monday.com’s $36 Million AI ARR Turns Software Agents Into a Usage Pricing Test

The revenue is real, but the larger transition depends on whether credit consumption expands faster than seat growth while gross margin and cash conversion remain durable.

The monday.com AI ARR figure reached $36 million in the third quarter of 2026, giving investors the first clear evidence that the company can sell artificial intelligence as more than a feature bundled into a traditional software subscription. The figure matters because monday.com is attempting two changes at once. It is moving from a work management product toward a broader enterprise platform, and it is adding a usage based economic layer in which people still buy seats while software agents consume credits.

That combination creates a more demanding investment test than the headline suggests. Third quarter revenue rose 19% to $377.5 million, GAAP operating income reached $31.2 million, and non GAAP operating margin expanded to a record 24%. At the same time, management expects fourth quarter revenue growth to slow to roughly 14%, gross margin has begun to absorb higher artificial intelligence costs, and quarterly operating cash flow fell sharply from the prior year. The new revenue stream is visible, but it is not yet large enough to settle whether artificial intelligence will accelerate the business or mainly alter the way existing value is packaged.

The central question is therefore not whether customers will experiment with agents. Six million agent interactions in less than two quarters already show meaningful activity. The question is whether those interactions become recurring, expanding and economically attractive workloads. monday.com must prove that credits add a second growth engine without weakening the subscription engine that financed the transition.

monday.com AI ARR Changes the Meaning of Software Expansion

The third quarter earnings release filed with the SEC provides a strong operating picture. Revenue increased to $377.5 million from $316.9 million a year earlier. GAAP operating margin improved to 8% from negative 1%, while non GAAP operating income rose to $89.4 million from $47.5 million. Net income increased to $35.2 million from $13.1 million.

The company also continued moving upmarket. Customers producing more than $50,000 in annual recurring revenue increased 26% to 5,038. Customers above $100,000 increased 33% to 2,127, and customers above $500,000 increased 58% to 123. The largest deployment exceeded 87,000 seats. Total remaining performance obligations rose 27% to $946 million, while the current portion rose 22% to $759 million.

These figures show that the core platform still has enterprise momentum. They also explain why the artificial intelligence strategy deserves attention. A company with large deployments can distribute new tools through an installed base that already contains workflows, permissions, boards, data and administrators. An agent inside that environment can act on live organizational context rather than arrive as an isolated chatbot that employees must teach from the beginning.

The economic model is changing with the product. Under monday.com’s official artificial intelligence pricing model, new work platform customers purchase seats and artificial intelligence credits together. Seats cover the people using the platform. Credits cover supported artificial intelligence usage. A customer can increase credits without adding seats or changing the underlying plan.

That separation is strategically important. Traditional collaboration software monetizes the number of users. Agent based software can monetize the amount of work performed, even when the number of human users changes slowly. If one project manager coordinates ten digital agents, seat growth could understate the economic activity running through the platform. Credits create a way to charge for that activity.

The opportunity is not automatic. Usage pricing exposes the vendor to customer scrutiny that seat pricing often avoids. A seat is easy to budget. A variable pool of credits requires administrators to forecast consumption, allocate limits, evaluate outcomes and decide whether another unit of automated work is worth its price. The transition can deepen revenue when customers see measurable productivity. It can also create hesitation when consumption feels unpredictable.

The Quarter Shows Better Profitability and Slower Cash Conversion

The income statement improved substantially, but the cash flow statement introduces caution. The table below separates the strongest evidence from the unresolved parts of the thesis.

Metric	Third quarter 2026	Third quarter 2025	What it tests
Revenue	$377.5 million	$316.9 million	Core subscription growth remains healthy
GAAP operating margin	8%	negative 1%	The company crossed into quarterly operating profit
Non GAAP operating margin	24%	15%	Cost discipline created substantial operating leverage
Gross margin	approximately 88.0%	approximately 88.7%	Artificial intelligence and cloud costs require monitoring
Operating cash flow	$20.4 million	$95.1 million	Working capital weakened quarterly cash conversion
Adjusted free cash flow	$63.0 million	$92.3 million	Profit growth did not translate fully into current cash flow
Net dollar retention	108%	not disclosed in the release	Existing customers still expand, but the rate has softened
Artificial intelligence ARR	$36 million	not disclosed	A new monetization layer has become measurable

The margin improvement is real. Sales and marketing expense fell to $149.1 million from $165.8 million even as revenue increased. Measured against revenue, that expense declined from roughly 52% to about 40%. Research and development expense rose to $85.6 million from $79.5 million, but it declined as a share of revenue. These movements show that restructuring and operating discipline are producing leverage.

The cash flow decline requires a different interpretation. Operating cash flow dropped to $20.4 million from $95.1 million. The SEC filed cash flow statement shows that accrued expenses and other liabilities used $87.1 million of cash in the quarter, accounts payable used $15.0 million, and deferred revenue decreased modestly. These movements can be seasonal or timing related, so one quarter does not prove structural deterioration. They still matter because a software company is valuable partly for its ability to convert contracted revenue into cash before recognizing all of that revenue.

Adjusted free cash flow of $63.0 million remained positive and represented roughly 17% of revenue, but it was below the $92.3 million reported a year earlier. For the first nine months, operating cash flow declined to $180.4 million from $274.0 million. Management continues to expect full year adjusted free cash flow of $282 million to $288 million, equivalent to a 19% to 20% margin. The fourth quarter must therefore carry meaningful cash generation.

This tension is familiar across the current technology cycle. Block2Learn’s analysis of DigitalOcean’s artificial intelligence cloud growth showed that revenue associated with artificial intelligence can expand faster than gross profit when infrastructure costs rise. monday.com owns a different layer of the stack, but its economics still depend on model access, cloud infrastructure and data processing. A software interface does not remove the cost of the computation behind it.

Credits Create a Second Meter Beside Seats

monday.com has designed credits as a common unit across several products. Its official credit documentation states that simple agent tasks may consume roughly 10 to 50 credits, intermediate tasks about 50 to 150, complex tasks about 150 to 250, and extra complex tasks more than 250. The same pool can support agents, sidekick, workflows, notetaking, artificial intelligence blocks and vibe applications.

The company’s feature catalog lists a standard price of one cent per credit in the United States, although plan design, introductory discounts and enterprise contracts can change realized economics. A meeting hour for the notetaker consumes 120 credits. An artificial intelligence action on a block consumes eight. A successful sales agent call of up to five minutes consumes 150. These units make consumption visible, but they do not make every use case equally valuable.

This is where the business model becomes more sophisticated. A customer does not buy tokens of generic computing capacity. It buys outcomes inside workflows. One hundred credits used to classify low value records are not economically equivalent to one hundred credits used to qualify an important sales opportunity, summarize a legal process or prevent a project delay. The billing unit is common, while the value of the result varies widely.

That creates room for expansion. Once administrators see which use cases create measurable benefits, they can allocate more credits to those workflows without increasing seats. It also creates governance risk. Customers may cap usage aggressively if employees trigger expensive tasks without a clear return. The product therefore needs not only capable agents but also monitoring, permissions, limits and reporting that help finance and technology teams control consumption.

The structure resembles a utility layer inside a subscription. Seat revenue provides a recurring base. Credits convert incremental activity into incremental revenue. When it works, the model can raise revenue per customer and make the platform more valuable as automation increases. When it fails, customers treat credits as a tax on features they expected to receive inside the subscription.

The distinction matters because the initial pricing model applies to work platform customers who joined on or after May 6, 2026 and were offered seats plus credits. The transition will therefore unfold gradually across cohorts. New customers may adopt the combined model from the beginning, while existing customers can remain on different arrangements. Reported artificial intelligence ARR may grow through new logos, migration, higher usage, or mandatory minimum credit purchases. Investors need more disclosure before they can separate those drivers.

$36 Million Is Evidence, Not Yet Scale

The $36 million monday.com AI ARR figure is meaningful because it converts product activity into a financial metric. It is also small relative to the overall company. Comparing annual recurring revenue with annualized third quarter revenue is not an accounting comparison, but it provides a rough scale indicator. Annualized third quarter revenue is about $1.51 billion, placing disclosed artificial intelligence ARR near 2.4% of that run rate.

That is enough to demonstrate demand. It is not enough to offset a broad slowdown in the core business. Management expects fourth quarter revenue of $380 million to $382 million, which implies growth of about 14%. Full year guidance of $1.473 billion to $1.475 billion implies approximately 20% growth. The growth rate is moderating even as artificial intelligence ARR becomes visible.

The company reported in its second quarter release that artificial intelligence ARR had doubled from the first quarter and represented 17% of net new ARR. By the third quarter, the absolute figure reached $36 million and agents had completed more than six million interactions. Those milestones suggest rapid early adoption. They do not reveal retention, gross margin, paid credit consumption, or the proportion coming from minimum bundles rather than voluntary expansion.

The quality of the $36 million matters more than the speed of the launch. High quality revenue would come from customers that repeatedly buy additional credits because agents perform useful work. Lower quality revenue would depend on introductory packages, required minimums, or one time experimentation that does not renew. The decisive metrics are cohort expansion, credit utilization, paid top ups, renewal behavior and margin after compute costs.

This is similar to the distinction in Block2Learn’s review of Indian technology services and artificial intelligence pricing pressure. Artificial intelligence can increase output while reducing the number of billable human hours. Software vendors face the mirror image. They must capture some of the value created by automation even when human seat growth slows. Credits are monday.com’s proposed answer.

Gross Margin Is the Most Important Economic Boundary

monday.com reported GAAP gross margin of roughly 88% in the quarter, down from about 89% a year earlier. The change is modest, but the company’s 2025 annual filing explicitly states that cost of revenue includes artificial intelligence compute, model usage and data processing. It also says the company expects gross margin to decline modestly in the medium term as adoption of artificial intelligence offerings increases, before stabilizing as efficiency and scale improve.

That disclosure defines the boundary of the strategy. Usage revenue is attractive only if price exceeds the variable cost of the models, infrastructure and support required to deliver it. A traditional software seat can generate very high incremental margin after the platform is built. An agent that performs repeated model calls has a more visible marginal cost. Complex tasks can involve planning, data retrieval, several tool calls and multiple model responses.

The credit system can protect margin by connecting price to resource consumption. More complex tasks consume more credits. Customers can buy larger buckets as usage expands. The risk is that competitive pressure pushes credit prices down faster than model and infrastructure costs decline. Vendors may also include free credits to encourage adoption, which can accelerate usage without producing equivalent revenue.

The long term advantage may come from orchestration rather than raw model access. monday.com can combine organizational data, permissions, workflow state and application logic in a controlled environment. A customer could obtain similar language generation from many providers. It may be harder to reproduce an agent that understands the company’s boards, tasks, ownership rules, deadlines and connected applications.

That is also why security and governance become part of the economic moat. Enterprise customers need administrators to control which agents can access which records, what actions they can execute, and how much they can spend. Trust can increase usage, while a serious error can halt it. The value proposition is not simply intelligence. It is useful action inside a governed system.

Block2Learn’s analysis of Tempus AI’s path from data to margin highlighted a related principle. Proprietary context can make an artificial intelligence product more valuable, but the economics improve only when that context produces repeatable revenue at attractive margins. monday.com possesses workflow context rather than clinical data. The financial test is the same.

Enterprise Expansion Can Carry the Transition

The strongest evidence in the quarter may be the growth of large accounts. Customers above $500,000 in annual recurring revenue increased 58%. Those customers now represent 7% of annual recurring revenue. Customers above $100,000 represent 31%, up from 27% a year earlier. The concentration of revenue in larger accounts gives monday.com a practical route to scale agents.

Large organizations contain many repetitive workflows. Sales teams qualify leads, service teams classify tickets, project groups summarize updates, finance teams reconcile requests, and managers prepare reports. A small reduction in effort can have substantial value when multiplied across thousands of employees and millions of tasks.

Enterprise customers also impose stricter requirements. They expect identity controls, auditability, data protection, regional compliance and predictable spending. The move from software assistance to software action raises the cost of mistakes. A generated summary can be corrected. An agent that changes records, contacts a customer or triggers a workflow can create direct operational consequences.

The product must therefore balance autonomy with approval. Too little autonomy reduces productivity. Too much autonomy increases error and liability. monday.com’s installed workflow layer could help because agents can operate within existing permissions and structured processes. The company still needs to prove that adoption extends beyond experimentation and that agents can handle enough real work to justify recurring credit purchases.

The large account trend also protects the base business. Net dollar retention was 108% overall, 112% for customers with more than ten users, and 114% for customers above $50,000 in annual recurring revenue. Those figures remain above 100%, meaning existing cohorts expanded. They are not accelerating. The success of credits will be visible when retention stabilizes or improves as artificial intelligence products mature.

The Core Subscription Still Funds the Experiment

The company’s profitability improvement came largely from operating discipline, not from artificial intelligence revenue alone. Sales and marketing expense declined in absolute terms. Share based compensation fell. The company recorded restructuring charges and has described a broader effort to become faster and more focused.

This matters because transitions are easiest when the existing business generates cash and retains customers. monday.com does not need the $36 million artificial intelligence ARR to support the entire company today. It can develop the new layer while subscription revenue, enterprise expansion and remaining performance obligations provide financial stability.

The risk is that management optimizes the old business while investing too little in the new one. Research and development expense increased only about 8% year over year, below revenue growth, although non GAAP research spending increased more quickly. A lower expense ratio can reflect scale and focus. It can also become a warning if product quality, model choice or agent reliability falls behind larger competitors.

Competition is unusually broad. monday.com competes with project management vendors, customer relationship systems, service platforms, productivity suites and model providers that are adding agents. Microsoft, Google, Atlassian, Salesforce and many specialized companies can bundle artificial intelligence into existing enterprise relationships. The company’s defense is flexibility, ease of configuration and the ability to coordinate work across departments.

Block2Learn’s analysis of Jio’s digital monetization problem made the broader point that a platform earns a premium only when its technology becomes measurable revenue, margin and cash flow. monday.com has crossed the first boundary with disclosed artificial intelligence ARR. The remaining boundaries are durable expansion, unit economics and cash conversion.

Three Scenarios for monday.com AI ARR

Base scenario: credits deepen enterprise accounts gradually

In the base scenario, artificial intelligence ARR continues growing from a small base, but the core subscription remains the main driver of revenue through 2027. Large customers add agents to selected workflows and buy additional credits after measuring productivity. Net dollar retention remains near its present range, gross margin declines modestly, and operating margin stays healthy because sales efficiency offsets higher compute costs.

This scenario would validate the product direction without transforming the valuation immediately. Investors would see credits as a useful expansion mechanism rather than a separate business. The most important confirmation would be steady paid usage across several product categories, not one spectacular agent application.

Favorable scenario: agent activity becomes a second growth engine

In the favorable scenario, agent interactions rise rapidly and customers purchase credit top ups well above required minimums. Artificial intelligence ARR becomes a material portion of net new annual recurring revenue, large account retention accelerates, and the platform expands without equivalent seat growth. monday.com preserves gross margin through lower model costs, efficient routing and pricing linked to task complexity.

This outcome would change the company’s economics. Revenue could grow through both people and work performed. The installed base would become more valuable because every workflow could support additional automated activity. Operating leverage could remain strong if product led adoption limits the need for proportional sales spending.

Adverse scenario: usage grows but value capture disappoints

In the adverse scenario, customers test agents but limit paid consumption. Required credit bundles create initial ARR, while utilization and renewals remain weak. Competitors include similar tools inside broader suites, pressuring prices. Compute costs rise with activity, gross margin declines, and core revenue growth slows before credits become large enough to compensate.

This would expose the danger of confusing interactions with economics. Six million agent interactions can demonstrate engagement without proving that customers receive enough value to expand spending. If cash conversion remains weak and retention declines, the artificial intelligence strategy would look more defensive than additive.

What Investors Should Monitor

The first variable is paid credit expansion. Management should eventually disclose how much artificial intelligence ARR comes from minimum packages, voluntary top ups and enterprise contracts. A rising share from voluntary expansion would demonstrate that customers are choosing more usage after observing results.

The second variable is net dollar retention. Credits should lift expansion inside existing accounts. If artificial intelligence ARR grows while retention continues declining, the new revenue may be replacing weaker seat growth rather than adding to it.

The third variable is gross margin. A modest decline can be rational during adoption. A persistent decline without faster revenue would indicate that price is not covering model and infrastructure costs adequately.

The fourth variable is cash conversion. Quarterly operating cash flow was affected by accrued liabilities and other timing items, but full year guidance requires a strong fourth quarter. Investors should compare adjusted free cash flow with GAAP operating income, deferred revenue growth and restructuring costs.

The fifth variable is large account deployment. Growth above $100,000 and $500,000 in annual recurring revenue provides the distribution base for agents. Those customers should show higher retention and broader product use if the platform thesis is working.

The sixth variable is agent reliability and governance. Enterprise adoption depends on permission controls, audit trails, error handling and predictable spending. The market will reward interaction volume only when those interactions produce safe, useful work.

The Block2Learn Assessment

monday.com has produced enough evidence to take its artificial intelligence strategy seriously. The $36 million ARR figure is not a demonstration metric. It is contracted recurring revenue associated with a portfolio that includes credits, agents and vibe. Six million agent interactions show that users are engaging with the products. Enterprise account growth gives the company a credible distribution advantage.

The quarter also warns against treating every positive metric as confirmation. Revenue growth is slowing. Gross margin has edged lower. Operating cash flow weakened materially, and management expects the artificial intelligence transition to create near term operational pressure. The company has improved profitability partly by reducing sales and marketing expense, a strategy that cannot substitute permanently for product led expansion.

The credit model is economically logical. Seats monetize access by people. Credits monetize work performed by software. The two meters can coexist, and the second can grow even if headcount at customer organizations does not. That could protect monday.com from the possibility that artificial intelligence reduces the number of users required to complete the same amount of work.

The decisive question is whether customers view credits as a budget they want to expand or a limit they want to avoid. That answer will emerge through utilization, top ups, retention, gross margin and cash flow. The current evidence supports cautious optimism, not a completed transformation.

monday.com AI ARR Is a Better Test Than Interaction Volume

The monday.com AI ARR milestone marks the beginning of a measurable business model shift. It shows that the company can attach recurring revenue to artificial intelligence products while improving operating profitability and expanding large enterprise accounts. It does not yet show that usage pricing can carry the next stage of growth.

The transition will succeed when three conditions appear together. Customers must increase paid consumption after initial adoption. Gross margin must stabilize as model efficiency and scale offset higher compute activity. Cash generation must remain strong enough to fund product development without relying on cost reductions alone.

If those conditions hold, agents can become more than a feature inside a work platform. They can create a second economic layer in which monday.com earns revenue from both the people who organize work and the software that performs it.

If they do not, the $36 million figure will remain evidence of demand without evidence of durable value capture.

Continue Through the Block2Learn Learning Path

Evaluating monday.com requires more than following one quarterly growth rate. Investors need to understand recurring revenue, annual contracts, net dollar retention, deferred revenue, gross margin, operating leverage and free cash flow. They also need to distinguish a product adoption metric from the economics of the product that generated it.

The Block2Learn Learning Path develops that structure progressively. Free Start introduces the language of markets and business models. Foundation explains risk, valuation and capital allocation. The Investor Operating System helps convert quarterly evidence into a repeatable decision process rather than a reaction to a headline.

The more advanced layers connect technology adoption with competitive advantage, unit economics and portfolio construction. That framework is essential when a company is changing both what it sells and how it charges for it.

Artificial intelligence can increase activity quickly. Durable shareholder value requires revenue quality, pricing power, margin discipline and cash conversion to move with it.

Information is abundant. Structure is rare.