Raju Malhotra

I am a chief product and technology officer at Certinia.

Certinia helps professional services companies sell, deliver, and manage their business. So what I mean by this is they're a professional services company like IBM, PwC, KPMG, Pure Play, professional services companies. And then there are companies that are professional services organizations within a technology company. Salesforce, Google, AWS, Philips, Cisco, HPE. They have consulting organizations, professional services. In both of those cases, we really help our customers B2B really sell those services. Because selling services is different than selling software or selling widgets or manufacturing. You have to be able to find out who's available, who's not available, which time zone, etc., in terms of people. And then deliver those project manage, resource manage, etc., and then make sure we have ongoing relationships that we can help our customers establish with their customers. So long story short, really, it's about professional services automation that we provide to our customers.

It is a full stack. Interestingly, we work with the CRM, ERP, HCM, human resource systems with them. Think of the Certinia as the industry specific solution for the professional services.

Yeah. So the companies I mentioned, they're all our customers. IBM, PwC, KPMG, of the world.

You said it better than I could have said. Professional services is the most disruptible because AI is all about really thinking about either augmenting or in some cases replacing the manual inefficiencies or manual work, custom work. And professional services, consulting services work is very much by nature is bespoke business. So yes, that is actually the most disrupted industry within the enterprises. And interestingly, Certinia is a software as a service provider. So we are actually providing software as a service to professional services companies. So you think about people talk about the SaaSpocalypse or the professional services disruption. So we're really right in the middle of all that change.

A ton of change, ton of change, which also presents a lot of opportunity for our customers. Really, I think this is a big opportunity that we engage with our customers on a very regular basis. We have executives from all these large companies that we work very closely with. They bet their business on running this. So we are all in many ways figuring it out together. I would say last two years, probably last nine months particularly since Claude Code became a thing early part of this year when it really kind of blew up has been a fantastic journey to really think about how we apply, where we apply these learnings. And that's what we've been working very closely with customers. We launched our AI offerings. We call it Veda, V-E-D-A, which is a Sanskrit word for wisdom or intelligence. And that has been a great traction, fantastic opportunity really to work with our customers and help them move faster.

So Veda is effectively accomplishing what we have tried to help our customers through software as a service for a long time. And what software as a service has done has really put the user in charge and the workflows that are built around them, data that we have from CRM, from HCM, from ERP, from different systems. Really at a place where the users can actually do their automation very, very efficiently. Apply all those learnings to agentic, where automation evolves into autonomy. And the difference really is we visualize a world where humans and agents are working together. And that changes the workflow in many ways. That changes the way we get our work done.

In many ways. So I think you made a couple of interesting points. So I think first of all, workflows are still very important in a world where humans and agents are working together or agents are actually getting some stuff done by themselves, autonomously. But I think the foundation of this really has to be the context, the context engineering, which is about structured data, which is about unstructured data. So Zoom meetings, Slack, emails, et cetera. It's about metadata. So how you describe your data, the schema, how things are kind of organized. And it's also about telemetry. So the context engineering, which is really the foundation for making agentic actions and the human workflows really get to the next level. So I think the power of this is now really highlighting how critical it is to have that unified data, the foundation, activated in the right way. And this has been an age-long kind of issue for enterprises where, yes, we all go back to data. That's easy to say. But never in our history has it become so critical to see the gaps in that data that become very unactionable or ineffective and you apply agentic to it.

So not quite. Not quite. So when I say context engineering, I actually mean those four very specific data points. I mean the structured data that is CRM data, HCM, ERP data that is actually stored in some kind of repository. I mean unstructured data, I mean telemetry, I mean metadata. That is the foundation. So think about this as almost like a five layer cake. The first two layers of that cake are context. So structured and unstructured data, metadata and telemetry. And then think about the third layer as the business logic or application logic, which is I think what you're referring to when you think about the institutional knowledge. So in many cases Software as a Service codifies that institutional knowledge, the ways of working together, the ways of how we actually, in our business, for example, how you manage the set of resources across different geographies is very much specific to a customer, specific to the professional services organizations. They actually have some criteria. If there's a project comes up, this is how I want to sort of manage that project. I want to resource that with this type of composition. This is sort of related to their business model, etc. But that is the business logic. Think about the fourth layer as a reasoning layer where picking up which parts of the business logic, which actions are appropriate in which areas, and the top of this is the user experience. So in an agentic situation, we still need the context, the two layers of context I mentioned. We still need the business logic, which is the institutional knowledge, which is really the ways of working, so to speak. They may have been expressed as APIs. Now they're expressed as agentic actions are available through MCP. We need the reasoning layer, which is the judgment of which action to pick in which area. Now with the UX, it can be headless. We have a certain app that for the last 10 plus years, our customers have used. We have two million users that use the application across 1400 enterprises. But we also have the headless experiences that we don't force you to use that certain application experience. You can actually use it from Claude. You can use it from OpenAI. You can use it from Copilot. That's sort of the power of this is wherever the work happens. But the foundation of this, I think just stepping back, is the context. Then you apply the business logic and the ways that we have learned across many, many years of what really makes the professional services organizations sell better, manage their projects better, deliver their projects better. We codify that. We have a reasoning layer. Then we make that available across different UIs that our customers care about.

I think yes. I think in, I would say, majority of cases, it's about transformation that our customers would be called on to. And it's based on technology transformation, AI transformation, making sure the data pieces are together, et cetera. The focus a lot of times is, how do I make this disruptive technology work for me in a very persistent way, not just as a pilot or experiment, but really as a persistent, something that I can actually change the outcomes that I want to drive, hopefully for the better, using this technology. So that's what our customers are.

No, I think that's not quite true, because our customers do have very specific problems that they actually have a very good thesis where AI can actually help them solve those problems. And I would say that has been an evolution over the last even six to nine months, where it started out as an experimentation. Let's just see the potential. And interestingly, a lot of our customers, executives, users started using AI on a firsthand basis. They might have ChatGPT, their personal account, work account, et cetera. And now that has progressed to something that is enterprise-grade AI for them. It has a harness. It has some skills. People are actually now at least using AI for their enterprise applications in many cases. They might be even replacing some of the things that they might have used with their own custom applications, et cetera. But now, I think they actually have very good thesis where AI can help them grow their revenue, where it can help them engage better with their customers, where it can actually help them improve the utilization of their human teams, where it can complement the skills that they don't have, which they might have to not bid on the projects because they don't have those skills. Now, they can actually bid on those skills. So that means their overall, the top revenue attainment addressable market actually goes up, too. So I think it's actually a number of specific use cases that they're thinking of.

That is absolutely right. That is absolutely true. In fact, that's the reason why, if you think the progression in conversation has been, "Oh, we were running some pilots. Oh, we actually signed a contract with OpenAI or Anthropic, and we started deploying this thing." Adoption becomes like a requirement. "Oh, I got to have 80%, 100% of my developers using this, too. Oh, boy, the tokenmaxxing, and all this sort of budget impact that it actually has. I didn't really plan for that type." Yes, CFOs care about that, and CFOs are now thinking about it. On one hand, I thought this was going to create some more efficiencies for me, cost savings, or it might actually improve revenue for me. I didn't really expect a new expense item that is now becoming bigger and bigger. So it definitely is a CFO conversation. It definitely is a CEO-level conversation. I think it's become very much an executive suite conversation.

Yes, so in many ways, I would say the role of a CTO, particularly, is right in the thick of this, because you have to visualize where the future is going so we can productize and deliver that innovation in a very predictable package way to our customers. Because here is the challenge that this experimentation to movement to CFO creates. It potentially creates a world where everybody is solving it in their own bespoke custom way. And the opportunity that as a CTO that I see is there are patterns, there are ways that we know our customers need to solve certain problems in certain ways. How can we actually solve that in a highly leverage, highly scalable, highly efficient way, while delivering the level of innovation that they expect? So at the end of the day, AI becomes a tool for them to drive their business. And instead of having them maintaining all those tools in-house, they get to focus on still their business. So that's how they make money. That's how they have their revenue and margins and utilization and other sort of goals. We really want to help them achieve those goals by standardizing the stack, standardizing their technology, standardizing the products through AI.

Yeah, so I'm actually not talking about personalization at all.

That's exactly right. So that's the whole thesis for why products exist versus everyone having their own custom solutions. In many cases, by the way, custom solutions are needed and customization of existing product is needed, et cetera. All I'm saying is there is a ton of value in having a standard set of services, capabilities, and tools that work within your enterprise fabric. So you get to actually spend time and energy and focus on your core competency. Otherwise, you end up becoming your own collector of those services and try basically inefficiently spending resources.

Yeah. I think this goes back to the fundamental question that has been a perennial question. Just because you can do something doesn't mean you should do something. What you should do should be based on what your strategy is. So the conversation in those cases is exactly as you describe, which is, yes, you can actually build something once. And by the way, you can build a code. You still need to have the data foundation. You still need to upkeep it. You still need to maintain it. How are you going to think about the orchestration of different models, orchestration of security layer, orchestration of user authentication? The whole system that comes down to is still, it's actually a productized enterprise grade system that in some cases our customers have hundreds of thousands of users within their organizations. Are you actually now going all the way from something that works for maybe 10, 20 people to hundreds of thousands to even touching millions of your end customers? That is something you have to make a strategic choice if that's the business you are in. Or are you actually in the business which is, what your CEO says, this is how we make money as a business strategy.

Yeah, so that is a key question that, as you said, is really top of mind for everyone, I would say without exception. In fact, it's so important that just a few months ago, we conducted a research. We worked with a third party, reputed analyst and really looked into this problem. The answers actually are counterintuitive to what the general narrative seems to be within professional services industry. This is actually validated by a lot of customer feedback, as I mentioned, that we work very closely with. Here is the overall thesis. With the AI, a few things change. Number one, the delivery teams. That's what our customers care about. The service delivery teams become hybrid. They're no longer just human teams. They're actually human plus agents or agentic capabilities or, shall we say, overtime digital workers. I don't think we have digital workers level quite yet, but we're heading in that direction. The point is that the teams become hybrid at a delivery site. What does that do? That actually unlocks the demand side in a very, very fundamental way. You are able to bid on projects that you couldn't do because of the skill constraint, because of the human constraint. Do I have the right people in the right time zone, in the right location to deliver those services? That actually constraint goes away to a large extent because you can augment those skills with your agentic capabilities. We did this analysis and we saw that your addressable pipeline actually goes six to eight times more for a customer, the type of customer that we serve. Six to eight times more opportunity. That does require a reconfiguration of how you manage your businesses, how you manage the delivery, how you engage with customers. The net-net point is that the counter narrative to maybe a general narrative might be is that it actually creates a six to eight times bigger opportunity as opposed to shrinking the overall pie. It's a different pie, but it's actually a much bigger pie than what exists today.

So keep in mind the pie from an opportunity perspective goes six to eight times. You still get to decide how you deliver those services.

So no, I actually do think AI is a great democratizer in the sense you can actually pretend to be a much, much bigger, quote unquote, bigger guy with the power of your hybrid teams delivery. You can have a lot of agents and digital it is a capital investment. It is an investment on OpEx, but you can actually deliver that level of service. And especially when things are changing very fast, I wouldn't take it for granted that the incumbents have advantage. Incumbents actually are the ones who are getting disrupted. So I actually think it is an equalizer in the sense if you are a relatively small professional service organization and you specialize, say the right now the lot of focus is on forward deployed engineers, and there are very few of them overall, especially if you actually have the right tooling at your disposal, and you're able to actually help deliver that transformation for your customers. I think you have a probably a better shot than you might have had a couple of years ago. Now, there are, I'm not underplaying the benefits of incumbency, right? Incumbents have a lot of advantages. True, but I think the technology has always been a democratizer, equalizer. AI being the most disruptive technology, at least in our lifetime, I think is also the most equalizer in that sense. So I'm not there that it somehow favors one group or the other. I think there are pros and cons of it. And the story is yet to be written to be found. I don't think it is given that somehow the big customers, we work with large customers. I understand. I also would tell you that there is a ton of focus and anxiety to actually make that change and not be complacent about it because this can actually unlock opportunities for not just the incumbents, but for a lot of other size companies.

And I think there's another part to it, which is often overlooked. Trust is important. Don't get me wrong. Trust is foundational. The reason why humans are critical in this equation and more important is because of accountability. And I'm not saying the word accountability in a very generic sense. I mean, there is a legal regulation sense of expectation in terms of service delivery, in terms of product delivery, in terms of anything that you're servicing a customer and getting paid for. And that accountability does lie with the company, with the individuals, with the humans. And of course, the trust is important because that's how you get the business. But accountability is how you sustain it. You actually grow the business because, you know what? A lot of times there are so many things that we love about generative AI. You have a lot of creativity, a lot of applications. But in many cases, it's not okay to be 99.999999% right. You have to be 100% right. When you are in an account receivable, account payable, financial transactions, like all of our customers are, we can't just write an invoice which is five nines correct, ten nines. It has to be 100% correct. Otherwise, guess what happens? You get sued, right? And it's a very fundamental requirement, which is even more important in the time when everybody's sometimes using, overusing the AI technology and forgetting it's a tool. Still, I think the accountability is with the human team that is engaging with that customer.

Yeah. And some of those things could be transitory. I think the challenge, and I think one of the examples I saw someone change their LinkedIn profile to say something very specific about I am like so and so within the AI world. And guess what? When you do the Google search, the Gemini results come up about XYZ person, and it actually picks up exactly the same words. So in a way, it's a very gameable kind of system. But I also think with the pace of technology and innovation, some of those things would get fixed. What wouldn't get fixed alone, just with technology is that accountability.

I would say for me and for our team within Certinia chief product and technology office, I think there's a huge transformation that is underway for us. Right. We've talked about what's at stake for our customers, and we obviously care, we ship the products, we give them Veda agents and a lot of capabilities. But to do that, at the rapid pace that things are happening, we ourselves are going through a lot of change of how we build products, how we deliver software, how we deliver AI. And it is a wholesale transformation all the way from thinking about engaging with the customers, thinking about requirements, thinking about opportunity analysis to use that experience to design ops to really thinking about how those become much more rapidly coded using, obviously, Claude Code, peer reviewed, and then they go back to the right harness with the data that is very applicable for the customer scenarios. And then that gets evaled. So we're actually being very confident before we deliver something. So this is a new way of building products, right? So the AIDLC versus the SDLC: the software development lifecycle is now an AI development lifecycle. And really, it's a transformation based on new tools. It's an entirely new process. It requires very different structure, much, much smaller speed boat type of view, even smaller than what we used to even like we used to have six or seven people. Now we have teams of two or three people really delivering the same or more level of productivity. And the, frankly, the incentive and the role definition and training and enablement, etc. So that level change is something that we are driving internally. We're basically not only experiencing this, but I think this is the root of, you can think about this, anxiety, or this is the biggest case for change, right? In many cases, it's exciting to be an engineer, even though I think there's a negative case to be made about redundancy from an engineer perspective. If Claude Code can write the code as good or better than engineer then what do I do as an engineer? I think it just opens up a question of what is the way that we want to have an impact? Was my impact writing code that is syntactically correct? The IDEs can do that in much better way even 10 years ago. Or is my way to deliver products or deliver outcomes that my customers care about? That fundamental notion of helping drive the innovation for the benefit of our customer base at large. How do I do that in this new world? I think that is something that is a big driver of change.

Yeah, I also think it changes the way leaders communicate, even amongst themselves or with their teams. And I think the time frame for delivery and time frame for this transformation is compressing so much that it's actually not about communication and then wait for somebody else to respond. I think of this as all of us are in it together. Communication really becomes more of a deeper level of engagement and an ongoing basis because you communicate when you are actually clear about here is the product stack, here is the tool, here is the methodology, like this is set in stone, here's the governance process and go do it. A lot of times we have that level of clarity. We have a clarity on what our vision is, what our choices we are making from a strategy perspective. But frankly, we are also very open about things that are changing. Yesterday this was this model. We started out, by the way, just in the last few months, our own journey was we went from Cursor to Claude Code. We used some of the OpenAI to et cetera to now thinking about in a little bit more gateway sort of approach. But that being said, we are making some bets right now for Anthropic Claude and making sure everybody is trained, everybody has skills et cetera. But the point here is it's about engagement. It's about making sure that while things are changing, there are some things that don't change. The vision, our overall strategy, how we are changing, we do need to really get even better at learning faster, driving the things, experimenting and discarding things that don't work and focusing on the few things that matter because at the end of it, that creates value. So I think that changes the way we communicate because we can't just say, here is the memo, let's go do it. But at the same time, we got to make sure everybody's coming along together. Different people react differently.

Yes.

So it starts with transparency. Transparency on what we care about as leadership team, as success criteria. So being very clear about those. Second, it's also our transparency of dashboards. We've always been very clear about, here are the velocities, teams, different sort of feature crews, et cetera, et cetera. Now we actually have added the new metrics around AI production, the budget, the cost, how different teams are delivering. And the cost is one part of it, but the impact. I think we're also very clear about it's okay in some cases to spend a lot of tokens for certain things, as long as we see the value potentially can be bigger. So I think a lot of these things are about being transparent about how we measure success, what are the right kind of measures right now, and understanding where the bottlenecks are. Because we have a few hundred engineers across our team in Europe, in India, in North America, and making sure the collaboration happens to share the best practices and identify what doesn't work as a community of innovators, community of product and engineering team members is also very important. So a lot about this is enabling everyone to be able to share while things are moving very fast is critical. And then kind of supporting through the right investments in tooling and data and harness in whatever ways we can.

Yeah, so the surprises basically go back to questioning the stereotypes. And I'll tell you some of the stereotypes were, oh maybe people who have been very seasoned, a long time at the company, are sort of set in their ways and they may not be very adaptable to change. Not true. Domain actually matters quite a lot in those cases. So that somehow shouldn't have been a surprise. But that actually has been a surprise to me how useful having an opinionated editorialized point of view about what success looks like for our customer, even if you deliver through SaaS, user experience, headless, agentic, those are how you deliver, but what matters is even more important. I think the second big surprise for me personally has been how different roles have melded together. So effectively, we have two main roles now. We have obviously functions in design and user experience and documentation and content creation, etc., etc. But really two kind of main roles. One is a "product manager," right? But their job is to identify requirements and verify and figure out what to build, why to build, and then the software engineer or developer. Really, those are the two pivotal kind of things. And they always have been, but we have had folks who would do maybe some of the roles and other functions move into either you are actually working very closely with the Claude Code type of skills and identifying and peer reviewing, etc., so much more of a closer to the code delivery. And other role which is closer to the customer understanding opportunity, why we should do, and then the requirements, etc., and both of those two working together. So the amalgamation of roles is actually something super interesting. I think it creates a lot of opportunities. It creates a lot of potential risk because someone's identity was "I am the best designer that the world has seen." And we have a lot of people, by the way, with that. How do you actually transform that into "How do I become a creator of the best design system that I contribute to with the design library, design assets, so that it can actually unlock product managers and engineers?" I think there are a lot of those surprises that are part of the transformation.

Completely agree with you. And I'm actually glad you use the word "velocity." That is my favorite word. And I would say within physics you have speed and velocity. We're talking about velocity because velocity is speed with direction. It's a vector quantity. So again, you're talking with engineers. You basically want to be very clear, not only doing things with speed is important, which you can do with a systematic approach, which you can do with a lot of AI tooling, but doing it with the speed, with the taste, with the judgment, with the taste, with the creativity, with your own personality, like the company personality, the product personality, the tone of voice, that actually makes the speed of velocity. And that's why I think that that's even more important when, you know what, there's so much AI slop, just we can see this on LinkedIn post. You can see this a lot of content. And that's one example of that. There's a lot of potential AI slop in the code, right? Because everybody's kind of doing something similar ways to use the libraries or do the same ways. It is way more important to actually be vigilant so that when you have thousands of lines of code that you can write in a day through a team, how do you make sure the quality assurance is still there? That it doesn't create more problems than it solves. But more importantly, how does that still embody the way you solve the problems, the way you communicate with customers? That type of personality is even more important to bring in that systematized code. It is doable very much. By the way, this is why the harness that you focus on is important to make that in a scale, but that is important. And that is something that can be lost if all you focus on is speed.

I think when everything changes, which by the way, I agree with you, everything changes and it actually raises the bar for us to now rediscover the purpose and the impact that we can have, we should have. And I think that's actually a good question to ponder because we just do sort of contextualize this as an example in a sort of maybe different way. One of the things that within professional services, customers do, and usually you're very familiar with, which is you go to Zoom meeting, you go to team meeting, you go here, you go different things. A lot of those unstructured conversations may be happening about writing a new proposal, writing a statement of work, delivery this, something is behind. And those things, you come back to your computer, you kind of think about it as you write it down, et cetera. And yes, even in the AI days, there is transcription from all the Zoom calls, et cetera. You still have to read it, you can do this. We just made an acquisition a couple of weeks ago, and we have this new product in the market now very rapidly called Veda Accelerate. All it does, by the way, on a very simple level, is actually connects the unstructured data, puts that into structured data, of course, with your human approval. But just imagine the world now, I go to different meetings, I have Zoom, I have team, I have email, et cetera. After, say, a few hours of work, I come back, it actually has told me, oh, by the way, for this project you're working on, these are the risks that were identified in this meeting. These are the things that, oh, you should probably deliver on this project, et cetera. And with a little bit of observing, I can actually say it becomes part of my structured data. Why? Because now the unstructured and structured makes it actionable for me and my agents to take some the next step. So in many ways, did I actually enjoy remembering all the action items? I said, no, I wanted that to be done. What I enjoy and what this technology allows me to focus on more is actually the problem solving is what do I do with this? Because all that grunt work, which I might have spent hours and hours and maybe would have missed half of that anyway with the cognitive load that most of us have, I can actually now focus on understanding how we address those risks. Make that sort of customer happy or those type of things. I think that's the power of AI in many ways is that it allows us to do things that maybe we were born to do. It unlocks...

I think the medium depends on the context the business is operating in. There's a very different answer for me as a consumer, me as a person, versus me trying to get the work done in the best possible way and drive the accountabilities I have. On a very simple level, the medium in which AI helps me is the channel I'm operating in at a given time. I could be in Slack, for example, communicating with my team or with customers. I want to be able to get the information right there, without leaving and going somewhere else in Certinia, trying to find it, and then coming back. That type of dissonance, that type of context switching, is not very helpful. That's what our headless AI strategy can help with. In other cases I might be in Claude. It might be email. I think the medium, or the context in this case, is: what is my workflow, what is my way to deliver that value, and can the tool really be a tool that helps me, versus me serving the tool by going to different contexts.

Yeah, I think it's the glue. That has been the focus for enterprises, which have almost always been fragmented silos of HR and sales and this and that. We also realize the work gets done when they're not different islands, when they're one connected place. It has been a challenge. It always is a challenge. I don't think we're going to a world where you put all the data in one repository and act on that. That's not happening either, because by the time you finish that data warehousing project, the technology is already obsolete. People have moved on. I think you're spot on with that view. Whether it's AI or integration, whatever it is, it's about connecting the data sources so the intelligence can support humans in driving the outcomes we care about, because connectivity is a prerequisite for AI to be successful too. I think we conflate things, because ChatGPT became very much a UI chatbot type of experience. For a lot of us, the first aha moment was seeing, "Wow, ChatGPT 3.5 can do this. Holy cow, I never thought it could write this haiku." That almost served it wrongly, because that is one interface, one way to do this. The fundamental quality of generative AI becoming agentic AI, actually doing tasks you care about, with decisions and long-running jobs, can be expressed in so many different ways. One of the things people say is that AI is the new UI. I think that is true. At the same time, headless conjures a very odd way to describe it, but that is also true. It's the new world of connections that come through endpoints that APIs unlock. But I think the question really is: how do I tap into the intelligence that operates on my unique data and context, takes advantage of the domain knowledge we have as an organization, so you can put your best foot forward in driving the outcomes and KPIs you care about, like revenue, margins, and utilization? I think that is the opportunity it unlocks.

Oh, mostly talking now. It's the game changer.

And it is much more natural. It takes away that "I have to do this because it's a very enterprise thing to do." It's as close as it gets to being a coworker, because it has its own personality. It has its own ways.

Voice as an input medium, in fact video, is very powerful. But the output in our case is still code, design, how you get things done, all the artifacts that are important. But I think you're right. The power here with AI, especially the way we think about Veda, is that your average agent is a reflection of your best employee. Just think about it. Take our staffing agent. It's almost like a reflection of the best staffing manager of the thousands of staffing managers you might have had, because it's learning from that. So when you ask it to do something, you can be confident the average agent is doing the best any human could. You still have to double check, but more and more it creates the confidence that the average agent is your best employee, because it's trained on the best data. It's giving that advice with a level of context that isn't humanly possible.

We do that right now. Here's how it materializes. We have agents that have their own set of authorizations and permissions, what they can and can't do. In every case, no exceptions, those are a subset of the permissions of a human they're connected with. So we are not, at least from our perspective, at a place where we have 100% autonomous agents floating through the enterprise. Those agents are very well defined, almost like being set up in your directory with specific roles. But they are assigned a subset of the permissions of a human who is accountable at the end of the day, and that's how we provision them, almost like a new employee.

Correct. It's definitely not one-to-one in that sense. It's one-to-many. But the point is that accountability works, first, when it's tied to a human identity, and second, when the permissions and authorizations it has are a subset of that.

Sounds good. Thanks for having me.