Stop Flying Blind on AI Consumption

Hey, everyone. Welcome back here to Techstrong TV. I'm really happy to have my next guest back with us.

The last time I checked in was at RSA time back in March. He is Sanjay Ramnath. Sanjay is the VP at 1Password, and Sanjay, welcome back to Techstrong TV.

How you been? I'm doing well, Alan. Thank you so much for having me back.

All right. tv or on the OTT app. You can go back, but for those who weren't lucky enough, Sanjay, give us a little bit of your background and how you came to be VP over at 1Password.

Yeah. So I run product and solutions marketing at 1Password. I've spent the last couple of decades or so in product roles across early to mid-stage, late-stage startups.

Been through a few acquisitions, through a pretty broad spectrum of the cybersecurity space, seeing things evolve rapidly. Prior to 1Password, I spent a few years in the application security space and what excited me about 1Password was, at the dawn of AI, there's a clear realization that the identity security stack needs to evolve. Exciting new problems, exciting new challenges, and really an opportunity to help organizations transform with powerful new technology brought about by agents and AI tools.

And 1Password was uniquely positioned to solve some of these problems, and it's an opportunity I couldn't miss. So that's why I'm here. I love it.

Thank you for that. Congratulations on, it's a great story. So look, I think I said this before, we're actually customers of 1Password here at Techstrong.

We use it and happy customers, I might add. But a lot of people think of 1Password as a one-trick pony. It's a great place to store your passwords.

Mm-hmm. Right? A password manager.

But of course, there's more to 1Password than being a password manager. Right? Everybody wants to be a platform.

No one wants to just have features. But 1Password's a company that actually kind of lives that, right? You do a lot of security research a lot.

My friend Dave Lewis, my friend Wendy Nader- Mm-hmm ... out there evangelizing an awful lot. But Sanjay, you guys recently released a whole new, well, to me anyway, kind of a whole new layer of functionality here around AI spend and consumption management, which is of course, on everyone's radar today, right?

With the token maxing and all that. Talk to us about that, if you don't mind. Yeah.

Before telling you what we built, I think the context is important, right? So almost every organization today is pushing AI adoption hard. And employees are rising to the occasion, right?

But AI pricing is fundamentally different from traditional software or SaaS pricing. You're not paying for seats and users anymore. You're paying per query, you're paying by models, you're paying by tokens.

And this new cost model moves very fast. It's quite dynamic and unpredictable, if I may say so. So the tools that IT and finance teams put in place to live in the world where things were a little more regulated and predictable and linear, if you will, they just don't work in the world where you don't have this fixed cost renewal in place.

It has to shift towards real-time consumption, and that's the gap. That's where we saw problems emerging. And AI spend and consumption management, the capability we just built into our SaaS manager product, is our answer to that gap.

We are giving IT and finance teams a shared view, real-time view, of what is being consumed, who is consuming it, how much is it costing, and give them that forecasting ability before cost becomes a surprise. We want to make sure that you don't get hit with this large unpredictable bill, if you will. So that's what we chose to solve for with AI spend and consumption management in the SaaS management product.

I love it. But does this move 1Password into sort of FinOps? Yeah, I think there is a natural convergence already between what was traditionally considered the role of IT, and the role of finance, if you will, right?

IT has the usage context, but they don't have the cost context. Finance has the cost context but doesn't have the usage context. And as AI tools proliferate, I think, just like DevOps, DevSecOps emerged as these converged categories in the world of IT, especially around consumption and spend, there is going to be a natural convergence between what IT teams require and what finance teams require.

So let's say IT FinOps is going to be a real thing. And, they both need the same visibility. They need shared visibility, they need visibility that's consistent, that is in one place.

And then they can make much more smarter, informed decisions around where to invest, how to invest, making sure they're getting the maximum return for that investment as well. So I come from a security background too, Sanjay, right? And one of the things we used to talk about in security back in the day was this notion of actionable intelligence.

Mm-hmm. And with the advent of AI I think the cheese has moved a little bit where it's not enough just to have actionable intelligence, but can I actually act? Mm-hmm.

Right. I think that's a great metaphor for generative AI to agentic AI. Sure.

Right? So I guess my question to you then is, in helping organizations optimize AI spend, are you just- Mm-hmm ... advising me what to do?

Or can I say, "Look, I'm going to put Sanjay on a hard limit of, I don't know, 10,000 tokens or $100 a month," or whatever the number is. Let him know when he's within 10%, 5%, and when he gets there, maybe I got to cut him down. Yeah.

I think those are natural next steps, right? The first step, and this is where I think most organizations are today, and what we're hearing from the customers that have adopted this tool, is to get that visibility, is to understand, what am I buying? Why am I buying it?

Is it delivering value? And then making sure that-- High consumption by itself may not be a bad thing. There could be a team that's consuming a large number of tokens and delivering a ton of return on that.

There's also a possibility that the reason why the tokens are being burnt at a fast rate is because there is an AI loop that has a misconfigured API call. So getting that visibility, being able to make that distinction, and really optimize usage is the first step. I think you can take it one step further and talk about limiting consumption or having thresholds and things like that.

But I don't think that's really the problem to solve right now. That could be a problem to solve once the optimization is done. The first step is who's doing what, where is the spend, and also being able to calculate burn rates and see if a threshold is being met, get that alert early.

So either you increase the threshold or optimize for how the usage is actually creeping towards that threshold. And then at some point in time, you'll see natural guardrails and natural governance mechanisms, both through process and through technology, emerging as a result. Fair enough.

I agree with you. And we were talking off camera about how AI time seems to speed everything up, right? Yeah.

And I think sometimes we forget things like this, tracking AI spend and consumption, moving it from a concept to a product, from a product to enforcement. Mm-hmm. That's a process that you can't-- And you don't want to be cutting people off from using their AI prematurely either, right?

Yep. It's important you don't. So I think that's something we tend to lose track with.

Just because we may want to go fast doesn't mean we should necessarily go that fast. We got to give these things a chance to-- How do they get internalized? How do we- Yeah ...

get used to them before we move on? I think in the AI world, speed is a given, right? Organizations are going to move fast.

That is a given. They should move fast. Organizations must innovate.

It's an incredible piece of technology that unlocks potential that we've never seen before. They are going to spend more doing that. So that is not the question, right?

So the question really is, do you have the tools to understand how that pace, that speed, that innovation is actually moving? And whether it's actually delivering value. So the companies that will win will retain that acceleration, retain the innovation, retain the speed, and continue spending, but then in parallel, figure out ways to measure without getting in the way of that speed, without getting in the way of innovation.

Figure out ways to govern, figure out ways to optimize so you can move faster, but you move faster and smarter at the same time without flying blind. So that shift from unpredictable speed to more visible and optimized speed is really what we're trying to enable through spend and consumption management. Fair enough.

So I get all of this information that you guys are collecting now. Where does it live? How do I put it somewhere where I could share it with the rest of the team?

You know what I mean, the other piece of this. Yeah. So there's two parts to this, right?

First, the spend and consumption management capability is part of the SaaS manager product. So you get one place where you can govern and optimize your SaaS applications, and now your AI applications as well. So you get the complete picture from one tool.

And then beyond that, we have integrations, both with the ITSM integrations, we have integrations with SIM-type platforms. So there are ways to distribute that data, distribute that information to the rest of the organization as well, depending on who's looking for it. Like for example, you can create reports for finance teams, you can integrate with platforms that your operations teams might use, your IT teams might use.

So there's multiple ways to do this. Agreed. I agree with you.

Sanjay, does this foretell an enhanced or broadened scope for 1Password? Yeah. Look, we started our journey securing digital credentials and digital identities for millions and millions of new users.

8 8 million credentials in our vault. Wow! " So we extended our platform to include application governance.

Then the next step was customers coming to us and saying, "Hey, now I have AI tools in the mix. I have AI agents in the mix. " And we've built that out, so we have ability to integrate our vault with both machine and agentic workflow, so we can provision credentials and secrets in a highly secure, just-in-time manner.

We just made an acquisition recently of a company called Epono, that allows us to also provide zero standing privilege access to cloud, to Kubernetes, and to infrastructure environments. So when a user is accessing critical infrastructure, they don't have any standing privileges that might become a breach vector. So that's the 1Password story.

We are in the business of making things simple for customers and building the identity and access layer for humans, for machines, and for agents on one platform in one place. And all the work we've done so far, all the products that we've built, and all the capabilities we're integrating so far are really aimed at making things easy for the customer to adopt AI safely, securely, and in a smart manner. I love it.

Good stuff. Sanjay, I forgot to ask this upfront, I should have. This AI spending consumption management functionality is available now?

Yes, we just released this capability as a beta. We have a number of customers that any SaaS manager customer has access to the capability. We have hundreds of customers that have already adopted it and are using it.

We're getting a lot of feedback from them. And we will act on that feedback and continue improving things. We'll add to what we've built so far.

But yes, if you're a SaaS manager customer, you can use the capability today. Excellent. Hey, I want to thank you for coming on here and informing our audience about this.

This is exciting. And I think as we said in the outset, it's timely, right? Everybody's worried today about, "How much am I spending on this AI?

What's my token consumption? " Though I think that's the red herring. It's not a question of cheaper having a person doing it.

A person's going to be involved anyway. I think how much more effective is the person with this tool than without it? And that's the economic decision that you should make based upon consumption management.

Absolutely. But that's me. What do I know, right?

No, absolutely. I agree with that. Yeah.

Absolutely. It's an enabler, not a control. Yeah.

Sanjay, 1Password will be out at Black Hat, I believe, yeah? Yes. We will be at Black Hat.

We will be showcasing all our products. We have an amazing booth, amazing bunch of people that are just ready to meet you all, so come by. Meet you there.

All right. Sanjay Ramnath, VP of 1Password here on Techstrong TV. We're going to take a break.

We'll be back with more in just a minute.