For most of the known SaaS era, gross margin has operated within a fairly predictable range, typically above 80%. Revenue was subscription-based and largely seat-driven, cost structures were well understood, and finance teams could orient around established benchmarks with reasonable confidence.
AI adoption is disrupting that predictability in ways that impact pricing models, cost and unit economics, as well as the metrics investors use to evaluate a company’s health. This requires a new level of sophistication to manage proactively and build a sustainable business model.
Companies that work through these implications early gain a competitive advantage in this rapidly shifting market landscape.
COGS in an AI Product Company
For SaaS companies, cost of goods sold (COGS) was historically well understood, with established classification practices and abundant benchmarking data that gave finance teams a reliable foundation for reporting and decision-making.
Now, the costs associated with building and delivering an AI-enabled product increase dynamically, materially altering the economics of the business in ways that existing cost categories were not designed to capture. Companies that don’t develop deliberate policies around cost classification early tend to end up with margins that don’t accurately reflect what the business actually costs to run.
Cloud infrastructure is a common starting point. Most companies today run their full technology environment — development, QA, testing, and customer-facing production — on platforms like AWS or Google Cloud, which means a single vendor relationship encompasses costs that belong in multiple categories. Internal development and testing environments are operating expenses; the infrastructure supporting the customer production environment is cost of sales.
The same principle applies to DevOps: personnel and systems dedicated to running and maintaining the platform customers access are delivering revenue-generating products, and should be classified accordingly.
Technical Support and Customer Success present a similar classification challenge, compounded by the fact that these functions can often be defined differently across organizations in earlier stages.
The relevant principle is what the work actually involves: time and resources spent directly supporting and serving customers in the delivery of the product belong in COGS; account management, business reviews, and activities oriented toward retention or expansion are sales expenses. Getting that boundary right matters because misclassification here distorts both gross margin and the unit economics that inform business decisions.
AI costs introduce a genuinely new category, including AI inference costs and associated costs such as infrastructure, monitoring, observability, and support costs. Companies need to be deliberate about separating what is consumed in customer-facing product delivery from what is used internally for development, research, or testing — and that separation requires intentional systems and organizational awareness to work reliably.
Without visibility into how AI resources are being consumed across the organization, usage tends to expand without clear attribution, and the cost base becomes difficult to manage or defend. Building visibility early and embedding cost-awareness into how engineering and product teams operate is a prerequisite for understanding what AI actually costs the business to deliver.
Pricing for an AI Cost Structure
The COGS challenge is compounded by a structural tension in how most SaaS companies generate revenue. Traditional subscription pricing — whether seat-based or tied to a fixed tier — was designed for a cost structure where the marginal cost of serving an additional user was relatively low and stable. AI changes that equation in ways that have significant implications for gross margin. When inference and delivery costs scale with usage, the relationship between customer engagement and profitability becomes complicated: higher-usage customers may generate the most cost without a corresponding increase in measurable value delivered. Inference costs, in other words, are not a reliable proxy for outcomes.
Navigating that disconnect requires a deliberate response at the pricing and packaging level, as well as new analytical frameworks and metrics — Salesforce’s introduction of “agentic work units” as a pricing concept is an early indication of how the industry is beginning to work through this. Many companies have not yet fully adapted to these dynamics, and the ones that engage with them early will be better positioned to build a model that holds up under scale.
The response taking shape across the industry is a shift toward hybrid models that combine a base subscription with usage-based or outcome-based components, allowing revenue to scale in closer proportion to the value being delivered and the costs being incurred. Structuring those models well requires finance to be involved in product and packaging decisions early — with sufficient understanding of the cost architecture to help design something the business can sustain as AI usage grows and the underlying cost landscape continues to shift.
That structural change also has implications for how investors evaluate gross margin. High gross margins have historically been treated as a signal of SaaS efficiency and scalability, but in the context of AI-enabled products, they can also raise questions about the depth of AI integration. An AI product presenting margins in the 80 to 90 percent range may prompt investors to ask whether AI is central to the product’s delivery or largely peripheral — a distinction that carries weight in how a company’s long-term positioning is assessed. A credible account of how margins will evolve as usage scales and model costs optimize tends to be a more durable part of the investor narrative than a static margin figure.
Gross Margin Is One Variable
Gross margin is an important indicator of business health, but it is one component of a broader economic picture. The equation that ultimately matters is EBITDA and free cash flow — what the business generates after the full cost structure is accounted for. AI is reshaping gross margin in ways that require active management, but it is also creating meaningful opportunities to reduce operating expenses across the business. Companies that approach the COGS transition with discipline while using AI to improve operational efficiency are working toward a sustainable economic model, and those two efforts are most effective when they are coordinated rather than addressed in isolation.
What that coordination requires from finance is a level of operational involvement that goes beyond reporting. Understanding how the product is delivered, where costs are accumulating, how the pricing model is evolving, and what the business needs to look like at the next stage of growth — these are the inputs that make financial guidance useful at this moment.
The established SaaS playbook provided a relatively stable framework for answering those questions. What has replaced it is more complex and more company-specific, which means the finance function’s ability to engage with the business as a genuine operational partner has become more consequential, not less.
Attivo Partners works with venture-backed companies to build the financial infrastructure needed to scale with confidence. If your business is navigating the shift to AI-enabled products, our fractional CFO team can help you think through the cost architecture, pricing model, and investor narrative that fit where you are and where you’re going.
