Context
This episode is a live conversation from the Masters of Scale Summit in San Francisco, hosted by Jeff Berman. The guest is John Furner, Walmart's president and CEO, eight months into the job and 33 years into his career at the company. He talks about how Walmart is changing in the age of AI: a fast leadership overhaul, a new short description of what the company is, and AI shopping agents that customers can control. For PMs, the episode is a look at how a very large company decides what AI to build, how it keeps customers in control of automation, and how leaders push change through a big organization without breaking trust.
The Big Idea
At Walmart, AI is an enabler, not the strategy: it should help customers say what they want in plain words and help employees do their jobs faster, while people stay in control.
Furner sums the company up as "people-led, tech-powered, and omni-channel retail," and every AI move he describes fits that order. Agent profiles with spending limits, AI tools that frontline workers build for themselves, and plain-language shopping all start from a human need and use technology to serve it.
Key Insights
Make big changes from strength
Furner moved fast in his first months: he overhauled the C-suite and joined up a tech stack that had long been split into silos. His reason is a phrase he thinks about a lot, that the best time to fix the roof is when the sun is shining.
- Why now: he sees the AI era as a time when customers and retail will change quickly, and "the customer is going to move" whether Walmart is ready or not.
- The goal: have the right people in the right roles before that shift, so Walmart stays strong over the next few years.
- A recent example: Walmart combined its Walmart+ and Sam's Club+ memberships, so buying one gets you the other. He says it took a lot of coordination across departments and silos to land.
For PMs, the lesson is that a change is easier to make when the product is healthy than when it is already losing, because you have the room and resources to do it well.
Customer intent is the hard problem
Furner describes AI's main job for shoppers as understanding what they mean when they say it in their own words.
- The request: a customer might say they need dinner tonight, they want tacos, there are four people, and it must be gluten-free.
- The complexity: matching that to a catalog of about a billion items, knowing where inventory is, getting the price right, and delivering it in about 25 minutes.
- The scale: Walmart serves about 275 million customers a week, and Furner says the aim is to understand each one at the individual level.
He calls the result something that can "feel magical," but the work behind it is joining language understanding to inventory, pricing and delivery. For a PM, the value of an AI interface depends on the systems behind it, not only on the model that reads the request.
Agents need customer-set limits
Walmart was about to launch agent profiles in the coming weeks. Whether a customer uses their own AI agent or Walmart's, they can delegate authority to it.
- The customer can set what the agent is allowed to buy.
- The customer can set a spending limit, or limits on how often the agent can buy in a day.
- Walmart's systems will recognize that an agent is shopping on behalf of a specific customer, not the customer directly.
Furner compares it to giving a family member your credit card, like his kids in college who have a card with a lower limit than his. His reason is that customers and employees both want to stay in control. He also admits it is early: "It's hard to say where the world will be in a year or two." For PMs building agent features, control and limits set by the user are part of the product, not an add-on.
AI should help frontline workers too
Furner sees a second, less obvious AI opportunity: Walmart's 2.2 million employees. The goal is to remove friction for them, give them the information they need on their phone, and suggest the best next action for their team. He ties this to Walmart's business model as an everyday low-cost operator, so internal speed directly supports low prices.
- Bottom-up building: an associate in Canada named Michael wrote a script for himself on an internal platform Walmart rolled out the previous December. He told Furner it saved him 15 to 18 minutes per shift.
- Training at scale: about 80,000 associates have taken AI certifications. The first to finish was a 40-year associate named Darlene.
- The belief: Furner says "everyone can be a builder" with the tools available today.
For a PM, internal users can be a large and fast-moving audience for AI tools, and giving them the means to build can surface uses a central team would not think of.
Pushback is often doing the job
Furner says the hardest part of moving fast is that employees were hired to do a specific job, with incentives and metrics they are held to. When they resist a change, they are often just doing what leadership asked them to do before. His answer is empathy, plus constant context and alignment.
- He holds a session every Friday with about 500 top US leaders to get feedback and ideas.
- He still believes change has to come from the top, because "the bottom can't change everything on the way up."
- He thinks trust between people is what makes speed possible, which is why Walmart stays a very in-person business that makes decisions quickly.
This links back to the first insight: a fast overhaul only works if the people who must change understand why, and if their goals are updated to match.
Fixed values, changing methods
Furner separates what never changes from what should change. Walmart's core values (respect for individuals, striving for excellence, acting with integrity, and service to others) and its mission to save people money so they can live better stay fixed. What Walmart does and how it delivers can and should change over time. He also says loyalty in retail is just the absence of something better, so the company always has to look for ways to get better for customers. For PMs, a stable mission is what lets a team change the product boldly without losing its identity.
Mental Models & Frameworks
People-led, tech-powered, omni-channel
This is Walmart's new short answer to "what does Walmart do?" Furner says that before, his honest answer was "how much time do you have?" because the company is so big and complex. He says he loves both the phrase and the order of its parts.
- People-led: people come first, and technology serves them.
- Tech-powered: technology helps employees succeed in the shortest time possible, as it has since Walmart's early technology bets in the 1970s and 1980s.
- Omni-channel retail: serve customers however they want, at the counter, at the curb, at their home, or in their fridge.
A PM can use the same kind of ordered statement to settle debates about whether a feature puts the technology ahead of the people it is meant to serve.
Agent as a trusted family member
Furner's mental model for AI shopping agents is a family member you have handed your credit card to. You trust them to buy for you, but you set rules on what and how much. This framing helps decide which controls an agent feature needs: what the agent can buy, how much it can spend, how often, and how the store recognizes that it is acting for someone.
Innovation at the speed of trust
Furner says "innovation happens at the speed of trust." In his view, trust forms between individuals first and then extends to the team. He argues it is hard to trust someone you have only seen as a box on a screen, which is why Walmart works in the office and brings people together. For a PM, this is a reason to invest in relationships with engineering, design and partner teams before you need fast decisions from them.
Support, not headquarters
At Sam's Club, the main office building is called "club support," not the home office. The name reflects that its job is to support the clubs where members meet employees. Furner says the home office's job is to support store teams, whose managers now run e-commerce, a delivery station, picking and stocking, and a supercenter all at once. Product and platform teams can use the same framing: the central team exists to serve the people closest to the customer.
Decision Principles
Principle: Change while you're ahead
- When: a product or team is doing well, but you can see the market shifting.
- Why: Furner moved fast on leadership, tech and memberships because change is easier with momentum and strength than after the customer has already moved on.
Principle: Ship controls with delegation
- When: building any feature where software acts for a user, like an AI agent that buys things.
- Why: Walmart's agent profiles let customers set what an agent can buy, how much it can spend and how often, because Furner says customers want to stay in control.
Principle: Treat resistance as misaligned incentives
- When: a team pushes back on a new direction.
- Why: Furner says people resist because they are following the metrics and incentives they were given, so the fix is context and new alignment, not frustration.
Trade-offs & Nuance
Top-down change versus listening
Furner holds both views at once. He says the best ideas come from Walmart's people, and he runs weekly listening sessions with about 500 leaders. But he also says change often has to come from the top, because the front line cannot change everything on its own. The balance is to listen widely while leadership aligns on the big moves and keeps the organization moving at pace.
Self-service versus human connection
The host raised the loneliness crisis and asked how Walmart balances convenience with being a community gathering place. Furner says customers often value their local store manager and pharmacist, and that his daughters and their friends enjoy shopping in person, but sometimes just want it taken care of for them. His answer is that flexibility matters, and so does connection. Automation and AI tools should free up time, not remove the human part of the experience.
Speed versus early uncertainty
Furner is pushing hard on AI agents while openly saying it is early and nobody knows where things will be in a year or two. His approach is to position teams so Walmart can follow the customer wherever they go, rather than betting on one fixed prediction.
Practical Application
Write your ordered one-liner
- Do: write a three-part description of your product, like Walmart's "people-led, tech-powered, omni-channel retail," where the order of the parts shows what comes first.
- Then: use it in roadmap reviews to check whether each AI feature serves the people in the first part.
- Why it works: Furner says this phrase replaced a vague "how much time do you have?" answer and gives the company a clear order of priorities.
Design agent permission settings
If you are building an AI agent that acts for users, list the limits a user would expect, based on Walmart's agent profiles:
- What the agent is allowed to buy or change.
- A spending or usage cap.
- How often the agent can act in a given period.
- A clear way for your system to know an agent, not the person, is acting.
Test plain-language requests end to end
Take a real multi-part request, like Furner's example of gluten-free tacos for four delivered tonight. Trace what your product would need to answer it: understanding the request, finding the right items, checking stock and price, and delivering on time. Note which step fails first, since that is likely where to invest.
Give internal users build tools
Find a group of frontline or operations staff and give them a simple way to automate their own tasks. Track time saved per person per shift, the way a Walmart associate reported saving 15 to 18 minutes per shift with a script he wrote himself.
Update incentives before launching change
Before rolling out a new direction, list the metrics and goals each affected team is measured on. Change those goals or explain the conflict up front, since Furner says what looks like pushback is often people doing exactly what they were asked to do.
Questions to Consider
- If our customers could send an AI agent to use our product for them, what limits would they want to set on what the agent can buy, spend or change?
- Is our product strong enough right now to make a big change, and are we waiting until it is losing before we act?
- When a team pushes back on a new direction, have we checked whether their current metrics and incentives are telling them to resist?
- Could we describe our product in a short, ordered phrase, the way Walmart uses "people-led, tech-powered, omni-channel retail," and would our current AI roadmap fit that order?
- Which internal teams could build their own AI tools if we gave them access, and how much time per shift could that save?
Bottom Line
Walmart's approach to AI puts people first and technology second: understand what customers mean in their own words, let them set firm limits on agents acting for them, and give employees tools to build their own fixes. Making those changes fast works only when leaders move while the business is strong and give teams the context and trust to follow.
Case Studies Mentioned
Walmart agent profiles
Walmart was preparing to launch agent profiles in the weeks after the summit. Customers can let their own AI agent or Walmart's agent shop for them, while setting what it can buy, a spending limit, and how often it can buy in a day. Walmart will recognize the account as an agent shopping on behalf of a named customer. The lesson is that trust in an agent feature comes from visible user control.
Walmart+ and Sam's Club+ merged
The day before the conversation, Walmart combined its two membership programs, so buying one membership gets you the other. Furner says there were many reasons for and against it, and it took a lot of coordination and management across silos to land. He frames it as giving customers access to more of what Walmart offers, done from a position of strength.
An associate's self-built script
In Canada, Furner met an associate named Michael who had written a script for himself on an internal platform Walmart had launched the previous December. Michael said it saved him 15 to 18 minutes per shift, and he saw it as a way to help his team be more productive. It shows how giving frontline workers simple build tools can produce real gains without a central project.
People to Follow
John Furner
President and CEO of Walmart, the sixth person to hold that role in the company's 64-year history. He started part time in a Walmart garden center 33 years ago, and has worked at Sam's Club and Walmart US and lived in China and Mexico. He is leading Walmart's AI push, including agent profiles and AI training for associates.
Notable Quotes
"The best time to fix the roof is when the sun is shining." (John Furner)
"In leadership, you can be frustrated when you get pushback, but it's not pushback. They're doing what we've asked them to do." (John Furner)
"It's like having a family member that you've given your credit card to." (John Furner)
"Innovation happens at the speed of trust." (John Furner)
"We have the tools today where everyone can be a builder." (John Furner)
