Walmart says training and tools are why associates feel more confident with AI

A confidence score can move because someone was taught, or because the question reached people who already liked the tool. The same percentage cannot stand for both, unless the survey shows the split. A certificate, a login and a feeling are three different facts, and a workforce of this size can hold all three at once without any one of them explaining the others.

Maren Waggoner's account is from the Workforce Executive Council Summit in New York, where she spoke with Leslie Picker. She is senior vice president and chief people officer for associate experience and digital at Walmart. The retailer asked its associates in June, for the first time, how they were experiencing AI. The workforce in that ask is more than 2.1 million people. More than 1.7 million responses came back. The finding the account is willing to lead with is modest and useful. Familiarity and comfort differ. Training and experience with the technology matter. That is a claim about a difference. It is not yet a claim about which number moved, or for whom.

The two population figures have to stay apart, because they answer different questions about different people. Seventy-seven percent of office-based associates said they are excited about AI's potential to improve their work. Among US frontline associates, 84 percent said they are confident using digital tools. The account treats that frontline confidence as a reflection of AI-powered tools those associates now use in the role. Reflection is the article's word for her reading. It is not a test that held the tools constant and watched confidence change. Office excitement about potential is not the same sentence as frontline confidence with tools already in the day. A slide that averages 77 and 84 has invented a third population the survey did not report.

What she offers as the reason is the practical provision, not a slogan about culture. "We've given them certificates, we've given them training programs and tools, and so they're using it in their day-to-day." The "so" does the work in that sentence. Certificates, programmes and tools are the cause she names, and day-to-day use is the effect she names. The account does not publish how many people hold a certificate, which programmes those certificates came from, or what share of the 1.7 million responses came from people who had done the training. A reader who hears "so" as a measured effect has gone past what the page can show. The page can show that this is the explanation she gives.

The partnerships are narrower than the survey. Over the last 13 months Walmart has worked with Google and with OpenAI on learning and on AI certification for associates, and the account says those programmes are at no cost to the associate. The figure attached to training, as distinct from the 77 and the 84, is the OpenAI one. Associates who went through that training said they had a 27 percent gain in confidence using AI. Said is her verb for their report. The account does not print the question, the baseline, or the number of people inside the 27 percent. It does not give a matching figure for the Google programme. Treating 27 percent as the confidence of the office, or of the frontline, or of both partners, folds three findings into one cell.

She sets the associate picture against a national one, and the two are not the same sample. An August poll from the Pew Research Center, as the account cites it, found that about 52 percent of US adults say they are more concerned than excited about AI. Walmart's own data, she says, shows greater comfort and excitement among associates. "Providing the tools, providing the training; they feel more confident." Adults in a national poll and associates inside one retailer are different people, asked different things. The contrast is the one she wants. It is not evidence that the programmes moved Walmart's people past the country's mood, because the page does not show the country's question asked of the associates, or the associates' question asked of the country.

The posture she claims for the company is older than this wave of tools. Walmart, on her telling, has a mission of being a people-led, tech-powered omni-channel retailer, aimed at helping people save money, and she says the people-led and tech-powered piece has been around for a while. The account calls AI the largest technological change in the company's 64 years, and then says the approach to transformation has not changed. That is a claim about continuity of method. It is not a description of the method. A mission line does not say who teaches, who signs a certificate or who is allowed to put an agent into a weekly report.

A certificate records that someone finished a sequence. It does not record the workflow that changed. The account shows no before and after on a named task, a report, a roster or a customer conversation, for the people who earned one. The Mexico manager is the closest it comes, and even there the sales result stays inside her phrase. An L&D lead who needs a task, an owner and a check will not find them in the 27 percent. They will find a feeling, a named partner and a claim that tools plus teaching are why the feeling moved.

Where the work itself is discussed, the account is more concrete, and still short of counts. She says Walmart is using AI to automate processes in fulfilment centres, to reduce friction, to use robotics for help and to focus the remaining work on human judgement. At the innovation lab in Dallas, efforts to take physical tasks off people have, on her account, created more quality-assurance roles rather than broader job losses. More is not a number. Broader job losses being absent from that lab is not a claim about every role in the company. The account does not say how many roles changed, what a quality-assurance role now consists of, or whether anyone left. An L&D lead who briefs the board that the lab proved there were no losses has added a finding she did not make.

The frontline of the store is a different choice again. She says customers actually like talking to a human at checkout. That preference, on her account, has let associates put their effort into using AI on customer problems, rather than into removing the person from the till. Liking a human at checkout is a customer fact. It does not say how many associates now use a tool in the aisle, or whether the 84 percent who feel confident with digital tools are the same people solving those problems. It does say the company is not treating the checkout conversation as the task to delete.

The example she is willing to put a person inside is one store manager in Mexico, met earlier this year. He built his own AI agent to look at reports and to drive fresh produce sales. "He's using the tool, but he's still running the store. The store manager job is still there. He's using the tools and technology to drive his business. And by the way, his fresh sales are through the roof." One manager is one manager. Through the roof is her phrase, not a percentage, and the account does not print the sales figure a finance partner could check. The useful part is the split she insists on. The tool is in the reports. The job of running the store is still his. A programme that celebrates the agent and goes quiet on who still decides has dropped the half of the sentence she bothered to say.

What an L&D or workforce lead can lift, without promoting a survey into a causal study, is the order of the claims. They asked the whole associate population, and they got a very large response, before they talked about feelings. They did not let an office percentage and a frontline percentage become one workforce. They attached a separate, smaller claim to the people who had actually been through a named partner's training, and they left the other partner without a matching number. They treated certificates, programmes and tools as the reason use became daily, and they did not publish the certificate count that would let anyone test the "so". They kept a human at checkout, pointed automation at physical work in fulfilment, and described one manager who built an agent without handing the store to it. They did not, on this page, say how many people finished, how the 27 percent was calculated, or what the fresh sales were.

1. Ask the population before you brief the feeling. More than 2.1 million associates were surveyed in June, and more than 1.7 million responded. The finding to keep is that familiarity differs and that training and experience matter. The finding to refuse is a single confidence score for the company.

2. Keep the three confidence lines in three cells. Office associates, 77 percent excited about the potential. US frontline associates, 84 percent confident with digital tools. OpenAI trainees, a 27 percent gain they reported in confidence using AI. The Google programme is named and not numbered. The Pew figure, about 52 percent of US adults more concerned than excited, is a different country and a different question.

3. Treat the certificate as a claim that still needs a count. Her reason for day-to-day use is certificates, programmes and tools. The account does not say how many certificates were issued, or what a person had to do to earn one. A programme no one can count is a different fact from the one she stated.

4. Keep the job in the sentence with the agent. The Mexico manager built an agent for reports and fresh sales, and she is explicit that he is still running the store. The Dallas lab, on her account, turned physical tasks into more quality-assurance roles, and she did not give the count. Customers liking a person at checkout is a third decision. None of the three is a promise that no role changes anywhere else.

An AI training session is a practical place to take one role a team already has, separate the people who finished the learning from the people who merely feel confident, and write down what that role still decides once a tool is in the report.