Home Depot

Home Depot runs 2,359 stores, sells to Pro contractors who visit their stores an average of 60 times a year, and closed fiscal 2025 with $164.7 billion in net sales. 

Their Pro relationship shaped almost everything the company has built with AI over the last two years.

By mid-2024, the company had more than 175 machine learning and generative AI pilots running at once, according to then-CIO Fahim Siddiqui. But only a few of these (Magic Apron, the AI voice agents, Blueprint Takeoffs, and Material List Builder) have turned into tools a customer or a Pro can use.

This case study helps you understand how Home Depot uses AI to run its business, why they had over 175 pilots running at the same time, how AI is used to serve their customers, how they operate their deliveries, and what Home Depot has built for its own associates. 

Home Depot splits its customers into two groups:

The company has spent the last two years buying its way deeper into the ‘Pro’ group, acquiring SRS Distribution in fiscal 2024 and then GMS through SRS in September 2025.

Roughly 80% of Home Depot's sales still occur through physical stores. But the company now runs a $25 billion e-commerce business, and about half of its US online orders are fulfilled through a store rather than a distribution center. It accounts for more than 6 billion visits a year across stores, website, and its mobile app.

Home Depot moved its enterprise data warehouse to Google Cloud's BigQuery, growing it from a 450-terabyte on-premises system to more than 15 petabytes. 

Because of this data foundation, Home Depot could run dozens of AI pilots at the same time without rebuilding infrastructure for every new dataset.

Angie Brown, EVP and Chief Information Officer since May 2025, has been quite vocal about Home Depot not trying to narrow its AI initiatives down to a handful of flagship projects.

Its AI efforts are focused around three priorities : 

In March 2026, Home Depot assigned the EVP and Chief Technology Officer role to Franziska "Fran" Bell. This was the company's first executive position scoped specifically to AI, data science, and product engineering.

To understand the breadth of Home Depot’s AI strategy, it helps to look at how the company uses AI to serve customers and improve its internal operations. 

A typical Home Depot store stocks 30,000 to 40,000 items. The online catalog runs past a million, roughly 30 times that. That’s more than any associate could ever walk a customer through by hand. 

To tackle this challenge, the company launched Magic Apron on March 6, 2025. It is built to answer all kinds of questions that an orange-aproned associate would answer in an aisle.

It shipped first across millions of product pages on homedepot.com and in the mobile app, running against Home Depot's own knowledge base plus general web information. 

The company has since added an Outdoor Assistant that lets a customer photograph a plant for care and sunlight guidance, and integrated its product catalog into ChatGPT for visual browsing and comparison. 

As of August 27, 2026, Home Depot finished rolling Magic Apron out to 2,000+ stores nationwide, adding aisle- and bay-level product mapping, real-time local inventory, and voice, text, and image input in multiple languages.

According to Broggi, they get millions of questions per month on Magic Apron, and it continues to grow.

As a customer, calling a Home Depot store meant listening to a list of options, pressing a number, and waiting before a representative could help. Home Depot built AI voice agents to skip this long process.

It tested AI voice agents built on Google Cloud’s Gemini Enterprise for Customer Experience in 50 stores before expanding the rollout. According to the company, the system understands what a caller needs in about 10 seconds, roughly four times faster than a traditional phone tree. 

The company says its voice agents can also build a shopping cart from a spoken project description or translate in real time for a non-English speaker.

Home Depot plans to expand the system to stores nationwide over the following year.

Pro customers make up close to 50% of Home Depot's revenue. Pros want to get their estimates faster and the materials right the first time. 

Knowing these are two different problems, Home Depot built two separate tools for each.

Blueprint Takeoffs turns a process that took weeks into one that takes days. A Pro brings their project to a Home Depot Pro Desk associate or an Outside Sales Representative. The associate runs the scope through the AI tool to produce a complete material list and quote for a single-family construction project. 

This is their self-service counterpart, aimed at smaller, faster-moving jobs rather than full blueprints. A Pro can describe a bathroom remodel in plain language, dictate a list hands-free on a jobsite, paste in something already typed into a notes app, or start with a template for a common project type. The tool returns a draft list grouped by phase, which the Pro edits; then the tool adds pricing and inventory information before placing an order with a single supplier.

The supply chain side of the business is where the benefits of using AI show up most clearly. 

Home Depot has also handed AI to its employees. The table below shows how thousands of corporate associates at the company's Store Support Center use AI.

Home Depot treats an AI experience as an extension of the person wearing the orange apron, and it holds it to the same standard.

Angie Brown has described this as an ongoing effort: "We're continuing to evolve so that the brand voice shows up the way we want it to." She also mentions controlling what data an AI partner can access, so the interaction still feels like it came from Home Depot rather than a generic model.

Weigh your AI portfolio by what moves the numbers, not by what's easy to show

Tools like delivery risk scoring, fulfillment routing, and warehouse automation are harder to show a board than a chatbot, but they're what actually produced Home Depot's efficiency and satisfaction gains. 

Running many small AI pilots is a strategic move

Experiment with several small pilots instead of committing to a few large ones. This process filters the options to those that fit your use case, so you don’t commit to one that may not work. 

Buying AI capability works as well as building it 

SRS, GMS, and SIMPL brought automation and distribution expertise that Home Depot didn’t have, and it would have taken years to build them. When a gap is specialized enough, buying the team/tool that has already solved it can beat building from scratch.