1Password’s CFO Greg Henry warns token budgets are breaking, launches AI Spend controls

On Tuesday, 1Password launched AI Spend and Consumption Management, a new capability embedded in its SaaS Manager platform, built to solve a problem Greg Henry called out directly: token consumption is moving faster than traditional enterprise budgets were ever designed to manage. Henry, 1Password’s chief financial officer, framed the stakes bluntly for executives: teams want to build faster with AI, but that speed is creating “a new kind of spending pressure,” because developers are consuming tokens at a pace that finance and IT struggle to forecast, justify, and reconcile.

So 1Password built the missing visibility layer. The new feature gives IT and finance a unified, real-time view of how their organization consumes and spends on AI services from vendors including Anthropic, Cursor, and OpenAI. Henry’s core argument is that the cost drivers are not only complex, they are also often invisible until too late. With this release, token-level consumption data is pulled daily via vendor admin APIs, normalized across providers, and surfaced in a single dashboard that connects usage to teams, users, vendors, and models.

The product is not a standalone tool, and it does not arrive with a separate add-on fee. AI Spend and Consumption Management is available to all 1Password SaaS Manager customers, which can activate it by connecting supported AI vendor API keys. For now, it is in public preview, with broad availability planned for fall 2026. That matters because many enterprises are already midstream on AI procurement, and the question is less “do we care?” and more “how do we stop surprise invoices?”

Why token spend is different, and why the old playbook fails, comes down to pricing mechanics. Traditional SaaS is typically priced on a per-seat, per-year basis, which makes budgeting and reconciliation comparatively straightforward. AI pricing does not behave the same way. Every API call to models like Claude, GPT-5.6, or Cursor-powered coding assistants consumes tokens, and the cost varies by model, by input versus output, and by task complexity. Henry’s warning point here is operational: a single engineering team running agentic workflows can burn through a prepaid token budget in weeks, while the finance team may not notice until the invoice arrives.

Henry also made a historical analogy executives will recognize: consumption-based pricing isn’t new. Enterprises saw a similar shift with cloud infrastructure when Amazon Web Services, Microsoft Azure, and Google Cloud popularized consumption-based pricing for compute and storage in the 2010s. That transition took years to manage because organizations initially lacked the tooling and disciplines to monitor and optimize cloud bills. Over time, the FinOps ecosystem emerged, including companies such as CloudHealth, Spot.io, and Apptio, focused on answering basic questions like what you’re spending and why. Henry is betting AI token spend will go through the same arc, and that organizations that do not build visibility now will end up, in his words, “paying far more than they needed to, for far longer than they should have.”

The company points to the likely scale of the demand to justify that bet. Goldman Sachs has estimated that token consumption from AI agents alone will grow 24 times by 2030, driven by the expectation that autonomous AI systems will increasingly execute multi-step workflows, rather than just responding to prompts. If agents are booking travel, writing and deploying code, or managing customer service interactions, the number of API calls can balloon beyond what humans generate in chat-based usage.

Technically, the “AI Spend and Consumption Management” dashboard is built around four capabilities. First, it aggregates token usage and spend across Anthropic, Cursor, and OpenAI into a single normalized view, reducing the need to toggle between multiple vendor dashboards with different reporting formats. Second, it supports budget controls like vendor-level spend limits, percentage-based thresholds, and automated alerts when prepaid balances approach depletion, with alerts configurable via Slack and email. Third, it disaggregates consumption by team, user, vendor, and model so finance and IT can identify not just what got spent, but where the spending came from. Fourth, it places AI spend within the broader SaaS portfolio so teams can see how token costs relate to total software investment.

A particularly important detail is that the system captures token consumption at the API level regardless of whether a human or an AI agent generated it. Henry said organizations get the total consumption picture, including spikes that agent loops can create, which can be among the hardest usage patterns to catch before they become a problem. For now, the product alerts but does not enforce spending cutoffs. When asked whether 1Password will eventually allow automatic enforcement when thresholds are crossed, Henry said the company is “actively evaluating” automatic enforcement, but emphasized the prerequisite: “You can't enforce what you can't see.”

Finally, the launch partner list reveals where 1Password believes budget strain is most intense right now. Henry said the company chose Anthropic, Cursor, and OpenAI based on customer demand, where token consumption can move fast and get ahead of the teams responsible for managing it. Cursor’s inclusion is especially telling because it sits in a workflow where AI output is continuously generated as developers write code, not just on-demand like a chatbot. Henry also acknowledged the ownership question is still messy: when spend is fragmented across vendor dashboards and finance reconciles monthly, you are “always behind,” and AI spend can’t be treated as a finance-only or IT-only problem. In the same way cloud pushed FinOps into existence, token-driven AI spending is now forcing CFOs, IT, product, and engineering leaders into conversations “in ways they never had to before.”