normal-purpose fashions confused merchandise designed for killing weeds with these for stopping them. That mistake can spoil a garden.
“Completely different merchandise, in case you use one within the improper place, would even have a really adverse end result,” Fleites notes. “However these are form of synonyms in sure contexts to the LLM. In order that they had been recommending the improper merchandise.”
The answer was a brand new structure. SMG created what Fleites calls a “hierarchy of brokers.” A supervisor agent routes queries to specialised employee brokers organized by model. Every agent attracts on deep product data encoded from a 400-page inner coaching guide.
The system additionally modifications the dialog. When customers ask for suggestions, the brokers begin with questions on location, objectives, and garden situations. They slender prospects step-by-step earlier than providing solutions. The stack integrates with APIs for product availability and state-specific regulatory compliance.
From drones to demand forecasting throughout the enterprise
The transformation runs throughout the corporate. Drones measure stock piles. Demand forecasting fashions analyze greater than 60 components, together with climate patterns, client sentiment, and macroeconomic indicators.
These predictions allow quicker strikes. When drought struck Texas, the fashions supported a shift in promotional spending to areas with favorable climate. The reallocation helped drive optimistic quarterly outcomes.
“We not solely have the power to maneuver advertising and promotion {dollars} round, however we’ve even gotten to the purpose the place if it’s going to be a giant weekend within the Northeast, we’ll shift our area gross sales sources from different areas up there,” Baxter explains.
Client Providers modified as effectively. AI brokers now course of incoming emails by Salesforce, draft responses based mostly on the data base, and flag them for transient human overview. Draft instances dropped from ten minutes to seconds and response high quality improved.
The corporate emphasizes explainable AI. Utilizing SHAP, SMG constructed dashboards that decompose every forecast and present how climate, promotions, or media spending contribute to predictions.
“Usually, in case you open a prediction to a enterprise particular person and also you don’t say why, they’ll say, ‘I don’t imagine you,’” Fleites explains. Transparency made it potential to maneuver useful resource allocation from quarterly to weekly cycles.
Competing like a startup
SMG’s outcomes problem assumptions about AI readiness in conventional industries. The benefit doesn’t come from proudly owning essentially the most subtle fashions. It comes from combining general-purpose AI with distinctive, structured area data.
“LLMs are going to be a commodity,” Fleites observes. “The strategic differentiator is what’s the further stage of [internal] data we will match to them.”
Partnerships are central. SMG works with Google Vertex AI for foundational fashions, Sierra.ai for production-ready conversational brokers, and Kindwise for pc imaginative and prescient. The ecosystem strategy lets a small inner staff recruited from Meta, Google, and AI startups ship outsized impression with out constructing every little thing from scratch.
Expertise follows impression. Standard knowledge says conventional corporations can’t compete with Meta salaries or Google inventory. SMG provided one thing completely different. It provided the prospect to construct transformative AI functions with instant enterprise impression.
“When we now have these interviews, what we suggest to them is mainly the power to have actual worth with the newest data in these areas,” Fleites explains. “Lots of people really feel motivated to come back to us” as a result of a lot of massive tech AI work, regardless of the hype, “doesn’t actually have an effect.”
Workforce design mirrors that philosophy. “My direct experiences are leaders and never solely handle folks, however are technically savvy,” Fleites notes. “We all the time are always switching arms between creating or sustaining an answer versus technique versus managing folks.” He nonetheless writes code weekly. The small staff of 15 to twenty AI and engineering professionals stays lean by contracting out implementation whereas retaining “the know-how and the path and the structure” in-house.
When innovation meets immovable objects
Not each pilot succeeded. SMG examined semi-autonomous forklifts in a 1.3 million sq. foot distribution facility. Distant drivers within the Philippines managed as much as 5 automobiles directly with sturdy security data.
“The know-how was really actually nice,” Baxter acknowledges. The automobiles couldn’t raise sufficient weight for SMG’s heavy merchandise. The corporate paused implementation.
“Not every little thing we’ve tried has gone easily,” Baxter admits. “However I believe one other essential level is you must deal with just a few crucial ones and you must know when one thing isn’t going to work and readjust.”
The lesson tracks with semiconductor self-discipline. Investments should present measurable returns inside set timeframes. Regulatory complexity provides issue. Merchandise should adjust to EPA guidelines and a patchwork of state restrictions, which AI methods should navigate appropriately.
The gardening sommelier and agent-to-agent futures
The roadmap displays a long-term view. SMG plans a “gardening sommelier” cell app in 2026 that identifies crops, weeds, and garden issues from pictures and supplies immediate steerage. A beta already helps area gross sales groups reply complicated product questions by querying the 400-page data base.
The corporate is exploring agent-to-agent communication so its specialised AI can interface with retail companions’ methods. A buyer who asks a Walmart chatbot for garden recommendation might set off an SMG question that returns correct, regulation-compliant suggestions.
SMG has launched AI-powered search on its web site, changing key phrase methods with conversational engines based mostly on the interior stack. The longer term imaginative and prescient pairs predictive fashions with conversational brokers so the system can attain out when situations counsel a buyer could need assistance.
What conventional industries can study
ScottsMiracle-Gro's transformation affords a transparent playbook for enterprises. The benefit doesn't come from deploying essentially the most subtle fashions. As an alternative, it comes from combining AI with proprietary area data that opponents can't simply replicate.
By making basic managers accountable for each enterprise outcomes and know-how implementation, SMG ensured AI wasn't simply an IT initiative however a enterprise crucial. The 150 years of horticultural experience solely turned helpful when it was digitized, structured, and made accessible to AI methods.
Legacy corporations competing for AI engineers can't match Silicon Valley compensation packages. However they will supply one thing tech giants typically can't: instant, measurable impression. When engineers see their climate forecasting fashions straight affect quarterly outcomes or their agent structure forestall clients from ruining their lawns, the work carries weight that one other incremental enchancment to an advert algorithm by no means will.
“We’ve got a proper to win,” Baxter says. “We’ve got 150 years of this expertise.” That have is now knowledge, and knowledge is the corporate’s aggressive edge. ScottsMiracle-Gro didn’t outspend its rivals or chase the latest AI mannequin. It turned data into an working system for development. For a corporation constructed on soil, its largest breakthrough is likely to be cultivating knowledge.
