Dec 12, 2024 · 52m · mad
Dataiku's Secret to Scaling AI in Global Enterprises | Florian Douetteau, CEO, Dataiku
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of The MAD Podcast, host Matt Turck interviews Dataiku CEO and co-founder Florian Douetteau about his background in French tech, the creation and platform evolution of Dataiku, and how global enterprises can effectively govern and scale predictive and generative AI.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Matt holds 22.4% of the talking time here. How this is scored →
Matt as informed peer 4.4 Guest teaching 3.4 Guest disagreement 0.7 Matt pushing back 1.1
speaking balance: gold is Matt, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 36:32 Rejecting Gen AI replacing traditional ML
Florian firmly rejects the popular narrative that Generative AI renders traditional predictive machine learning obsolete, explaining why probability-based business decisions cannot be handled by LLMs.
Hardest push from Matt ▶ 14:56 Challenging platform-first and enterprise-first strategyMatt presses Florian on violating standard SaaS venture wisdom by building a platform layer immediately instead of starting with a single product wedge and selling to Bay Area tech startups.
Biggest teaching moment ▶ 42:10 Dismantling chatbot productivity as false low-hanging fruitFlorian educates Matt on why generic chatbots and email summarizers are false low-hanging fruit that fail to impact enterprise P&Ls, contrasting them with complex multi-step industrial reports.
Matt holds his own ▶ 14:29 Framing platform timing against startup strategyMatt demonstrates sharp venture insight by probing the difficulty of selling a massive platform early versus securing a targeted product wedge in enterprise sales.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Matt as informed peer | Guest teaching | Guest disagreement | Matt pushing back | Why |
|---|---|---|---|---|---|---|
| Welcome and Florian's Global Travels | 2 | 1 | 0 | 0 | Matt establishes a warm, informal atmosphere, asking about Florian's recent global travels and background. Florian shares early memories of programming on an Amstrad CPC 6128 and OCaml in France. | |
| Elite Education at ENS and Shifting to Startups | 4 | 3 | 0 | 0 | Matt demonstrates knowledge of the French tech ecosystem by contextualizing ENS and Exelit as the premier French talent hubs. Florian explains his pivot from pure mathematics research to language models and search engines during the 2000s compute hardware shift. | |
| The Inception of Dataiku and Democratizing Data Science | 3 | 3 | 0 | 0 | Matt prompts Florian on the founding of Dataiku in 2013 during the early big data ecosystem. Florian describes the core thesis of democratizing data science across business and technical roles. | |
| Collaboration Philosophy, Naming Dataiku, and Enterprise Focus | 5 | 2 | 0 | 0 | Matt draws an astute comparison between Dataiku's collaboration thesis and Datadog's bridge between devs and ops. Florian breaks down early miscommunications between business units and data scientists. | |
| Pragmatic Platform Strategy and Non-Tech Enterprise Selling | 6 | 4 | 1 | 4 | Matt challenges Florian on defying standard startup playbooks by launching a broad platform rather than a narrow tool wedge and targeting Global 2000 enterprises early. Florian explains why non-tech enterprises lack engineering capacity to build custom stacks, justifying their strategy. | |
| Dataiku's Platform Evolution and Orchestration Layer | 4 | 4 | 1 | 1 | Matt asks about Dataiku's evolution into an orchestration platform. Florian offers a nuanced reframe, noting that rapid AI/tech innovation acts as a stress factor for enterprises, positioning Dataiku as a dampener layer. | |
| Composability, Building Blocks, and Data Asset Lifecycle | 4 | 3 | 0 | 0 | Matt highlights the end-to-end data asset lifecycle. Florian details his Lego-block philosophy where metrics, models, and agents are managed as long-term enterprise assets. | |
| The Critical Role of Data Preparation and Pipelining | 5 | 4 | 1 | 1 | Matt brings up the unglamorous necessity of data preparation in enterprise AI. Florian explains why point solution startups get stuck when scaling agent fleets without pipelining capabilities. | |
| Governance and Operational Control as the AI Bottleneck | 5 | 4 | 1 | 1 | Matt synthesizes the governance and human process bottleneck in scaling AI. Florian reframes the future AI challenge, asserting that model output quality will become commoditized while operational control will be the primary limiting factor. | |
| Unifying Predictive Machine Learning and Generative AI | 6 | 5 | 2 | 4 | Matt directly addresses the narrative that Generative AI replaces traditional predictive machine learning. Florian reframes the premise by distinguishing statistical risk optimization from LLM generation, arguing both must co-exist. | |
| Real-World Enterprise Use Cases Beyond Basic Chatbots | 5 | 5 | 2 | 2 | Matt frames initial Gen AI deployments as low-hanging fruit like chatbots. Florian gently pushes back on chatbots as false low-hanging fruit, pointing instead to complex multi-step workflows like patent research and industrial safety reports. | |
| The Long-Term Roadmap and Scaling Enterprise Platforms | 4 | 3 | 0 | 0 | Matt wraps up by reflecting on the 10-15 year timeline required to build enduring software companies. Florian compares building a company to raising a child into adulthood over two decades. |
Statements from this episode (19)
Douetteau: Top ENS math students realistically target winning a Fields Medal
“And very smart people. I think you're like in a group of 50 where the top 10 people there are probably expecting to get a field medal in their lifetime and have a decent sense of it.”
Douetteau: Exalead spawned 20 to 40 successful tech startups
“Yeah, 2030, 40 good startups actually came up out of out of it”
Douetteau: Dataiku was founded to address a global shortage of data scientists
“You had a need for the enterprise to do more in terms of data science, An ML, but a scale gap, because there was not enough data scientists on planet Earth in order to fill that gap.”
Douetteau: Adding data scientists does not bridge the enterprise business-data gap
“It should not be solved by adding more data scientists, because anyway, there's not enough data scientists, and if they don't understand the business, there's no point. It should be solved by finding a way for people from the business that are very much into d…”
Douetteau: Silicon Valley groupthink would have killed Dataiku's platform strategy
“A thing that it required then some will or independence of mind, which was easier for us because I'm not sure if I have had the will or the mind power to Keep this line of thinking if I had been in the Bay Area. Because around me, everyone would have had a dif…”
Douetteau: Dataiku intentionally chose not to sell aggressively to tech companies
“But our perspective was that, how do we solve for the problem of every other companies, which will not follow the same route, which cannot follow the same route, and take the perspective of, like, we will never try to sell aggressively to tech companies, per s…”
Douetteau: Dataiku cannot maintain a fixed 3-to-5-year product roadmap
“This constant integration of technologies has been part of the DNA of Dataiku. And it's uncomfortable when you do enterprise software, because it means that like every quarter or so, twice a year at least, you've got something else going on, and you're not con…”
Douetteau: Data science and analytics teams tend to rebuild work from scratch
“When you're building in data science, people tend to redo from scratch, or even analytics and so forth.”
Douetteau: Analytics metrics in data-driven enterprises get reused for up to ten years
“If an enterprise is Becoming data-driven or AI-driven, it means that every piece of analytics is actually bearing lots of value long term because you're rebuilding lots of intelligence in your company based on this. Even if it's a simple metric that you're bui…”
Douetteau: RAG tools and AI agent builders are becoming commoditized
“Ultimately lots of those technology can become, in isolation, pretty much commoditized. Meaning it's not hard to build RAG or agent builder.”
Douetteau: Data science bottleneck is data prep, not model tuning
“What's hard in many data science projects is not the model tuning and so forth, in most instances, in fact, is getting the customer data right and manage all of the edge cases or situation where you've got empty columns or not, and whether they matter in terms…”
Douetteau: Enterprises will run 500 to 1,000 AI agents
“Our vision is that enterprise at large will not have one agent, but like more, 500 or 1000 of them.”
Douetteau: Point-solution LLM capabilities will become platform features
“All of those capabilities could be startups by themselves, but the same way we thought about the platform required for the enterprise back in the day, 10 years ago, we think that for the world of LLM, all of those capabilities will just become features of plat…”
Douetteau: AI model quality won't be the key bottleneck in 2-5 years
“I, for instance, don't think that the problem two years or five years from now of agents will be relevance, per se, as in the quality of the models. The problem will be very boring stuff, such as, is it the right data, or have we actually updated the data for …”
Douetteau: Generative AI lacks statistical consistency for business goal optimization
“Generative AI, as it is built today, is not meant to provide any consistency in terms of how statistically you reach a business goal through the output or through your predictions.”
Douetteau: Enterprises will always need traditional predictive ML models
“You will still always need as an organization to be able to build those models because there are actual business decisions in them. Like actual risk profile with conscious decision in it.”
Douetteau: AI agents will combine predictive ML and LLMs for autonomous tasks
“What we will build are agents that are combining those models or those models together in order to do more autonomous tasks end to end.”
Douetteau: Enterprises must build proprietary AI agents to differentiate
“Then enterprise, in order to differentiate and be, transform themselves, will also need to build their own agents. On everything which is about stitching together the data from one vendor to the other, or the overall process, or the thing they actually built, …”
Douetteau: Building a mature tech company takes 20 years
“I think it's at least 20. I mean, I think it's almost, it's probably very comparable to the number of years in order to get a kid in or even out of college, in and out of college, and maybe being being financially independent, you know? That's kind of like the…”