The old engine still works. MONDAY.COM IS REBUILDING IT ANYWAY.

Start with the number, because the number is not the problem.

In the second quarter of 2026, monday.com reported revenue of $364.6 million, up 22 percent from a year earlier. Non-GAAP earnings came in at $1.48 a share against a consensus near $1.11, a beat of $0.37. Revenue topped the roughly $355.6 million analysts expected. Non-GAAP operating income was $61.1 million, a 17 percent margin, several points above where the company had guided the year. The largest customers, the ones spending over half a million dollars a year, grew 68 percent. Remaining performance obligations, the contracted revenue not yet recognized, rose 34 percent to $937 million.

This is not a company in trouble. It is a company beating expectations across the metrics that usually settle the argument.

Management said as much. The co-chief executives, Roy Mann and Eran Zinman, told investors that "Q2 reinforced our conviction that our strategy is working and that it was time to move faster." The chief financial officer, Eliran Glazer, said the results "demonstrate the underlying strength of the business."

Hold that in place. Beat, widening margin, fastest growth in the biggest accounts, confident management. Now look at what the same company did around the same moment.

Three weeks before the quarter was reported, monday.com told the market it would cut about 20 percent of its workforce, a reduction outside reporting put at roughly 600 to 630 jobs. Companies that are failing cut to survive. This company was not failing, and it said the cut was not about survival. In its filing with the Securities and Exchange Commission, monday framed the reduction as a move to "align the Company's organizational structure with its strategic focus on the AI Work Platform," part of an "ongoing transformation of its product, marketing, and go-to-market strategy" toward "a leaner, more focused operating model." It expected to keep hiring in what it called key strategic areas.

Read the sequence, not the headline. A profitable, growing company chose to remove a fifth of its people, and it described the reason as a redesign, not a retreat.

The financial mechanics confirm the intent. The restructuring carried estimated charges of $45 to $55 million. At the same time, monday raised its non-GAAP operating margin outlook for the year, from roughly 13 percent toward 15 percent, and by the time Q2 was reported the full-year margin guide had moved to about 16 percent. The cut did not patch a hole. It funded a change and lifted profitability while doing it.

A 20 percent reduction is the most expensive signal a management team can send, because it is the hardest to reverse and the most closely watched. monday spent that signal not to cover a miss, but to move resources toward a version of the business it does not yet run at scale.

To see what monday is building, you have to see how it used to grow.

For its entire history, monday expanded an account one way: by adding people. More seats, more upgrades, more teams inside the same customer. The software was priced per user, so growth inside an account was, in effect, a headcount bet. It worked. It is most of how the company reached this size.

That is the engine now being rewired. For new customers on the monday AI Work Platform after May 6, 2026, the purchase is no longer only seats. It is seats plus AI credits. In monday's own words, "seats cover the people using the platform, while AI credits cover supported AI usage." Credits are consumption. When the software's AI agents do work, they draw down credits, and when a customer wants more, they buy more. The support documentation is explicit about what this unlocks: a customer can "increase your usage as your team adopts more AI capabilities, without changing your seat count."

That last clause is the whole point, and it deserves plain language. Under the old model, a monday account got bigger mainly when the customer added employees to it. Under the new one, an account can get bigger because the customer's AI usage grows, even if the number of humans logging in never changes. The unit of expansion is no longer only the person. It is now also the work the machine does.

Management narrated the shift in the present tense. "Before the changes in the product and the pricing, the only way customers could expand was to add more people and more seats," Zinman told investors. "This is the first time since we added the new agent that we see customers expand not only on the seats for humans, but on AI consumption." Mann put it in operator's terms: customers "top up and reach the end of their consumption bucket and then add more, which is, for me, like the best indication that they get value and want more of it."

The early numbers are small but pointed. Revenue from AI products doubled from the first quarter to the second and made up 17 percent of net new recurring revenue in Q2. That is not yet a second engine. It is the first evidence that a second engine has been switched on.

A pricing change is a decision about economics. What monday did next was a decision about people, and it points the same direction.

The 20 percent reduction was not spread evenly. The chief revenue officer, Casey George, told investors that "the restructure of the go-to-market organization was primarily focused around non-quota carriers and downmarket resources," and that this "will afford us the opportunity to accelerate our investment upmarket." Translated: the company thinned the parts of its sales organization that chase smaller customers and roles that do not carry a quota, and moved that capacity toward larger accounts and toward a new model of embedding engineers with customers to get AI actually deployed.

That is a company changing not just how it charges, but who it sells to and how. The forward-deployed engineer is a tell of its own. Zinman said customers "want to adopt AI, but a lot of them don't know how to do it," and that helping them "deploy products" is now "very strategic for the company." Selling software to a buyer is one motion. Standing up an AI system inside a large enterprise is another. monday is staffing for the second.

It went further at the product line. Zinman said monday is "sharpening our focus on monday service and monday CRM," and that "each will operate with its own dedicated product development, its own go-to-market motion and its own investment road map." A company confident in one universal motion does not split its products into separate motions. It does that when it has concluded that different buyers now require different machines.

None of these are isolated announcements. Pricing, headcount, sales coverage, and product organization all moved in the same quarter, and all moved the same way: away from the broad, seat-led, downmarket motion that built the company, and toward a narrower, consumption-aware, enterprise-led one.

Here is where the traditional scoreboard starts to disagree with itself, and where the most useful number is the one most people skimmed past.

That number is net dollar retention. It answers a simple question: take the customers you already had a year ago, and measure whether they are spending more or less with you now. Above 100 percent means the existing base is expanding on its own. It is, for a software company, the closest thing to a reading of whether the growth machine still compounds without new logos.

monday's net dollar retention was 109 percent in Q2. A year of history sits underneath that figure and it points one way. It was 112 percent for full-year 2024, 110 percent for full-year 2025, and 109 percent now. Management guided it to 108 percent for the full year. Glazer attributed the softness to lapping "prior pricing actions taken in fiscal years 2024 and 2025" and to tier upgrades and multiproduct expansion running "slightly below our original expectations."

The lazy reading is that retention is falling and that is bad. That is not the QNT/L reading, and it is not what the evidence supports. Gross retention, the measure of customers simply staying, is strong. Enterprise seats are still growing. The metric is not describing a business coming apart.

The metric is describing a business whose engine of expansion is being rebuilt underneath it. Net dollar retention was designed in and for a seat-led world, where an account grew because it added people. monday is now, deliberately, introducing a second way for accounts to grow that the seat-era version of the metric was never built to capture cleanly. The number is not broken. The business model is becoming larger than the number.

That is why the decline matters, and it is also why it is easy to dismiss. A metric losing a point a year does not look like an event. It looks like weather. But an instrument built to measure one kind of expansion will read low precisely while a company is shifting to another kind, and it will keep reading low until the new mechanism is large enough to show up in it. The instrument is not failing. It is early.

Management all but said so, in the one decision that should be read most carefully: it declined to raise full-year revenue guidance to reflect the Q2 beat. Glazer was direct about why. "We're observing the near-term cost of the 20% workforce," he said, "and this is something that we had to take into account because there is going to be a short-term impact. And we did not want to lay an aggressive top line raise on top of the execution risk that hasn't fully played out yet." The market heard it. Shares fell on the guidance, not on the quarter.

Do not overread that into distrust of the business. Read it precisely. Management itself refused to treat one strong quarter as proof that the transition was already safe. When the people with the most information decline to bank the beat, the beat is not the most important thing they told you.

Put the two halves of the quarter back together.

Revenue grew 22 percent. Margin expanded. The biggest customers grew fastest. Gross retention held. By every backward-looking measure, the system monday has run for years is still working. And while it was working, management cut a fifth of the company, rebuilt its pricing around machine consumption, pulled its sales force upmarket, split its products into separate motions, and warned that its headline expansion metric would keep declining. The old engine still works. monday is rebuilding it anyway.

That combination has a name worth giving it. Call it elective re-architecture: a company deliberately redesigns a functioning revenue system before the existing one has visibly failed. It is the harder version of the decision, and the more revealing one. Companies forced to rebuild a broken model are reacting. A company that rebuilds a working one has concluded something about where value is going and decided to move first. The pattern is not new. Adobe rebuilt a profitable license business into a subscription one while the licenses were still selling. The move looks premature right up until it looks obvious.

The QNT/L judgment is not that monday's seat model is dead. It plainly is not. Enterprise seats are still growing, and the company said so. The judgment is narrower and firmer: monday's management has decided the seat model is no longer sufficient by itself for the phase it is entering, and it is spending real, irreversible resources to prove that decision correct before the numbers force its hand.

This is what makes the transition consequential rather than routine. AI is loosening the grip that the human seat has held on how software companies grow inside an account. When a machine can produce more output without a proportional increase in people, expansion stops being purely a headcount event. Every company whose growth economics were built on seats now faces the same question monday is answering out loud: redesign monetization, coverage, and forecasting for that world, or wait for the old metrics to make the problem undeniable. monday is not the only company that will face this. It is an unusually clean, unusually early case of a company acting on it while it still has the option to.

The honest counterweight is that monday may get the sequence wrong. The new engine may not scale before the old one slows. Consumption revenue may prove lumpier than seats. The upmarket bet may thin coverage in a base that still pays the bills. These are real risks, and the softening retention figure and the cautious guidance are exactly where they would first appear. But every one of those risks is a risk of execution, not of direction. The direction is no longer in question. Management has already placed the bet.

monday.com may execute this transition superbly. That is not the signal, and it is not our claim. The claim is that the transition is already underway, funded out of a quarter the company beat, and that the instruments most people watch are the last ones that will describe it.

Revenue grew. Margin improved. Earnings beat. None of those is the number that will tell you whether the rebuild worked, because all of them measure the engine monday has decided is not enough. The question is no longer whether monday is changing its revenue architecture. It is whether the new architecture compounds before the old one loses momentum. That contest is now the entire story, and the strong quarter that opened this brief is the least informative thing about it.

We do not report the signal. We interpret it. This is our interpretation: the scoreboard said monday won, and monday, reading the same scoreboard, decided winning was not the same as being built for what comes next.