On August 19, Lowe’s put a cumulative number on its AI shopping agent for the first time. Mylow has fielded roughly 25 million customer questions since launch, and shoppers who engage with it while browsing online convert at three times the rate of shoppers who don’t.
The same Q2 report carried a second set of numbers. For the quarter ended July 31: total sales of $26.0B, up 8.3%. Online comparable sales up 15.7%, the second straight quarter above 15%. Overall comparable sales up 0.2%. Comparable transactions down 2.1%. Comparable average ticket up 2.3%.
Read those together and the story inverts.
The agent moved conversion, not demand
Transactions fell 2.1% and ticket rose 2.3%. Multiply the two and you land almost exactly on the 0.2% comp. Fewer people bought from Lowe’s this quarter. Each of them spent slightly more. The pool did not grow.
So the 15.7% online number is not incremental demand. It is the same shrinking pool ordering somewhere else. What Mylow’s 3x conversion rate did was not turn non-renovators into customers; it kept the customer who would have driven to a store to ask one question on his phone instead.
Lowe’s own guidance confirms the direction. The company cut FY2026 to the bottom of its prior range: roughly $92B in total sales, down from $92B–$94B; comparable sales now flat, down from flat-to-up-2%; adjusted EPS of about $12.25; operating margin of 11.2%, down from 11.2%–11.4%. CFO Brandon Sink told analysts the company is assuming the second half looks a lot like the first.
Online up 15.7%, full-year outlook down. In this report, AI is an efficiency line, not a growth engine.
Whom were those 25 million questions being asked before?
The number worth sitting with is 25 million.
Home improvement is not apparel and it is not grocery. Customers walk into Lowe’s rarely knowing the SKU they need. They have a leaking sink, a breaker that keeps tripping, a floor lifting at the seam. The first thing they need is not merchandise. It is a judgment: which unit, which fittings, whether this is a job they can do themselves. That judgment used to come from the person standing in the plumbing aisle who had worked the department for fifteen years.
25 million questions is 25 million of those conversations relocated to a model.
Lowe’s built the store-side half too. Mylow Companion gives frontline associates real-time product knowledge across departments. EVP of Stores Joe McFarland told analysts that more associates are engaging with the tool every day, that it gives them the confidence to help customers outside their own department and lets them answer questions faster, freeing time for what matters most, serving customers.
That sentence is worth a second read. The associate does answer faster. The answer no longer originates with him.
What is being unbundled is expertise
We drew the line between codified and tacit knowledge on August 12: models eat the part that has already been written down, has a correct answer, and can be retrieved.
Home improvement product knowledge sits squarely on the codified side. Which fitting matches which pipe diameter, how many primer coats a given paint needs, whether a dishwasher model clears a 24-inch cabinet opening. All of it lives in product manuals, installation codes and closed work orders. A fifteen-year plumbing-aisle veteran’s professional value is substantially that he loaded those documents into his head and knows which line to pull when.
Mylow Companion moves that layer out of his head. What remains after the move? He can lift the 40-pound box, climb the ladder to the overstock, parse a customer’s badly described symptom, and read whether this person wants to save money or save a Saturday. Those capabilities survive. They do not price like expertise.
This is not a layoff. It is a repricing of what the job is.
Three big-box retailers, 48 hours, no headcount
The cohort comparison is unusually clean this week.
On August 18, Home Depot posted Q2 sales up 5.7%, crediting smaller projects and Pro demand. On August 19, Target’s Q2 release reported that digital traffic from external AI platforms like OpenAI and Google is growing more than 3.5x the industry rate. The same page reported 17 new stores in the quarter and more than 3,500 jobs created. That same morning, Lowe’s reported Mylow’s 3x.
Three of America’s largest retailers reported inside 48 hours. All three placed an AI surface at the center of the growth narrative. None of them attached a headcount number to it.
The silence has a cause, and we took it apart on August 4 in the Rackspace case: writing an AI pivot into a formal disclosure creates securities-litigation exposure. Retail carries a second constraint. Store employees are organizable and locally newsworthy in a way corporate functions are not. So AI enters the retail earnings call as a conversion rate and never as an hours figure.
Watch the schedule, not the WARN notice
If you work a store role in large-format specialty retail, this report will not reach you shaped like a layoff.
Retail does not adjust labor by announcement. It adjusts by scheduling. Total headcount can hold flat for a year while weekly hours allocated per store quietly drop 4%. The leading indicators are narrower than a press release: whether out-of-department questions increasingly come back as “check the tool first”; whether training budget shifts from product knowledge toward process compliance; whether new job postings swap “product specialist” for “customer experience.”
The clock here is slow. Mylow took more than a year to accumulate 25 million questions, and it worked the online side first, where no associate stood anyway. The pressure point is Mylow Companion’s adoption rate inside stores. Once more than half of associates reach for it as a first move rather than a fallback, the expertise premium is largely spent. Lowe’s did not disclose that rate.
Q3 is the next calibration window. Two things to watch: whether online comps clear 15% for a third consecutive quarter, and whether the company discloses a Mylow Companion usage figure for the first time. The first tells you how durable the channel shift is. The second tells you how far the store-side substitution has run.
If Q3 shows online still growing double digits while comparable transactions keep falling, aisle advisory work in home improvement has entered a contraction that will never require an announcement.
