Monday.com's AI Credit Pivot: How Selling More AI Could Make the Company Less Profitable

We didn't see a pricing model innovation on May 1st. We saw a confession. Monday.com's shift to AI credits isn't a growth story. It's a margin compression story wearing a hockey stick costume. The official line: Monday.com is evolving from a Work OS to an AI Work Platform. Under the new hybrid model, even non-technical team members can configure native AI agents with one-click connectors to Anthropic, Microsoft, and OpenAI. The pricing structure that went live in May 2026 is deceptively simple — Basic, Standard, and Pro plans now include 1,000, 2,000, and 3,000 AI credits respectively. Exceed that, and you pay between $0.01 and $0.0125 per credit. Clean, familiar, digestible. The market rewarded the announcement with a 12.6% bounce, interpreting it as "the bad news is out, and here's a fresh story." Except the bad news isn't out. It's just changed shape. The stock had already lost over 50% of its value since January. The 20% workforce reduction — 620 to 630 employees — was announced alongside this "new vision." CEO statements about adapting the company to the future accompanied restructuring charges of $45 to $55 million. And beneath it all sits a technical and financial truth that most coverage has missed: the unit economics of this pivot are fundamentally hostile to the company's historical gross margin profile. Let's do the arithmetic. Traditional SaaS delivers a 75% to 85% gross margin because software distribution has near-zero marginal cost. AI credits delete that advantage. Every credit burned represents an inference on someone else's GPU, routed through someone else's API. Based on my experience auditing tokenomics during the 2021 bull run, the model costs alone likely consume 30% to 50% of the credit price. That's not a theoretical concern. That's a structural reality of reselling third-party model inference without owning the underlying compute. The result: the more customers use this AI platform, the lower the blended gross margin becomes. That's not hyperbole. That happens on every single transaction. Consider the monthly billing path. Why does a monthly credit cost $0.0125 while annual prepayment drops it to $0.01? That's not customer friendliness. That's a 25% premium designed to pull cash forward. It's the same playbook cloud providers have run for a decade — prepay for certainty. The difference: cloud providers own the infrastructure generating the consumption. Monday.com is a reseller of other people's models. The "prepaid discount" here isn't subsidizing server costs. It's an attempt to stabilize revenue recognition and protect cash flow during a chaotic transition. That's not innovation. That's a bridge. The deeper engineering challenge compounds the problem. Billing for AI credits requires a real-time metering infrastructure indistinguishable from a lightweight cloud billing platform. You need to track token consumption per agent, compute usage, tool invocation frequency, data throughput — then map all of it to a billable unit. Building this system is not a marketing exercise. It's a multi-quarter engineering investment that explains why 620 people had to go. They're buying time to build the meter before the meter can start generating meaningful revenue. The pricing model has a known fragility. Enterprises hate unpredictable costs. Per-seat pricing answers a simple question: how many employees? Credit pricing demands a sales team explain "how many credits will your AI agent consume completing 100 tasks?" Every sales representative now needs to do value-based selling with limited historical data. The sales cycle doesn't just double. It can stretch from weeks to months. That's not an opinion. That's what happens when you remove predictability from procurement. The hidden assumption underneath this entire strategy is that existing customers will migrate without friction. But the installed base — those 225,000+ enterprise clients — didn't sign up for a consumption-based metered system. They signed up for flat-fee software. The transition from "unlimited-ish subscription" to "pay for what you use" is one of the most sensitive pricing changes a SaaS company can execute. Existing customers might downgrade to Basic plans to minimize credit exposure. If that happens, the core ARR base shrinks while the company tries to sell a new, more complicated product to new customers who don't have legacy expectations. Customers aren't the only constituency facing pain. Sales compensation becomes significantly harder to calculate. Quotas defined in dollars now depend on the variable consumption patterns of clients. Forecasting revenue becomes dependent on customer AI usage habits rather than headcount additions. The entire internal planning apparatus of a classic SaaS company — sales targets, customer success staffing, financial modeling — was built for seats. It's now being retrofitted for tokens. Now let's talk about what nobody wants to address: the platform's data strategy. AI agents require access to enterprise workflows to be useful. The most valuable asset any AI work platform can accumulate is the training signal embedded in how an enterprise's 30 different agents optimize its customer service pipeline. That's the moat. That's also the liability. Enterprise customers are terrified of their proprietary workflow data being used to train external models. Week after week, by default, the platform transmits sensitive business logic to third-party APIs. If Monday.com cannot secure zero-retention agreements with its model providers, adoption will stall — not because the technology fails, but because trust is absent. As a consequence, low-risk, low-value tasks get automated while high-value, sensitive processes stay manual, keeping credit consumption — and revenue — artificially low. The contrarian angle here cuts against the entire narrative of "AI transforms SaaS." The traditional counterargument — that AI-native startups will eat Monday.com's lunch — misses the actual threat. The most pernicious dynamic is what I call the AI Efficiency Paradox: as the AI agents get better at their jobs, customers need fewer credits to accomplish the same amount of work. In traditional SaaS, a better product leads to more usage and more seats. In a metered AI model, a better product means fewer credits consumed. The second-order consequence: revenue declines as the product improves. This is not a hypothetical. This is the fundamental difference between software economics and utility economics. The forward-looking metrics that matter have changed completely. Forget DAU/MAU ratios. The new KPIs are credit burn rate, agent activation rate, and agent retention. These metrics don't just measure user engagement. They measure the unit economics of the new business. But there's a critical measurement problem on the financial side: Does AI credit revenue count as ARR? Customers prepay for credits at a discount, and the credits might never be consumed. Under traditional SaaS accounting conventions, deferred revenue creates a liability. The biggest risk to the 19%—20% revenue growth guidance is whether the company can make its booking numbers reflect real, consumable value rather than hoarded prepaid credits. Let's evaluate the competitive landscape honestly. The model connectors are a defensive maneuver, not an offensive one. Being model-neutral sounds empowering in a keynote, but the long-term risk is that upstream model providers — OpenAI, Anthropic, even Microsoft — build their own enterprise agent orchestration layers. Microsoft Teams and Project are already competitors. Copilot has an unstoppable advantage in distribution depth. If Microsoft chooses to bundle agent workflow capabilities into its existing enterprise subscriptions, the middle-man value of "AI connectors" gets squeezed from both directions: upstream integration and downstream replacement. I've spent 18 years analyzing platform transitions. The pattern is always the same: the company that is most vulnerable is never the one with the worst technology. It's the one whose revenue model for the old world doesn't map to the new world. Monday.com's legacy seat-based revenue is a liability in a world where the value is in autonomous execution, not user headcount. The organization signaled its own understanding through the layoffs. A 20% reduction isn't a cost-cutting measure. It's a restructuring of the talent base. The people who built the legacy Work OS features are no longer strategically aligned with the company's future. That's a bet, not a plan. A plan would be new talent hired before old talent leaves. Instead, the company is racing to reclaim engineering capacity while simultaneously upgrading its technical foundation to support agent runtime, multi-tenant model isolation, and real-time usage metering. The technology comparison here is closer to cloud infrastructure providers than collaborative work tools. The entire product architecture needs to shift from a System of Record — a place to store and organize work — to a System of Action — a place where AI agents execute the work. This transition isn't just about adding a new feature. It requires re-architecting the permission model, re-engineering the workflow engine to support autonomous decision-making, and building an observability stack that lets non-technical users understand why an agent took a particular action. That's a multi-quarter delivery gap. During that window, customer experience fractures between the legacy features (still being maintained) and the new AI version (requiring new training, new documentation, and new support). The functional vacuum this creates is precisely what competitors will exploit. Customer success teams just shrank by 20%. At the exact moment when the product needs more education and support, the company's capacity to provide it has been reduced. Here is the other uncomfortable signal: 66% of the $45—55 million restructuring charges in Q2 are standard market practice for severance and real estate. But that money buys time, not revenue. The company is spending heavily now to prepare for a revenue stream that — even in the best case — generates lower margins than the old model. The bear call isn't that Monday.com fails. It's that Monday.com succeeds at becoming a lower-margin utility business instead of a higher-margin software business. The original bull thesis on the stock was the high gross margin, predictable subscription revenue. This pivot trades away both of those attributes. The story now is a potential multi-year growth premium. But if the AI-related revenue share stays below 10% for the next two years while margin compression from credits accelerates, the stock gets caught in an uncomfortable position: neither a classic SaaS valuation nor an AI infrastructure valuation. The nearest red flag to watch comes with the next earnings report. Management reiterated growth guidance of 19% to 20%, but that guidance is only credible if they provide clear disclosure on three things: when credit revenue is recognized — at purchase or collection — how unused credits are treated as liabilities, and the actual gross margin of AI credit revenue compared to subscription revenue. The market was right to bounce the stock on the announcement. But the bounce isn't pricing in opportunity. It's pricing in narrative relief. The real question is whether Monday.com can convert this expensive, risky pivot into a business model that doesn't require falling back on engineering excellence alone to make up for weaker unit economics. Here's what I'll tell you directly: cheap software isn't a revolution. Software that makes you money is. The AI integration is a revolution. The pricing could be a confession. The metered consumption model is the new standard, but it's also the architect of margin compression. The transition from pure subscription to utility pricing is the most significant change in SaaS unit economics this decade. Whether it's a triumph or a disaster depends on how the company handles the tension between selling innovation and managing the infrastructure that makes it possible. The stock bounce is a pause, not a pivot. The 20,000 — no, the 225,000 — paying customers are the collateral. Executives will call it transformation. I call it a calculation. The market has yet to see the bill. The broader lesson isn't about Monday.com specifically — it's about what happens when the industry's darling business model starts cannibalizing its own financial foundation. The next 12 months reveal whether this was a bridge to a new S-curve or a step off a cliff. The tools that were once sold as an operating system for work now behave like a metered casino — every move costs a chip. Understood. Now watch the profit margins.