Florian Douetteau, CEO and co-founder of Dataiku, explains how to navigate this complex landscape. He shares his insights on building a secure governance framework that provides clear guardrails, allowing your teams to innovate freely while ensuring your company's data remains protected and secure.

 
 
 

288 Audio.mp3: Audio automatically transcribed by Sonix

288 Audio.mp3: this mp3 audio file was automatically transcribed by Sonix with the best speech-to-text algorithms. This transcript may contain errors.

Florian:

Some of our customers as they pushed employees to actually discover and create agents by themselves, start to have some form of like agent sprawl, as in, you've got lots of people that created small agents doing this or that, and there is a question of like, who is doing what and for what purpose? So we're working with some customers on essentially helping them having a central access point for agents. And the core idea is that there is a need and an appetite for having more lifecycle management of agents. Hi, and thanks for having me today. Uh, so I'm Florian and I'm the CEO and co-founder of Dataiku and, uh, started the business in 2013. Uh, prior to that, meaning I started my career, uh, more on the tech side, uh, studying math, uh, but, uh, in 2000, uh, which is back 25 years ago, uh, actually started to work on natural language processing and web search and enterprise search technologies and um, meaning all by that time or the this early in the cycle, um, let's say it was very hard to get anything scalable up and running in the space of an LP. You had to build everything from scratch. And yeah, I started the Taikoo, uh, after having uh, uh, had multiple experiences, uh, with companies that were trying to scale their data science practice and were failing to do so. Um, and that this idea that meaning, if you can democratize AI, you can actually bring lots of value to the enterprise.

Craig:

Yeah. Um, and, uh, and tell us, uh, in short, what Taku does, uh, and, and what its main products are.

Florian:

Yeah. We have one product which is in essence, a no code AI workbench, a workbench where you can do everything end to end required to manage your data. No code so that anyone in the enterprise willing to do so can actually touch it and start to work on data and AI. And, uh, it's a workbench that can work on top of, uh, any cloud, any data. And the goal is really to bridge this gap between no code that sometimes is perceived as being, uh, prototyping for domains and enterprise production grade applications. And indeed, we have, uh, our customers are, uh, among among our customers. You've got large enterprise that use our platform in order to build thousands of data products and models and scale that scale that in two ways, meaning scale the number of things you need to build, but also scale the number of people that can participate and understand what is being built.

Craig:

And most recently, you have, uh, what you call an enterprise grade blueprint, uh, for a gigantic AI in financial services. Um, so on this, uh, workbench, uh, you have, uh, it's built for various verticals. And is this, uh, can you talk about this blueprint and how it works? Uh, on the workbench?

Florian:

So, indeed, um, we we built this, um, enterprise blueprint, uh, for FSI with Nvidia, and it relates to the agnostic nature of our platform. As you pointed out, our platform is used equally by a large manufacturer or life science companies or banks. And it makes sense, actually, when you want to do enterprise AI to actually tap into all of these types of sectors, because they all have these issues of big data, complex data, diversified data, siloed data, and people in the business don't understand how you can help them. So that we want to be agnostic in that nature. How can you actually help everyone? But we also want to be agnostic in the in the sense of the infrastructure itself. And so, for instance, we build our platform so that it's super easy to work with the various types of large language models and vector stores around in order to move from just doing ML to do a gigantic ML with our platform. And on that front, we are connected with the Nvidia enterprise AI stack. So that in particular in banks that typically have a regulated constraints, want to be able to host their data and all of that good stuff so that they could actually build this type of solution in an easy manner.

Craig:

And uh, the what are some of you you're saying that every vertical has It's, uh, it's, um, you know, it's it's particular needs. What what is it with, uh, financial services that that, uh, is different from from another, uh, vertical being, uh, built on the platform.

Florian:

I think that, um, for, uh, every vertical there is a particular need to support, uh, a bit differently, uh, security and infrastructure constraint, as in, like where the data is, what is the particular type of security setup you have in place? So for instance, in financial services it's fairly common to have the need of securing data, not just per who has access to it, but also on a project basis, making making sure that there is no leakage of information from one project to another. Because in financial services there are constraints in terms of who you can use the data, meaning you can't actually create any type of like insider, um, type of issues that, uh, you could get into if you don't have proper firewall between data in your system. Um, the way we support that in our platform is through its flexibility, but also in financial services we provide on our platform, um, different types of solutions, as in a way to just have prebuilt projects that help you get value faster because there is, like so many use cases of AI that many of our customers actually want to accelerate their discovery of them. It's a bit overwhelming to work in AI or to have to build AI systems in the enterprise today, because you live between this, uh, high level of expectation coming from your board and leadership and the reality of the field where earning getting to, uh, select one platform for large language model can be an issue in itself. And so our buyers are essentially the people managing this high level of expectation with the constraints of the field and our platform. Help them do that more efficiently.

Craig:

Yeah. Um, and for the agents, um, if you have at, this is there because this is no code. So is there, uh, how do you build an agent on the platform? Is, uh, or or is it something that an internal team would build and then make available on the platform?

Florian:

Yeah, that's something that, uh, the internal team of our customers would build on the platform. And in terms of agents, um, you can build visually agentic workflows in the platform, as in agents that would actually connect to another agent, connect to another tool, go through, um, various steps to achieve the task. And in this platform you can do it by connecting natively all of those platforms to your data and or more recent data, or your more recent models. So you can actually build a fairly sophisticated things, such as an agent that would use a predictive model to actually make a decision instead of like hallucinating that out. Uh, very useful, for instance, in banking, if you want to at some point identify business impactful, uh, situations such as, I don't know, agreeing on a loan. You don't want an LLM to decide, you actually want a good old statistical model to decide this part, for instance. And we also built the platform with lots of, uh, agentic evaluation capabilities. And in the enterprise, it's fairly important to have, um, a good level of auditability and testability of agents, as in having continuously this ability to tie them to the data, analyze the outputs, run type of back testing of agents before pushing them in production. When you when you want to support like more customer facing not so impactful applications, you can be more like I a b test type of mindset with agents and you can iterate in a certain way. But when you're in the enterprise, you need to have a different level of depth in terms of the test you're making, because the cost of one issue can actually be very high.

Craig:

Yeah. Um, and then, um, once these agents have been built, um, how do different people in the enterprise access? I mean, there must be a very strong permissioning, uh, controls.

Florian:

Yeah. Meaning you need permissioning controls that, um, agents can only be accessed by people having access to it. You want to be able to, um, apply the access control of the person? Using the agent to add the agent can do as in if as Florian, I can only have access to these rows in this data set. When I'm using any type of regular BI tool, I set up my it. I want the same permission to be applied when I'm using an agent, of course. And so this is the type of things that needs to be indeed built and solved for in the enterprise. A broader question that our customer has is like, how do you discover agents, as in, um, some of our customers have they as they pushed, um, the, their, their employees to actually discover and create agents by themselves, start to have some form of like agent sprawl, as in you've got lots of people that created small agents doing this or that. And there is a question of like, who is doing what and for what purpose, which is, uh, I think another, I think, key topic these days.

Craig:

Yeah. Well, how do you handle that? I mean, is there, um, a master dashboard that can see every agent in the system. And not only not only that issue, but, uh, you end up, uh, I mean, not simply Agent Sprawl, but you end up with agents talking to agents and passing data back and forth, and that creates a potential security problem.

Florian:

Yeah, potentially. I must admit that in the enterprise, the agent talking to agent type of scenario is still, let's say very early on hypothetical currently. Yeah. Uh, but but indeed, a practical issue is, first having some oversight on everything that is happening from the perspective of it. Can I actually tell what's happening? What were the projects, what are the ROI, what are the risks and so forth. And so for that meaning, it's essentially providing some governance framework and ability to to gather data from everywhere. The second thing which is funny is that we're working with some customers on essentially helping them having a central access point for agents. And the core idea is that there is a need and an appetite for having more life cycle management of agents. And then it's kind of like back to this question of ideation. Um, as someone in the business, I could have a great idea of like what an agent could be doing, and I want to be able to prototype this quickly. So and with agent prototyping, it's very easy. Like you prompt, you say this and this data and it's kind of like working. But lifecycle means that in the enterprise you would ideally have a step before this thing becomes shared to 1000 people in your organization, a step where it can check it, validate it, operationalize it with like, the proper data, the proper permission and so forth before it gets and becomes something official. And doing that you avoid two things. First is like issues in terms of accuracy, security or falseness in the enterprise. And, um, second, you create more trust because you can ideate things. But having a proper vetted catalog of enterprise agents that people start to use, uh, for real. And we've seen, uh, with this pattern, I think, good progress with our customers. And we think there is, uh, for us, a way to actually further help there.

Craig:

Yeah. Um, the, uh, you know, you said that not many people are doing, uh, agent to agent, uh, uh, systems yet. Um, why is that? Is it just is it a matter of, uh, trust? Because certainly, as a journalist, that's all I read about agent swarms, agent societies. Yeah.

Florian:

So? So, yeah, I think because it's, um. So bear with me. We have and we support and we have lots of use cases where an agent is calling to another agent. But in fact, the reality is that when you're it's almost a development pattern where when building something significant, you start by building small things and a small agent to be able to test it, and then another one, and then another one and another one. And then your real use case, you add a wrapper on top thing. You say you will do A, B, C and D and that's it. And you have built intermediary agent for each in order to actually be able to test the thing as someone building the system. So it's not yet agent to agent, as in like an agent that was uh uh. Knowing, uh, what are the best customers and another and running on its own and an agent thinking about the talent pool in the company and the two agents talking to one another to discuss and devise what should be the strategy of the company. We're not there yet, and we might never be. What we're talking about are genetic systems that can automate significant tasks, but indeed need to need to be built in an incremental in an incremental manner. And we've seen good success of that in the last, um, couple of months.

Florian:

Uh, we've seen advertising company using our platform in order to automate their campaign management and be able to just be able to do more, uh, campaign per employee and per manager of such campaigns, like by a factor of 50, which is amazing. Uh, we've seen retailers using agents in order to solidify, uh, access to information in the shops, as in how people can actually access inventory, can provide recommendations to customers on site and so forth by having more fluid interaction and just unlocking all of this data, which is into, uh, ERP and product catalog and so forth, that are too hard to access by, uh, people on the ground. And we've seen manufacturers and farmers using agents into their some of their complex R&D cycles in order to massage lots of information about state of the art into multi-step process, in order to refine this information and build internal drafts that serve their and accelerate their R&D practice. And so all of those things are not trivial. Meaning they took some understanding of the business to be built and they are very early, meaning it's a compared to what will agents compare to what agents will be doing? It's still very early, but very promising from my perspective.

Craig:

Yeah. Um, what is the most complex? Because a lot of what you mentioned sounds, um, a little bit like, uh, like search. Right? Or it's surfacing information that already exists. Um, you know, maybe, uh, Are tied to a chat bot or something. But what's the most complex, uh, agentic activity that that you've seen to date that's in production?

Florian:

Um, I think that in terms of, um, first, in terms of scope or number of employees you can touch, for instance, let's have things that are field activities, as in supporting, for instance, a retailer are indeed search. Search as in like, um, can you tell me what what should I recommend to this customer that is in front of me? It's search in a sense, but like when you dig into what you need to do to do there, you need, uh, the proper product catalog. You need to understand the customer question. You need actually also to know what to recommend based on prior history and all of that good stuff. So you need a supercharge, kind of like Amazon recommender system with a product catalog, but like being able to deliver that in front of the customer, kind of like in real time. And so those systems are maybe not complex in the sense of like the number of steps is not that many, but you need to connect them properly to have the level of experience you need in order to, to, to, to, to deliver the value. And so in that sense, they are sophisticated. The type of thing that we've seen that were the most sophisticated perhaps where uh, into uh, intellectual property, for instance, where you could start using agents in order to understand the state of the art, what was patented or not in your space. Um, mix that with your internal databases and use that in order to support or accelerate the management of your, uh, IP portfolio and where indeed you need to understand the data and your space and the science of your space in a very sophisticated way. And again, you could say it's search, but it's not just search. It's like synthesizing all of the information of your domain in a specific context in order to understand what's new versus not new. Which is indeed a complex multi-step setup where within systems that are literally calling. Thousands of times LLM in order to answer one question.

Craig:

Yeah. Um. And um, in terms of, of agents. You know, I, I was at a conference a couple of months ago and the guy was talking about, you know. Speculating about, uh, a headless organization someday that that there will be agent managers and. For various departments. And then beneath them, you know, groups of agents, uh, doing all the various tasks that the organization needs to do. Um, and then, you know, maybe, uh, Overlord agent that watches everything. And. And.

Florian:

And. Do you want to be the Overlord agent, or should I be like. It seems to be a good position to be the Overlord agent.

Craig:

Yeah. I suppose. Uh. Uh, but but could you envision that sort of thing happening? And what kind of more complex actions do you see agents being able to carry out?

Florian:

Yeah, I think that's, um. Indeed, it's an easy way to map Agentic activities, to map them to the organizational structure of a company, because it makes it easier to for people to discover them, to understand the role and responsibility over them. Uh, so indeed, we see our customers doing that, as in, in a given team. I should think about how many employees and how many agents. And that's. Yeah, that's interesting. And it's, uh, when you interesting agents are agents that do not just do a simple task, but that essentially accelerate or support a given process inside a company. And indeed, it's easier if you map if you. Wrap all of those agents that do, um, processes in a given domain, let's say sales or R&D or whatsoever in a master agent that can route to them. Because then you simplify the way people in the company can actually access the information and start using them and so forth. And so it's a pattern of, well, essentially delegation like like in real life, like in human life, uh, you've got the sales agent having different subprocesses and depending on what you ask, it will delegate to the other one. They will do its steps. So it's actually it's actually, um, the way people do it then uh, to to maybe answer the second part of your question in terms of the, the capability and the capability space.

Florian:

I think it's interesting to realize that agents, when you build them, you have some expectation of their qualities that for some of them mirror a bit human qualities, as in, in some scenarios you want agents that are very good at just repeating a given task, like it's a process and you want it to go through in a very predictable manner sometimes, and maybe you do to. I use agents for creative purposes to give me ideas or generate an image, help me with a slideshow or a presentation. So you use it as a as a creative head sometimes and meaning Literally all the time. I use agents for more like productivity and doing some assistant type of work, as in like meeting notes and whatsoever. And and it can help also with collaboration or some aspect of it. And so and there is some agents that are more maybe at the equivalent of a management of a function or management of a process type of capacity, as in you, uh, the you expect the agent to support or automate, uh, an actual business decision, as in, do I resupply my stock or not? Do I prioritize this product or not? Do I change my prices or not? This type of things.

Florian:

So those those things are different qualities. Meaning you don't expect a given person, usually in an organization, to be creative and very, very good at execution and very good at taking business decisions, or be a great developer or be a good production or manager or a great facilitator, and you don't expect the qualities, all of those qualities, to be the same person. Most of the time I don't actually. And so and I'm neither of those so the, the the for the same reason I think that when we will be building agents more at scale, we will start realizing that depending on the task, you expect different things from agents, and maybe you start using different types of system depending on the type of agents. Um, you, you want to use or create. And that's my the way I think about this ecosystem moving forward. It's very early because agent is it's like we're in 95 or 97 maybe and agents is website and MCP is a gopher or FTP. Not really. Not really HTTP and certainly not Https. Yeah.

Craig:

Yeah.

Florian:

So that's that's.

Florian:

And so we say agent but actually we mean. Blog or e-commerce or whatever else. Like you will have many type of agents, many categories of agents that will be actually created.

Craig:

Yeah.

Craig:

Okay. Can you talk a little bit about the uh, uh, dada, um, uh, mesh and, and how that, uh, works with, with the agent platform. I mean, is. Does the user decide have access to all the various, uh, open source or proprietary models and and make the choice? Or is there an agent in, uh, the, uh, ecosystem that that, uh, decides what model is best for what task?

Florian:

Um. Great questions. Uh, indeed. We built the. So what what I think in this gigantic ecosystem is that, uh, from an IT perspective, there will be more and more the, the, the question of abstraction of large language models because from a cost or management perspective, they are very peculiar type of systems in terms of their costs, their characteristics and so forth. And the fact that they you send lots of data to them, potentially, and so on. It's a new type of system. And so from my perspective that I could not entreprise at least large enterprise will end up having this view of, uh, having a way to abstract away this access to all of the LMS. And there are at least three big ways to consume LMS today. You can use directly the services of the OpenAI and anthropic of the world. Um, you can use, um, them hosted into your favorite cloud platforms, either the same models or more open source ones, and you've got lots of variants in data platforms and cloud platforms to actually host agents. And you can host them yourselves as your own servers or services, your end on a purpose manner to just, uh, self-host, uh, models. And you've got good reasons to do at scale. There will be good reasons to do all three of them potentially, maybe depending on security and sovereignty type of questions. So for nationally, we said like or products should be built so that from the customer perspective, it would be very easy to manage and switch from one to the other and to have a more holistic view on, the risk, the logs, the audit.

Florian:

Um, so the costs, the security aspect and security for agents is not just who has access to what, but it also, uh, the list of forbidden words or topics you should not send to this agent or that other agent. If I talk to about maybe you want to set up your system so that if you talk to or use a customer name, you should never do it on a public server agent and always use a self-hosted one? I don't know, for instance. And so that's the type of, uh, very IT mindset we use to, to, um, to, to build, uh, to, to, to build our product. And I think it's actually very generic, as in like this mindset should be the, the, the type of mindset that enterprise will have when they think about those, uh, AI systems, um, at scale. Uh, to your second question, indeed, today we don't think and we don't provide a magic, uh, Model that is, um, choosing the best model. And I think it's early to do that because it's more like on a per use case basis where you choose the model. When you create the use case, instead of doing it dynamically, it might be the case down the line that we do that, but it's a bit early in the ecosystem to do it right now.

Craig:

Yeah, you were talking about security and you have a function on the platform to set guardrails. Um, is is that essentially, uh, a prompting exercise?

Florian:

Um, Gabriel is not just a prompting exercise. It's also to have, um, dynamic guardrails are about having some form of filters that you can have on the data in or the data out, as in what you send to the LM or what you receive to from the LM. And this can be a mix of, uh, guardrails that are based on just keywords Our dynamic list of keywords, or the galleries that are based on topic guidelines on keywords could be like, I don't want my customer name, customer customer names. To be sent out or I don't want something looking like, um, security, social security number to be sent out. For instance, the type of guardrails, the, the kind of related to topics is, for instance, I don't want my LMS to be providing back any type of, uh, else advice or financial advice, for instance. And so you typically want to set up those guardrails, um, in a meaningful way so that either you do them universally for any type of usage within the enterprise or per category of applications. You allow one versus not the other. And so it's important for the enterprise, well, regardless of whether it's urgent use cases or more basic use cases of LLM, there is a good reason for enterprise to have a good control on those guardrails from the perspective of a compliance and compliance and security.

Florian:

Yeah.

Craig:

And on the platform, um, is it, uh, is there a conversational interface to implement these things or, um, on, uh, guardrails? Is there a list of checkboxes that the user can check? I mean, how you were saying it's largely no code. So how is a user, uh, telling, uh, you know, building this agent?

Florian:

So, so when we, when building an agent, uh, user would actually tap into, um, a LLM set up by the enterprise, it would drop in the list, the one that the Enterprise provided and was actually given to him, and those LM, when used by the platform and through the platform are already secured, as in, for instance, uh, the someone in the IT of the company would have said in this company we just use, let's say OpenAI and we can use it for three different reasons. One is like super secure and one is more like internal fluid and one in the middle. And depending, we set up three types of guardrails that are different for each level and maybe for the secure one. We are super strict in terms of like no health advice and no financial advice and no talking about chocolate. I don't know, whatever. Uh, and in the internal one we can do whatever. And so depending on their permissions, the user would have access to level one, level two or level three, let's say, or like what is being configured by the enterprise. And then in order to build the agent they can prompt their way and have access also to a set of tools, and same tools are paired a setup of a set of permissions related to what they can access to and what they can give access to in terms of applications and data. And so this way, a user can build an agent by prompting and sequencing prompt and and accessing to tools. Then the user can give access, can test this agents and give access to this agent to other users. And then indeed total security is about the other users when accessing the data would also have a set of security constraints being applied to them that is specific to them, so that you actually don't. Leak information by creating agents.

Craig:

And in the background, what's happening? Um, um, are there blocks of code that are pre-written, that are being assembled. Is is actually writing code in the background? Um, or are there, um.

Florian:

Uh.

Craig:

A catalog of agents that, that Iku is pulling from, uh, and presenting to the user?

Florian:

Yeah, we have a catalog of we have a catalog of tools that contain code. Uh, we've got, uh, the prompts are enough to actually create, uh, uh, create create the agent and then Dataiku would generate and the tools within Dataiku would generate code in the sense of, for instance, if it's about querying a database, it would generate SQL code, for instance, in order to actually support, um, the proper, the proper activity as a broader tool beyond agentic, meaning that Iku is also a platform to do any type of uh, ML and analytics. In which case it's also no code. As in, you can use it to transform your data and build your models and so forth. And in that process, you can also decide to generate code if you want. And we see that as part of the lifecycle of the platform. Sometimes our users like to be able to generate code in order to actually go one step further in terms of customization and adaptation to their use case. And so over the years, we've built the platform in a way so that it's no code, but also code friendly. And and it's back to this topic of collaboration. And a lot of people in the enterprise needed to work together in order to achieve goals. Um, we built the platform more and more so that it can be a collaboration place between the technologists and the non technologists in any given company, because I think that the lack of collaboration between both camps is usually what's creates struggles in organizations. If geeks and non-geeks could be friends. What would be better? And and so indeed, we build a no code platform where you've got a full fledged VS code and Python notebooks type of environment so that coders can also thrive.

Craig:

Yeah. And I mean, that's interesting, uh, just thinking about the organizations that I've worked in, I mean, primarily the New York Times, uh, the, the the people. Yeah, there's there's a huge, uh, divide, at least in my day, between it and, and and users. Um. In how, in your experience, um, the users of the platform Firm at enterprises. What percentage are business experts or, uh, you know, operational, uh, people and what percentage are the IT department? I mean, is this really getting out?

Florian:

Yeah.

Craig:

Of the IT department.

Florian:

Yeah.

Florian:

In enterprise it's typically 75, 80% that are in department and um, 2,025% that stays uh central i.t. And so the way I think about it is that, um, let's say you've got IoT and business. Great. But in fact, in it you've got a bunch of people in it that are very interested in working with the business and in the business. You've got lots of people that are more, uh, data experts or perform or Excel gurus or like, really interesting into the quantitative side of things. And so the goal of our platform is actually to bring those two groups together, the people in the business that are actually more like quantitative. I want to get my hands dirty doing AI by myself type of mindset and people in it, which, uh, don't care about, uh, building, uh, um, it ivory towers, but actually want to build, uh, it enabled for the business enabling the business type of mindset, which is not all it like. Bear with me. It's like and so but if you take those two groups together and you can actually have them share a platform, or you actually can achieve a great thing for the enterprise, which is to break silos.

Craig:

And is that something that you advise corporations on? Because I can imagine someone buying the, uh, the The platform for the enterprise. Maybe you know, the CISO or the CTO or even the CEO and saying, here, I want everyone to use this. Um, and it's it sort of sits idle on most people's, um.

Florian:

Screens. Uh, yeah.

Florian:

I think that there is what we advise enterprise is. In fact, a modern version of what should be a center of excellence for AI and center of excellence is like loaded word and like fairly old as a, as a, as a as a concept. But it's indeed effectively creating this place where you. Move it as an enabler of the business. And you create a dynamic where people can try by themselves, where they communicate, where you, you even have like sometimes hackathon. Real hackathon within enterprise where they can like innovate and test and so forth. Uh, where uh uh, instead of people being like, oh my God, I don't have access to data. You create a proper desk or way for people to ask or people to help on that front. And so that's part of the dynamic. We indeed help create creating the enterprise. And so this way we were successful in creating like fairly, fairly large groups of people by the thousands, sometimes like thousands of employees. Building models on the platform and so forth. And indeed, you need people in central ID to help that by having a proper center of excellence, because it does not. In large enterprise, nothing significant happens organically. In fact, large enterprises are very good at killing organic stuff. That's just a fact. I think that's a good thing, actually, because you need focus and vision and drive and rigor and discipline if you want to scale any type of organization. So for this type of collaboration to happen, indeed it needs to be in large enterprise, a proactive, proactive initiative.

Craig:

Yeah. I mean some people I've heard advise that every department have an IT person embedded or a tech person embedded in the department that everyone can come to, to. So it's kind of a dispersed IT department across the enterprise. Uh, that makes a lot of sense to me, because you're not going to have your you're going to have varying levels of skill and interest, uh, in the department. And if there isn't a dedicated person, um, and everyone starts coming to, you know, Joe in sales. Who knows how to use this stuff. Pretty soon his job is going to be, uh, just helping other people. So, uh. Yeah, I mean, that's that's an interesting challenge. Um, and and how does, uh, how do you sell the platform? Is it buy seats or.

Florian:

Yeah, it's bases, it bases.

Craig:

And, uh, and and how to enterprises generally distribute those seats. Is it, uh, I mean, are there some enterprises that give it to everybody above a certain level or, um, is there like one seat per department? How do you see people organizing that?

Florian:

Yeah, well.

Florian:

There are some enterprises giving it to everybody. Um, but not enough yet. Uh, but, uh, the most common pattern is that you start having a group of users into one given team, and then you expand to another by group, typically not individual users. Um, the goal being to modernize the way you do advanced analytics, ML or AI into this group and, uh, having a consistent set of use cases and data to work upon and consistent way to support this group. So imagine you've got a a big department of, uh, 1000 people. You would have a few dozens maybe kickstarting, using, uh, using Dataiku. That would be more probably the people in, uh, various form of ops function and data support or AI support function within, within that, within that department. And, uh, it would actually support them by making sure that, uh, their first use cases and access to data is, uh, properly Supported and they've got a good shared view on what are the use cases and the roadmap for that. And at the end of the day, the goal, meaning the goal for us is just to multiply the number of AI projects that people can deliver, that enterprise can deliver.

Craig:

Yeah. What's the most common, uh, agent that you see people building? And I'm sorry I'm so focused on agents. I know that you guys are not only an agentic platform.

Florian:

Yeah, yeah, uh, an agent. An agent. It's the shot are still to be called because we see lots of variation by, uh, per sector. Um, so to give you, uh, to give you to give that to you in a nutshell, for instance, uh, in manufacturing, we see some appetite to use agents to help with some of the forecast and demand and maintenance schedule and, uh, accelerating, uh, accelerating those, uh, those those activities more realistically. And in banking, we see quite a bit of appetite to rethink the anti-money laundering initiative, for instance, because it's aggregating a lot of information and in a very time consuming activity. Uh, and, uh, and so it depends. Yeah, I think there is not like one agent to rule them all. I think it depends a bit on the sector. Um.

Florian:

Yeah.

Craig:

Um, I mean, the reason I ask is I've got an agent to help me go through my emails. Um, and that seems like the most obvious use case. You know, I get tons of emails, and I want to.

Florian:

Oh, yeah.

Craig:

You know, answer some and and ignore others. And it does a very good job.

Florian:

Yeah.

Florian:

But my yeah, our focus at the telco is that well what we believe is that lots of productivity agents, for instance, will be delivered by existing application or OpenAI and anthropic down the line.

Florian:

I see.

Florian:

Yeah. And so that's not. And so what we focus on are the agents that are related to, uh, well, this type of core business process on which you actually need to control very well the data in the data out and why you took another decision. Not saying that your emails are not important, but I mean, uh, for the enterprise, productivity is, uh, let's say, uh, an individual per individual topic. And to some extent, productivity in email is like general purpose type of application. The the best could be the same for every almost for everyone on planet Earth. But for a given enterprise um or you if you're a manufacturer or you actually maintain or not, your maintenance schedule, for instance, is very, very specific to you as a manufacturer because at the core of your business. So these type of agents. Um, I think are the ones that we want to focus on, because I think that would be the one that would be the the hardest to crack in the enterprise in the long run.

Craig:

Yeah, yeah. You're right. Because the productivity you save, uh, you know, half an hour here, half an hour there. Um, it's it's hard to convert that into a real gain. Um, you know, the the individual employee may not necessarily contribute that half hour save to the enterprise, but if you focus on the core business, uh, then you're getting real, uh, um, a real impact. Um, how how, uh, you're you're a global company. Um, is this space, uh, crowded? Uh, I mean, you guys have been doing this for a while. Uh, how do you see the the industry developing your industry? I mean.

Florian:

Yeah, I. Think that there is, um. I think that the. It's a funny space because I still have the perspective that in the enterprise, people in the business are still underserved, even if, uh, even after a tremendous investments in AI. But lots of investments in AI have been pretty much focused on building data centers and cloud and data platform and developer tools, various kind. But ultimately, people in the business wanting to do things by themselves have not received, um, the same level of investments. And I.

Florian:

Think.

Florian:

That still a space which is, um, um, early compared to, to to what it should be. And so that's, that's the way I think about my, my, my, my, my ecosystem.

Craig:

Yeah. Uh, I mean, you're you're quite a bit younger than me. Um, I'm not, uh, I'm guessing you are. And, uh, um, uh, when you look at, uh, you know, I saw Sam Altman was talking recently, and he was talking about the 2030s, and, uh, I thought that was interesting because in my life, I've never thought, uh, you know, you can sort of have these vague dreams of what the future might hold, but I haven't thought very concretely, uh, you know, a decade ahead or thought of it in terms of a decade. And to me, that was very interesting. What are the 2030s going to be like? Uh, as opposed to the 2020s? Um, and do you have a view of, do you do you think in those terms like, wow, in the 2030s, every enterprise is going to have agents, uh, in its, uh, core business?

Florian:

Yeah, yeah, I think that.

Florian:

That's it's um, to me, there is a question of like urgency, as in, um, and two level of urgency. You've got the usual one, which is, as you know, if you remember internet technology, there was like this five year gap or ten years gap between adoption in consumer land versus adoption in enterprise land. And the fact that you, you, you had access to tools and technology or had the perception of as a, as a customer, as a user, um, as a consumer that you did not have as an enterprise employee. And so this gap and so there is one urgency which is like the play catch up. And in the open AI anthropic of the world and Google and even Apple, they will keep pushing great things for day to day productivity. And there will be some catch up. And like good old enterprise. Great. And that's one aspect of it. In the next five years, this race, the second one is also a race of, uh, well, sheer competitiveness, as in there was lots of discussion about, like, digital native, uh, overwhelmingly winning over traditional, uh, enterprise, which was, um, I think a thing back 20 years ago. And in the grand scheme of things, it happened a bit, but not that much. Of course, uh, media is very different today compared to where it was before.

Florian:

Of course, e-commerce and some aspect of retail changed a bit. But in the grand scheme of things, not that dramatically in most sectors of retail. But so there is this question of whether in the next five years you have another wave, but more actually bigger wave of disruption of traditional enterprise by newcomers because they will be able to provide the same service but like just 30 times cheaper or 30 times more efficient with AI. And I think that for most entrepreneurs, there is this realization that if they are not read enough so that they can defend themselves when the real competition begins, they will have to face. Well, uh, potentially, uh, well, understanding they are no longer a relevant business. And so this has maybe not happened that much, um, in during our lifetime. It made me it made me this type of, uh, drastic change of the economic principles. And fast enough might have happened to people in the 20th century, the first half of the 20th century, in terms of, like, dramatic changes of the economic principle that change where the the companies or the regions or the whatever winning versus not. And this actually might happen in the next 5 to 10 years for us.

Florian:

Yeah.

Florian:

And so which is why I believe that there is urgency for all enterprises, and especially traditional enterprises, to be state of the art in terms of AI and to enable people in their business and so forth. Because even if it's a very let's say I'm making lots of fluffy remarks about an uncertain future, trying to play like sad Batman about being a futurologist. But I still think that enterprises should hedge our bets by making sure that they can compete in AI land if they don't want to be overwhelmingly disrupted.

Florian:

Yeah.

Craig:

Yeah, yeah, I agree. I mean, it's it's a little more fundamental than, uh, I remember the, uh, you know, the early days of the internet, there was A big divide between companies that had web pages or a web presence and those that didn't. Um, and, and everybody caught up and I don't think anybody, uh, went out of business because they didn't adapt to the internet fast enough. Uh, but the Agentic AI and the Agentic stuff, uh, you're right. I can see newcomers in traditional industries popping up. Um, that can just operate far more efficiently. Um, I've been talking to people, uh, uh, in the, uh, distribution space, you know, about warehouses and robotics. And you can see that if, if, uh, that industry, it's all going to be, uh, the, the leaders are all going to be those that adapt and adopt robotics and AI. Um.

Florian:

So indeed. And we've got great customers there. And they not only are using, um, AI for robotics, but they are also using AI for every aspect of their business to do their demand planning and real estate planning and optimize their back office function and so forth, because the competition in that space will be acute in every aspect, every aspect of the business. So you can't just be running the business out of Excel, even if it's real estate. You need to be, uh, uh, very agile, understand about the market and the pricing and the efficiency of your sales force and your back office, and understanding how you can manage, like hybrid use of your warehouse that could be reconfigured depending on the type of demand and the energy supply. And like, yeah, it's, uh, it's becoming a different game.

Sonix is the world’s most advanced automated transcription, translation, and subtitling platform. Fast, accurate, and affordable.

Automatically convert your mp3 files to text (txt file), Microsoft Word (docx file), and SubRip Subtitle (srt file) in minutes.

Sonix has many features that you'd love including world-class support, automatic transcription software, share transcripts, upload many different filetypes, and easily transcribe your Zoom meetings. Try Sonix for free today.


 
blink-animation-2.gif
 
 

 Eye On AI features a podcast with senior researchers and entrepreneurs in the deep learning space. We also offer a weekly newsletter tracking deep-learning academic papers.


Sign up for our weekly newsletter.

 
 

WEEKLY NEWSLETTER | Research Watch

Week Ending 9.21.2025 — Newly published papers and discussions around them. Read more