Tech builds on AI. Finance protects the margin.

Tech builds on AI. Finance protects the margin. | Databricks Blog Skip to main content

Ask a tech company CFO where the quarter's margin is landing, and you will get a straight number. Ask what moved it, and the list starts: the popular feature whose compute cost climbed faster than its price, the usage revenue booked across mixed subscription and consumption plans, the reserved compute an automated scaling policy drew down faster than planned. Every number on that list is the product of multiple systems, and every one is increasingly shaped by agents. Finance's job is to see how those variables move unit economics and steer the company continuously, not once a month at the close.

Most finance teams are asked to do that on plumbing built for a slower business: extracts, spreadsheets, and metrics that reconcile monthly while usage, pricing, and compute move hourly. Every day of lag has a price. A repricing lands a week late. A metering bug survives until the close. A compute commitment gets signed on last month's picture of demand. How AI unit economics became finance's front line Tech and AI-native companies are built on growth. Usage scales, pricing spans subscription and consumption, and compute has become the largest variable cost in the business. Protecting the economics of that growth has always been finance's job. What changed is the speed: agents now shape how compute is consumed, how usage is priced, and how revenue is recognized, and the sector's numbers show the squeeze. AI-native gross margins reached about 52 percent in 2026, up from 41 percent in 2024 (ICONIQ), still short of the 70 to 90 percent that classic software earns. The strongest finance organizations are not waiting it out.  OpenAI's finance team describes putting AI agents to work on contract workflows, extracting and reviewing terms that once consumed analyst hours.  YipitData brought revenue operations and finance onto Databricks, where finance analysts write their own SQL and PySpark, and a NetSuite integration took manual effort out of financial reporting. The question for a tech company CFO is no longer whether finance will operate with AI. It is what foundation makes it safe.

Usage and compute scale by the hour. Unit economics are what finance protects as they do.

Why a word like ontology now matters to tech company finance Finance has always been good at finding the number, even when it is buried in complexity. But a number can be perfectly accurate and still not be correct, because it rests on a partial or dated picture of how the business works. What matters is the meaning behind it: the definitions, the products and plans, the usage and compute drivers, and how each of those changes as the business moves. For AI, this gap is the headline problem: inaccuracy is the most-cited issue enterprises report with AI, named by roughly a third of them (McKinsey, 2025). As Ali Ghodsi puts it, most enterprise AI is guessing with false confidence. That is a context problem, not an intelligence problem. An ontology closes the gap. It captures what the numbers mean and keeps that meaning current as the business changes. And the raw material is arriving faster than ever. Stripe payment data , transactions, subscriptions, refunds, and payouts now flow into Unity Catalog through  OpenSharing on  Databricks Marketplace : live, governed, and with no ETL to build or maintain.  Lakebase , the transactional Postgres database built natively on the lakehouse, extends the same principle inside the platform. Operational data and analysis share one foundation, so the answer reflects the business as it runs today, not as it was at the last sync.

Accurate is the right figure. Correct is the same figure, rooted in the product, the usage, and the plan terms.

Where Genie One comes in In an AI-native business, an understanding formed a few hours ago may already be out of date. So the ontology itself has to keep moving. It has to learn from the systems the business runs, sharpen with every question, and adapt as the business evolves. Databricks built Genie One for this. It works as a data-smart AI coworker: a finance leader asks a direct question and gets a trustworthy, sourced answer, grounded in Genie One's ontology and governed at every step. Amagi , the AdTech platform behind thousands of broadcast channels, already runs this way in production. Its finance team gets real-time billing and financial reporting on Databricks, with Genie One answering natural language questions, and because finance, marketing, and operations draw on the same governed data, executive meetings no longer relitigate whose number is right. Every AI-native finance team is working some version of three questions, and each answer sets up the next: ›  Where is real gross margin landing, by product and customer, once AI compute is in the mix? The price a plan carries and the margin it keeps are rarely the same number once inference and serving costs are counted. Margin is where most teams point Genie One first, and it is why design software, marketplace, and consumer internet companies are standing it up as the self-service entry point for their finance teams. ›  Where is consumption revenue at risk across mixed subscription and usage pricing? Consumption you can meter correctly is revenue you can recognize, and the value is in catching a gap as it forms, not at the close. A digital wagering company has taken this the furthest: it replaced fragile spreadsheet forecasting with governed revenue forecasting models, added liquidity analysis, and gave its leadership team a Genie One of its own. ›  Where is compute spend at risk of outpacing the runway as usage grows? The time to act on burn is before the next compute commitment is signed. A digital asset platform runs its financial cash pipelines and blockchain analytics on the platform for this reason, and a global delivery platform runs finance forecasting, finance apps, and workflow automation the same way.

Three questions, three outcomes, one mechanism. Genie learns the business, sharpens with every question, and shows its work.

And because every figure traces to its source, every permission holds, and the cost of the AI itself stays governed under one model, the answer is one finance can act on. Genie One readies the move, whether that is a repricing, a metering fix, or a slower drawdown, and a person in the loop makes the call. From first answer to finance...