Transcript: The Future of Enterprise AI: $100M ARR, Agents, Company Building, and Scaling Unicorns | Arvind Jain (Co-founder of Glean, Rubrik, ex-Google)
Full transcript: The Future of Enterprise AI: $100M ARR, Agents, Company Building, and Scaling Unicorns | Arvind Jain (Co-founder of Glean, Rubrik, ex-Google).
2025-09-13
Host: So I'll give you like the the provision for the future of work from our, you know, like that that we believe is going to, you know, becoming soon is every person who works is going to have this incredible personal companion.
Um, companion that knows everything about you and your work life. It knows your role, it knows the company that you work at. It knows what are your annual objectives, OKRs for the next quarter.
It knows what your career ambitions and goals are, what you want to actually become. It also knows your weekly task, things that you need to actually complete. It knows your day, you know, all the people you're going to be meeting today.
It's actually walking with you wherever you're going. It's listening to every word that you say, that you speak. It's listening to every word that you hear.
Uh, it's in the systems with you, looking at messages that you're sending, documents that you're writing, questions that you're answering. And with all of that deep knowledge of you and your work life, now it's ready to actually help you.
And help will come, you know, very much in a proactive manner. What is a company? A company is basically a group of people, set of people who are building something together, building products, building a business together.
And so in that sense, you know, I I do firmly believe that the scale of your business and is sort of proportional to the the number of people that you're going to be having in your business.
And so I don't personally believe in that hey, like, you know, like I can have a five-person company and I will generate a billion dollars of business with it. Um, the um, I think like companies will largely look the same way.
Host: Hello everyone. Today my guest is Arvind Jain. Arvind is the CEO of Gleen, which is a Work AI platform. Gleen has quickly become one of the widely adopted Enterprise AI platforms estimated to reach a revenue run rate of $100 million.
It's used by companies like Zeppier, Carta, Grammar, Data Bricks, and twice named as Forbes Cloud 100 companies. Um, Arvind is a quintessential Silicon Valley veteran, you could say.
Uh, he's the co-founder of Rubric, which is a cyber security and cloud data management company, recently valued at $16 billion and IPO last year. He was for a decade a distinguished engineer at Google and a founding engineer at River Bread Technology.
I'm super excited to talk to Arvind about what is Gleen building for enterprises? What is the future of work and agentic AI look like? And his secret sauce behind scaling two successful companies.
If this is the first time listening to startup project, don't forget to subscribe to startup project on Substack or wherever you're listening to the show.
Host: Arvind, welcome to the show.
Guest: I'm really excited to be here. Thank you so much, Nitraj.
Host: Uh, so just to set the context, um, you know, uh, my wife's company actually uses a Gleen. So I had a little bit, uh, you know, I was playing around to prepare for this conversation.
But I think most people if they if their company is not using it, um, I think they'll not be aware of, you know, what is Gleen and how does it work. Uh, so can you give a, you know, sort of a pitch of, you know, what does Gleen do today?
Uh, how is it helping enterprises?
Guest: Yeah. So, well, I mean, I think most simply, think of Gleen as, you know, it's like Chad GPT but inside your company. Um, it's a conversational AI assistant. Uh, employees, uh, can go to Gleen and ask any questions that they have.
Um, and Gleen will actually answer those questions for for them using all of your internal company's context, your data, your information, as well as all of the world's knowledge. Um, and and so that's very simple.
So like the only difference here like if you think about Chad GPT versus Gleen, Chad GPT is great, it's intelligent, but and it knows everything about the world's knowledge, but it doesn't know anything internally about your company, like who are the different people, you know, uh, what are the different projects, who's working on what, if that context, you know, is not available in Chad GPT and that's the additional power that Gleen has.
Um, you know, in our in our product. So that's that's that's the core of what Gleen does. We started out as a search company.
So before before these AI models got so so good, we didn't have the ability to sort of take people's questions and and just, you know, you know, produce the right answers back to them, uh, using all of that, you know, internal and external, uh, knowledge.
Uh, so in in the past, we would actually, I would call ourselves more like Google, uh, for your workplace where you would come to clean, you would ask questions and we'll surface the right information.
Uh, but as AI got better, we got like the the ability to actually go and read that knowledge and actually instead of pointing you to 10 different links for relevant content, we could actually just give you the answer right away.
So that's sort of the evolution of like how we, uh, went from being a Google framework place to being the chat GPT for a workplace. And that's that's the core of what we do. Uh, but we're also an AI agent platform.
So the the same underlying platform that powers our chat GPT like experience, um, is is also available to our customers directly to build all kinds of AI agents across the different functions and and departments, uh, on on on on Gleen and ensure that, you know, they're delivering AI in a safe and secure way, uh, to their employees.
Host: So, I mean, the interesting thing, you know, I I I was following Gleen before, um, you know, for a while. I think you started in 2019 and you start, you mentioned it, uh, you started as an AI search company. Yeah.
Uh, I mean, now I think it sort of feels very natural to build something, you know, a chat GPT kind of thing for enterprise because, you know, the value is sort of instantaneous and anyone can understand it.
But why did you pick the problem of like solving enterprise AI search because it was not like the hot thing or, you know, it's not a problem that's, you know, so obvious, um, to pick for like what what was your thesis initially to go for that problem?
Yeah.
Guest: Well, for me it was obvious, but yeah, for the for the for the because, you know, I was actually suffering that pain of, you know, like inside, like, you know, as you mentioned, like, you know, before we started Gleen, I was one of the founders of Rubric.
You know, we had really good success at Rubric, we grew very fast and and in four years, we had more than 1,500 people in the company. And as we as we grew fast, you know, we actually did run into a problem and that pro that problem was productivity.
Uh, there was this one year where we had actually, uh, doubled our engineering team, we tripled our sales force, but if you look at our like our metrics, you know, like how much code we were writing, how fast we were releasing our software, everything was flatlining.
Which just couldn't like, you know, it was almost like a ceiling of like, you know, we just couldn't do more than no matter how many people we have in the company, we couldn't actually produce more.
And and and one of the key reasons for why we were hitting that productivity, uh, uh, uh, you know, issue was, you know, because the company grew, uh, a lot very fast and there was so much knowledge, so much information that we had inside the company, it was fragmented across many, many different systems and our employees were complaining that, hey, I actually cannot find the information that I need, that I need to do my job.
I also don't know who to go and ask for help when I need help because there's no concept of who was working on what inside our company. There was no good employee directory.
And, and so when we saw this problem as being the number one problem that people are complaining about, I decided to solve it. Um, and my first instinct was like I'm a search engineers.
When people say that, um, you know, I'm not able to find things inside the company, we said, okay, well, like just go and buy a search product.
You know, uh, get something that can connect to like all of these, you know, 100 different systems that we have and make it easy for people to find things. And and that actually revealed to us that there's nothing to buy.
There's no product in the market that would connect to all of our SAS applications, you know, that we had, um, and give people one place, you know, where they don't have to remember anymore.
It's going to just simply go and ask questions and we can surface the right information. So that was sort of the origins like, you know, I felt like wow, like, you know, nobody has actually tried to solve the search problem inside businesses.
Of course, there's a great company out there like solving search in the consumer world, you know, which is Google. And so that got me initially excited about it. Uh, about wanting to solve this problem.
And at that time, you know, we were not thinking about building, uh, building a chat GPT like experience. Actually, nobody knew at the time, um, how fast AI technology will evolve. Uh, but our mission was actually very similar.
Our mission was to help, you know, save people time, to get them the right information as fast as we can. Um, answer questions, you know, whenever we can by extracting, you know, that information, extracting the answers from your knowledge.
Um, and also helping people connect with people like, you know, if you need help inside the company, connect you with the right experts. So that was sort of the problem statement that we started with.
And AI actually played a big role in it from day one.
Transformers were actually already, you know, uh, although the the rest of the world was not aware of it, but in the search, you know, uh, community, uh, search technologists like which is what I and most of our team is, you know, we're all we're all ex-search engineers.
We actually knew that Transformers were playing a big role and making search better. So we, you know, we actually, uh, deployed, you know, language models in our technology stack in 2019, well before anybody else did.
And in fact, that makes us the first Enterprise AI company out there, uh, in the world, uh, because, you know, Transformers was the technology was that was literally created, you know, to make search experiences better.
So it was, you know, obvious for us to actually use it. And then then over the years, you know, we've, you know, got more and more capabilities from, uh, from Transformers, which allowed us to actually become more like an assistant.
And that's actually an interesting story. I'll I'll also share with you. In 2021, we launched our company to the public and we called ourselves the work AI assistant.
Um, we didn't call ourselves a search product at the time because, you know, we could do more than search. We could actually help you with your questions, we can actually we could answer it for you.
Um, you know, as as much as we could could also be proactive and and actually help you sort of stay connected with the company. Uh, and and so we did a whole bunch of things and we launched our company like that, but it was a big crop.
Like nobody understood what an assistant is.
Nobody really sort of had yet like, you know, uh, like nobody had seen Chad GPT and so it was like from a marketing perspective, it was a bit of a failure and we had to sort of rebrand ourselves and and call ourselves a search company at the time.
Uh, and then of course, you know, with Chad GPT, uh, launching and like people now realizing how capable AI is and like it can really assist you and can really be a companion to you, which is when we sort of, you know, came back and start talking about our original vision.
Host: I think almost, uh, pre Chad GPT, no one called AI as AI because it was called ML or some other technical term. Yeah.
Like, I used to remember watching 2020, 2021, all the Pixel phone launches and Google used to do a lot of work in terms of, you know, creating products, AI first products. Um, very early on.
But for some reason, the tragedy is, Google is considered as they are not doing enough with AI. Obviously, in last couple of weeks, that's sort of I see that narrative sort of changing now.
Uh, but that was the tragedy because I clearly remember I'm a Pixel user and Google was AI first on Pixel very early on, way before it was talking about AI. So that was sort of like a narrative versus experience difference that was happening.
One question related to like why you picked enterprise search. One of the things, you know, um, uh, I had a CEO of Ward.
They're doing AI terminal kind of a product and one thing he mentioned was when you pick a really hard problem to work on, uh, a couple of things actually become easier and most people should actually pick a harder problem.
His thesis was, it's actually easier to convince investors because it's a problem that is so hard that if you're successful, the returns are going to be very high and very few people inevitably do it.
And also you can polarize people in a way that you can self-select, you know, people attracted towards solving hard problems, heading problem as well.
What what what's your take in terms of like picking a problem for, you know, when you're starting a company?
Guest: I actually agree with that, uh, assessment. I think solving, it's not like, you know, that you're trying to just pick something that is super hard, um, to solve as the main criteria.
Like the main criteria, of course, still has to be that you have to add value, you have to, you know, build a product that is actually going to be useful to people.
And and and I'm always uh, attracted towards working on problems which are very universal in nature. Um, something, you know, where we can bring a product to, uh, everybody, uh, around us.
Um, and and that and and the reason I like it is is both because of impact that you're going to make.
Um, you know, you know, building a startup like, you know, let's say is is not exactly a, you know, if, you know, fun and exciting ride at all times, it's actually a difficult journey and you have to have something, you know, something that sort of makes you go through that.
And and that something for me is impact. Uh, you know, you know, uh, solve a problem that is going to actually build a product that's useful to, you know, to a very, very large number of people.
Uh, and then second, when you think about solving problems, you have to think about what your strengths are.
Like, well, if you are, um, if you're a technologist, it's, uh, it's actually a luxury or it's it's actually a gift if the problem that you're trying to solve is a difficult one.
Um, because you'll be able to actually build, uh, that technology with the best team and and you won't actually get commoditized quickly. Uh, because you know, the problem is so hard to so hard to solve.
So, so with search, I actually knew how hard and difficult, you know, the problem is.
And so that was definitely an exciting part of like why I also started Gleen is because the I knew that, you know, if we solve the problem, uh, it's going to be a super useful product and it's also going to be it technology that others won't be able to replicate quickly and it'll allow us to sort of actually, you know, bring a distinguished product to the market and and and and and build a phenomenal company.
Host: So, uh, one of the things, um, that I often see for like chat GPT or like, you know, Gleen AI or M365, you know, in the enterprise context is when you're working on certain types of data, it's not enough that you're 90% times accurate. Yeah.
Um, if I'm reporting something to my leadership, you know, business and versus revenue numbers, I really want it to be almost 99.9% accurate, right?
So like I still feel like we are not yet there in terms of and obviously I'm not a regular user of Queen.
So can you talk a little bit about like what are the techniques or, you know, you are trying to implement in terms of like reducing, you know, for lack of a better word, hallucination or just to understand that context because for me, numbers are in a different context versus someone in my organization doing a completely different thing.
Yeah.
Guest: Yeah. So, so this is actually like AI is progressing quite quickly, uh, in terms of its capabilities. So there's also a lot of work that we don't do, but, uh, the platforms that we use, like, you know, actually, uh, do.
So like if you look at the models today, uh, whether it's GPT5 or Cloud Sonet models or a Gemini from Google, they are they're significantly different from, uh, the models that we had last year, uh, in terms of, you know, their ability to really, you know, deeply reason and think and and in fact, like review their own work, um, and give you a lot more confident high high, you know, uh, higher accuracy answers than what they could do last year.
So there is there is that general like, you know, uh, you know, improvement that has happened at the model layer, which which is actually reducing hallucinations significantly.
Um, the but then the second thing if you think about coming into the enterprise and your business use cases, the none of these models actually know anything about your company, your business, your, you know, you know, like all all of your knowledge that is private and internal inside your company, the models have no knowledge of it.
So when you start to now solve for specific business tasks, the typical workflow that you have is that you have a model that is thinking and it's also actually retrieving information from your different enterprise systems.
And you're going to use that as sort of the source of truth to actually perform its work.
And and so it becomes very important, um, for your AI product to ensure that, you know, for given any task, you're picking the right, you know, up-to-date information that is fresh, is high quality, written by the right subject matter experts and is the truth, um, and then give give that information to the model so that the model can actually, uh, produce the right answers.
Otherwise you will end up becoming a garbage in garbage out situation and that is what actually, you know, most people are struggling with right now.
Uh, as people are sort of building these, you know, like, you know, these AI applications and they're just sort of dabbling, you know, their way into retrieving information, then that's the first problem they run into and then they will complain to their customers that, well, your data is bad.
There's so much, you know, you know, inaccurate information out of data information inside a company. It's not our fault that, you know, the answers that the AI systems are are building is bad because you're feeding it the wrong information.
And and and that's actually not the right answer because, well, AI should be smart enough to understand what information is, you know, old, what information is new. Think about you as a human.
As a human, if somebody asks you a question, you'll go and try to find an answer, you will do research, you will actually read a bunch of documents and you'll have that judgment that hey, like this thing looks old to me. I should not trust it as much.
Um, I want to look at more recent information. If I'm not finding more recent information, I'll actually go and talk to somebody like who is, who I believe is working on this area right now. And so that is that's our thought process.
This is how we process and we you're still able to live and handle, you know, that low quality knowledge that obsolete knowledge inside your company and still ultimately as a human bring the right answers back to, you know, those actually asking you for that, you know, information.
And AI has to go the same way and that is what Gleen does. Like our in like what we do internally is when you bring Gleen to to an enterprise, we go and connect with all of your different systems.
We understand knowledge at a deep level, we understand what knowledge is high quality and fresh written by subject matter experts as I said.
Uh, and we ensure that models are being provided the right input so that they can actually produce the right output. So that's that's that's the core of it and like, you know, in fact, like our entire company is focused on that.
It's is focused on like how to sort of tease apart like good from the bad, like, you know, within your enterprise knowledge.
Host: You're almost like reinventing the page rank kind of algorithm for enterprise and, you know, making sure that it works well with the models, um, as the intelligence layer.
Um, can you talk a little bit about, you know, which models do you use and are you constantly experiment experimenting with different models and, uh, if so, like in your opinion, like what which companies are doing in terms of base level models, you know, what are working better than, uh, others?
Guest: Yeah. So we have this, um, layer in our technology called the Model Hub, um, which is where we actually go and establish connectivity with, uh, all the major model providers as well as open source models.
And and so we connect with, you know, we, uh, we support, uh, the GPT family, the O, the the O4, the O3 models from Open AI. We support Cloud Sonet and Opus families. Uh, we support all the Gemini models. And those are the the big three today.
Um, and then there's like a whole bunch of models that are in open source, whether it's Lama, Deep Seek and others. Uh, today what we're seeing, you know, especially for our use cases because our use cases are are quite open-ended.
Um, I described to you that you can come in clean and you can practically ask any question and you could be anyone, you could be a salesperson, you could be an engineer and we are now supposed to do all of the, you know, deep research, you know, inside your company to actually come back, you know, with the right responses back to you.
So the range of tasks that we have is actually quite unbounded. And and for for unbounded tasks like for, you know, where you cannot do pre-planning, um, these models actually work the best.
The the, you know, uh, you know, Open AI, Anthropic and and and Google models.
And they are actually developing like different, uh, expertise as well in the sense that, you know, I think for code writing, for a long time, uh, Cloud models have been actually doing better, uh, than others.
For reasoning, uh, the GPT and the O, the the O3 model had been the like, you know, real state of art.
Gemini has been, you know, better in, like, you know, when you're look looking to work with very large amounts of information, large context windows, uh, it also works better in our in in our testing, uh, with international like if you're trying to write in different languages, if you're trying to actually have a tone of the user, we've seen it do uh, better.
So they're all like developing different, you know, uh, capabilities, getting better at different tasks. Um, and we have to constantly test and evaluate because our end users won't be able to pick the right model, uh, for the right task.
So, so so behind the scenes, we'll, you know, continuously monitor like, you know, for what kind of questions, you know, what are the right models to choose.
Host: In In terms of like, uh, you mentioned agent AI agent platform, what are the typical use cases for which enterprises are creating agents and are there any like use cases that you can give which sort of crystallize, uh, you know, what the platform has?
Guest: Yeah. I'll pick some of the key ones and we'll let's go through a few different departments. So let's talk about the sales teams.
The sales teams, you know, they're most most of the team most of the time the team spend time on is on prospecting, number one. Like like basically reaching out. Lead generation, right?
Like, you know, reaching out to, you know, potential customers, sending them messages.
Um, and so now, so you can build a really good AI agent that does that and in fact can does that now in a faster and and higher quality than a human in in in many cases.
Um, like you know, people have built an agent on Gleen where a salesperson basically, uh, goes and says, hey, you know, like I I would like to prospect, you know, these five accounts today, help me out.
And then, you know, Gleen will go and for each one of those, it will actually do a, you know, good amount of research to understand like, you know, if you have already reached out to some people there, otherwise who are the right contacts and then and then it's going to actually, you know, you know, generate personalized, you know, outreach messages, you know, for the right individ, you know, for the that for the right targets.
And then our our team, we don't actually take the humans out of the loop. Like, you know, we believe that AI is an augmentation tool, not a replacement tool.
So the so like our our sales people can then, you know, review the work of AI and they can actually quickly go and like do thumbs up, thumbs down and and those messages and get sent out.
So like, you know, like it's a it's still uh, and and so it's like, you know, they can now prospect at the rate which is like, you know, five times greater than what they could do before.
Um, similarly, like, you know, let's say imagine that this was a customer, you know, a call between a customer and, you know, in in a vendor.
After this call is done, I have to actually follow up as a vendor with you with next steps, with action items and and sales people spend like, you know, two hours writing a great follow up, you know, with all the right supporting materials.
And and like, you know, people build agents in Gleen to to do like to actually, you know, generate those, you know, the right meeting follow ups and automatically send them. That's another one.
Um, if you are in enterprise sales, you're bidding on large projects, you have to fill RFPs. RFPs are like 500 questions long and they're always almost the same questions that you fill like, you know, so many times before.
And so you can have AI agents that can automatically go and fill those RFPs. Anyway, these are some examples for sales. Customer service is pretty easy.
You know, like day in day out your job is to actually answer questions that your customers are asking. Uh, you know, uh, and helping them their support tickets. And AI is actually pretty good at that.
Like, you know, people have built agents to auto resolve tickets, uh, you know, for for your customers. Um, for engineering teams, um, AI can actually be a really good code reviewer.
Uh, so we, you know, there is that the Gleen AI code review agent is actually quite popular. Um, at any at any at our customers when they use it, then the first person who does a review of any code that an engineer uploads is actually AI, right?
And and it can actually handle all the basics of, you know, pulling the right style guides and things like that. So the use cases, I mean, I'm seeing, you know, they're exploding right now.
And one of one of the things, you know, again different what's different from 2024 to 2025 is that last year it was all about engineering and customer support as the two big, you know, um, uh, departments that would actually that were going all in on AI and sales were starting to.
Uh, but now it's actually all departments.
So legal teams, for example, they are actually, um, yeah, a popular agent with Gleen is a a red line agent, which is when you get a third party paper like an MSA or NDA, um, Gleen's, you know, red lining agent will automatically create the first versions of red line.
And these are like, you know, 300 page contracts. So our cost of like getting an MSA from a third party customer and and responding to it, getting it to in good shape is like, you know, 10 to $15,000 and AI can actually really like short circuit that.
So so that's actually the democratization is happening now.
You're seeing more and more like, you know, every team, you know, inside a company, and you know, they're understanding that like, look, okay, this is the time to go big on AI and and try to save, like, you know, uh, save time and and move fast.
So so a lot of lot of progress this year, um, you know, across across different teams.
Host: In in one form, I feel like the better way to describe agents as like workflow agents because I think the farm factor where we really actually really see very powerful is some something like a Zapier or, you know, uh, N8N or apps which used to do, you know, connections between things.
Yeah. Yeah.
Now that's sort of workflow editing and creating a workflow, basically mimicking what you do in real life, repeatably and basically finding a common pattern, you know, you go to email and you read the email and probably then you synthesize the information from it.
And then you, you know, put it in Excel or somewhere.
So if you just break it down into three different compartments then and if Gleen has integrations into all these different tools, then you have that intelligence layer powering it behind it and you can now create a workflow agent.
And that's why I like the term workflow agent because it sort of is more intuitive than agent because I think agent sets a different sort of expectation like you ask Chad GPT and it goes does a bunch of stuff and comes back. Yeah.
Uh, but I think what I'm seeing is the more reliable way to actually have an agent is sort of like you tell and predefine a certain template on which the work has to be done and that is more sort of accurate.
And we might eventually get there where Chad GPT will actually do this stuff and it's almost accurate.
But I feel like right now we are in this abstraction layer where we have to define XYZ steps and those steps are executed and we fill the gaps with intelligence layer because we have access to your sort of LMs and everything integration.
And I would assume like one of the big challenges for a company like Gleen would be integrations, right? You you might be integrating with hundreds of apps.
And this abstraction can only work if you're integrated well with, uh, you know, different set of apps and today like every company uses like hundreds of SAS tools.
And the like so the stack of number of SAS tools that are been used at, you know, a thousand person companies is like innumerable. Uh, so like talk to me a little bit about that challenge. Like how how how how does that play into all this? Yeah.
I mean, you're you're spot on, number one that the agents, uh, that you build, they have to work on some of your enterprise data. They're going to use some model intelligence to mimic, you know, the work that the human was going to do.
And you also have to take actions, um, into your enterprise systems, you know, potentially to save that work or complete that work.
Uh, so there's a there's a very strong dependence on, you know, your ability, like, you know, if you're building an agent platform, you know, your ability to to actually both read information and take actions into all of your enterprise systems.
And so for Gleen, like, you know, the the good good news is that, you know, we've been working on that for the last six and a half years.
We have hundreds of these integrations, thousands of actions, you know, that we can support across your enterprise applications.
And so that sort of becomes, you know, just the raw material, like, you know, sort of the underlying platform, um, that allows you to actually build these different types of agents.
Um, and and I think that's a it's been like it's interesting like, you know, how hard and difficult it is to actually make make that, you know, get that to work because enter systems are very bespoke in nature.
There's a lot of complexity, um, and the dip the deeper you dig, the more you realize the number of problems that you're to solve. Like one of them is simply, you know, security and governance.
Well, you have all these systems, there's a lot of knowledge information there, but there's also a concept of permissions and access control and you cannot have agent, you know, an agent platform where you let those agents just read any data from any system, take any action that you want.
Like, you know, you have to sort of follow the governance architecture and the rules inside the company.
Uh, for example, like nobody should be able to look at the deep financials if you're a public company and if you're not actually, you know, in in the in the in the finance team.
And so you so you have to sort of go and, uh, like not only build these integrations, but also like, uh, work, you know, uh, work upwards from that like, you know, ensure that you, you know, you handle agent security as a problem statement, that you also solve for like making sure that you deliver the right data to these agents, not like stale or out of date information, which will end up like, you know, you know, these agents are going to make those mistakes.
So so so you definitely have to solve that. But I also want to actually, you know, comment on the first thing that you mentioned, which is that there is a concept of a workflow, uh, and then there's a concept of an agent.
And these are two different things in some ways. Workflow is basically fully deterministic. Right?
It's a sequence of events, you know, that you're going to undertake with, you know, the appropriate branching and, you know, and, you know, and and logic, you know, built into it, but given any input, you know, what is the output is fully deterministic.
And and so what is have, you know, there's and and these workflows are great, but they have been hard to build in the past and AI is actually really making it easy for you to go and build that.
For example, in the Gleen agent builder, you can actually construct these workflows, which are fully deterministic deterministic, but you don't have to actually go and