A nasty shock is hitting finance leaders across every industry right now: AI token bills that run ten, twenty, even a hundred times over what they forecasted, blowing holes straight through quarterly budgets. These leaders are all asking the same questions: How could this happen if they didn't approve it? Why didn't any of their systems alert them to the spike? And most importantly, what can they do now?
Yes, your company’s leaders told your engineering team to use AI. They told everyone to use AI, for everything. Build faster, ship more, and become "AI-native." The workforce did exactly that, and somewhere over the past three months, a few teams multiplied their token usage, a default model got swapped for a pricier frontier one, and a prepaid balance meant to last the year was gone by the first quarter.
This is what happens when tokenmaxxing catches up to you. For the past two years, AI tools have largely operated on an unspoken unlimited plan: experiment freely, burn tokens, figure out ROI later. That's starting to change, because the bill is coming due in a way traditional software never required.
AI tools don't behave like the SaaS apps that came before them. A traditional app is priced per seat, so your headcount tells you your bill. AI is increasingly priced by consumption: every prompt, model call, automated workflow, and autonomous agent, all add to the meter.
This leaves IT, finance, and AI program leaders asking three questions they often can't answer with any confidence: How much are we spending on AI? Who is driving the cost? How can I make my runway last?
AI spend is uniquely difficult to see and control, in ways that even seasoned procurement and FinOps teams haven't had to manage before.
Usage compounds fast and often silently. A vendor can quietly shift your default model to a more expensive tier in the middle of a billing cycle, and unless someone happens to notice, every request from that point on costs more with no change in behavior on your end. A coding agent left unsupervised overnight can burn through a week's worth of tokens on a single task it got stuck looping on, or on work nobody actually needed done. Add in a team ramping a new use case, or a wave of agentic workflows running with no one watching, and a monthly burn rate can double before anyone thinks to check. Token-based pricing means spend accumulates in real time, not just at renewal, so the damage is done long before an invoice surfaces it.
It's also tough to get a holistic look at AI spend, since every AI vendor has its own billing model, its own metrics, and its own dashboard. Getting a credible total means logging into each portal separately, pulling CSV exports, reconciling mismatched data, and maintaining a spreadsheet that's stale the moment finance asks about it.
Accountability is just as fragmented as visibility, since AI spend lives in the gap between IT, engineering, and finance. IT sees the SaaS landscape, while Finance sees invoices, often after the money is already committed. Neither sees the full picture, and no single team owns a reliable answer.
The AI spend conversation often overlooks something else: many of these AI tools were never sanctioned in the first place. According to 1Password's Access-Trust Gap Report, over a quarter of knowledge workers use AI-based applications their employer didn't approve. Some of that is personal experimentation with no cost to the company. But some of it does hit the corporate bill: an engineer spinning up an API key on a shared account, a team expensing a subscription, someone provisioning seats in a paid workspace, all without going through finance. Either way, IT has no record of these tools, no visibility into what data enters the AI's context window, and no way to revoke access when the person moves on.
Clearly, the need for visibility and governance over AI spend is as urgent as this month’s bills.
Today we're announcing the Public Preview of AI Spend and Consumption Management in 1Password SaaS Manager, available to every SaaS Manager customer.
Recently recognized as a Leader in the 2026 Gartner® Magic Quadrant™ for SaaS Management Platforms, SaaS Manager unifies AI and SaaS discovery, access governance, and spend optimization across more than 400 integrations to help organizations see and govern their full software portfolio. AI Spend and Consumption Management brings that same foundation to the fastest-growing, least-understood category of software spend: AI token consumption. At launch, it supports Cursor, Anthropic (Claude), and OpenAI (ChatGPT and Platform), with more vendors planned.
Setting up AI Spend and Consumption Management is simple. You connect your AI vendors' admin API keys, data syncs daily, and you get a normalized view without custom engineering or agents to deploy. With AI Spend and Consumption Management, teams can:
**See AI consumption in one place. **A single AI dashboard shows consumption and spend across Cursor, Anthropic, and OpenAI, with click-through to app-level detail.
Stay ahead of budget risk. Set budgets per vendor, configure budget thresholds by percentage, and track daily burn rate so you know the estimated exhaustion date for prepaid balances.
Understand what's driving cost. Break AI spend down by vendor, team, user, API key and model, so you can see exactly where consumption is concentrated and where optimization opportunities exist.
**Get alerted automatically. **Over-budget, depletion-risk, and data-gap notifications arrive in Slack and email, so nobody has to remember to check multiple dashboards.
**Optimize model selection. **Attribution data shows you when everyone defaults to the newest, most expensive model for work a cheaper one could handle.
**Rebalance the broader software portfolio. **With AI consumption visible alongside the rest of the SaaS portfolio, finance and IT can see how AI spend is growing relative to traditional SaaS, identify redundancy, and make confident consolidation decisions at renewal time.
Attribution by model and user does more than catch overruns. It lets you manage both sides of the equation: feed more capacity to the projects actually delivering value, while reining in the experiments quietly wasting tokens.
Consumption-based pricing has two edges, and most conversations only look at one of them. The same lack of visibility that lets costs spike unnoticed also means most teams have no idea whether they're getting good value for the dollars they're already spending.
Model choice is a good example: output costs across leading models today vary by roughly 300x for comparable work, from a fraction of a cent to well over a hundred dollars per million tokens. Most of that gap comes from defaults, not better outcomes. When a team reaches for the newest, most expensive model for a task a cheaper one could handle just as well, the tool is being used for exactly the right reason (faster development, better output) in exactly the wrong way. One engineer doing this barely registers. A few hundred engineers doing it every day, on every prompt, is how good intentions turn into a real line item.
The financial exposure is real and specific:
**Consumption-based spend wrecks forecasts: **AI consumption pricing produces swings that are hard to predict and harder to defend. For a public company managing earnings guidance, an unbudgeted AI overrun is a major forecasting and reporting problem.
**Pre-committed purchases come with use-it-or-lose-it terms: **Many organizations buy AI capacity in prepaid tranches. Run out early and you're renegotiating mid-contract or absorbing unplanned overages; run out late and unused commitment expires unrecovered.
**AI is reshaping the whole software budget; **As AI spend grows, it can compete with the rest of your portfolio. Organizations that can't see how AI consumption sits within the broader software portfolio can't make confident decisions about where to cut, consolidate, or double down on their SaaS usage.
The category forming around AI spend is full of partial answers. Finance-first tools capture cost only after it's committed, and only when it shows up on a corporate card. Spend-and-procurement platforms track invoices but operate at a distance from usage. Neither connects AI consumption to the discovery layer that tells you what's actually running in your environment.
1Password SaaS Manager closes those gaps. It continuously discovers AI applications employees are using across identity providers, SSO logs, finance systems, device agents, browser extensions, and 1Password vaults. The 1Password browser extension detects OAuth credential grants in real time, surfacing new AI tools to IT before they become governance gaps or offboarding problems. (When Flipdish connected SaaS Manager to their identity provider and finance systems, they discovered more than 1,000 applications in under five minutes.)
Discovered apps are matched against a library of 40,000+ pre-populated profiles to surface immediate risk context and compliance posture. So you don't just know how much you're spending on AI; you know which AI tools are active in your environment, whether they were approved, who has access to them, and what happens to that access when someone changes roles or leaves.
AI Spend and Consumption Management is an integrated extension of a platform that already governs your broader software portfolio, not another silo to stitch in. And it's available to existing SaaS Manager customers at no additional cost, because this kind of visibility is quickly becoming table stakes, not a premium add-on.
AI Spend and Consumption Management is now available in Public Preview for all SaaS Manager customers, and will be broadly available in Fall 2026. As a preview, it's production-close and built for real use, with room to grow as we incorporate your feedback. If you use SaaS Manager today, you can connect your AI vendors and start tracking consumption now. If you'd like a guided walkthrough, reach out to your Customer Success Manager to set up time with us.
The trial period of uncontrolled token spending is ending. The organizations that build AI governance now, while the category is still forming, will be the ones that can scale AI confidently instead of capping it in a panic mid-year.
__Want to learn more? [Read the press release](https://1password.com/press/2026/july/1password-introduces-ai-spend-and-consumption-management).__
__Want to get started managing AI spend with 1Password SaaS Manager? [Head here.](https://1password.com/solutions/ai-spend-management)__