1Password, long known for its leadership in identity and password security, is extending its platform to tackle one of the fastest-growing challenges in enterprise tech: managing the unpredictable costs and risks of AI usage. As part of the launch, I had the opportunity to dive into a Q&A with Sanjay Ramnath, VP at 1Password, on how the new AI Spend and Consumption Management capability fits into their broader vision.
Here’s the full conversation:
1. 1Password’s core strength is identity security for both humans and now AI agents. How does adding spend and consumption visibility fit into that broader platform vision? Is this the start of treating AI usage as another identity and access management challenge?
For more than 20 years, 1Password has grown with how our customers work. We started with passwords, and then expanded to secrets management, access control, and visibility into the applications teams rely on. SaaS Manager followed that same path by giving IT, finance, and security teams a clear picture of the software running across the business, who has access to it, and what it costs.
AI tools are now part of that software landscape. They come with different pricing models and different risks, but the underlying questions are the same ones we have been helping customers answer for years. Which tools do people have access to? Is that access appropriate? What is it costing the business? AI Spend and Consumption Management applies that framework to a layer that has been difficult to see clearly until now. For our customers, this is not a new problem. It is a familiar one showing up in a new place.
2. For teams that live inside Cursor, Claude, and GPT every day, what kind of “I had no idea we were spending that much on X” moments have you already seen in the public preview? Any early patterns around model choice, team behavior, or shadow usage that surprised customers?
It’s still early in the preview, but the consistent theme is that AI spend is becoming a real management problem for IT and finance teams. The size of the spend is part of the challenge, but what customers talk about most is how difficult it is to forecast. Unlike traditional SaaS, AI costs move with every prompt, and usage is spread across multiple tools, teams, and models. Most organizations have moved past the debate over whether to adopt AI. Now they are trying to understand what it is costing and whether that investment is paying off. With Goldman Sachs estimating token consumption from AI agents will grow 24 times by 2030, many customers are trying to build better visibility before AI spend becomes much harder to manage.
3. The feature lets teams get configurable alerts before prepaid balances run out. How granular can those controls actually get today — and how close are you to letting teams set different thresholds per model, per project, or even per type of usage pattern?
Per-vendor, model, and project-level spend management is on our roadmap. AI Spend and Consumption Management doesn’t just report total spend. It breaks down consumption by vendor, model, team, and individual user. That granularity lets IT and finance see when a specific team or user is driving a disproportionate share of consumption. From there, a CFO or CIO can look at what that team is building and ask whether the spend is earning its cost or needs to be optimized. AI spend decisions should not be made on usage numbers alone. The data needs to support a harder question: why is that usage happening, and is it delivering value?
4. Many organizations are still in heavy experimentation mode with AI. How should companies balance giving developers the freedom to move fast with the need to avoid runaway costs? Does this tool help create that balance, or does it risk becoming another layer of governance that slows teams down?
Visibility and speed are not in conflict. The problem is what happens without visibility. When organizations lack insight into AI consumption, they tend to make reactive cuts when costs spike without warning. Initiatives get shut down without anyone understanding what was driving the cost or what value it was producing. AI Spend and Consumption Management isn’t designed to add governance gates or slow teams down. It gives IT and finance the data to make targeted decisions: which teams are consuming the most, what they are building, and whether that usage is delivering business value. Organizations make better investment decisions when they understand both what they’re spending and what they’re getting in return.
5. You talked about helping organizations “understand where model usage is creating value.” Beyond tracking spend, what additional signals or frameworks do you think companies should combine with this data to actually measure whether their AI investments are paying off?
Spend data alone is not enough. To understand whether an AI investment is paying off, that data needs to connect to business outcomes. The challenge is that value looks different across every organization. There is no universal signal or framework. What matters depends entirely on what each team is using AI to accomplish. That is why the data works best as a starting point for internal discovery, not a final answer. Teams that ask what their consumption is actually producing, and whether the usage patterns they are seeing reflect meaningful output, are the ones who will make better decisions about where to direct AI investment and where to pull back.
6. As AI agents start making autonomous tool calls and consuming tokens without a human in the loop, what new spend-management and governance challenges do you anticipate? How might 1Password’s platform need to adapt when the “user” is an agent rather than a person?
AI Spend and Consumption Management captures token consumption at the API level regardless of whether a human or an agent is generating it. If an agent is running on your OpenAI or Anthropic account, that usage shows up in the vendor admin API data and flows into the dashboard, including the spikes that agent loops can create. The challenge organizations might face in the future is around AI agent attribution. When an agent is running autonomously on a shared service account or tied to credentials outside the managed environment, connecting that consumption to a specific team could become difficult.
That is where governance breaks down, and is why we see this as an identity problem as much as a spend problem. As an identity security provider, 1Password connects that spend layer to the access and credential layer across the platform.
This feature feels like a natural evolution for 1Password — extending their identity-first approach into the chaotic world of AI consumption. With token usage exploding and forecasting nearly impossible without proper tools, better visibility could prevent a lot of nasty surprises for IT, finance, and security teams.
Commentary attributed to Sanjay Ramnath, VP at 1Password. Thanks to the 1Password team for the details.
