May 31, 2024 · 1h 13m · mad
From Tiny Romanian Startup to Global AI Automation Leader | Daniel Dines, CEO of UIPath
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 UiPath co-founder and CEO Daniel Dines about the company's evolution from a small Bucharest startup into a global enterprise AI and automation leader. Dines shares insights on product pivots, platform architecture, global expansion, go-to-market execution, and the integration of specialized Generative AI models.
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 15.2% of the talking time here. How this is scored →
Matt as informed peer 2.2 Guest teaching 2.9 Guest disagreement 1.1 Matt pushing back 1.1
speaking balance: gold is Matt, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 49:05 Dismissing general LLMs for all tasks
Daniel directly reframes Matt's summary about dominant LLMs doing everything, stating 'I don't think so' and arguing that nature uses dedicated, efficient sub-models rather than heavy general cognitive engines for standard tasks.
Hardest push from Matt ▶ 33:10 Testing customer drag on product roadmapMatt pushes back on Daniel's description of intense customer focus by asking whether bowing to demanding Japanese clients pulled the engineering team away from their planned horizontal product roadmap.
Biggest teaching moment ▶ 49:05 Cognitive engine vs specialized modelsDaniel educates Matt using a tennis learning analogy, showing that human brains shift tasks to specialized unconscious subroutines once learned, proving why enterprise AI needs specialized, cheaper models over massive LLMs.
Matt holds his own ▶ 18:49 Reframing early screen scraper as computer vision AIMatt demonstrates technical vision by connecting UiPath's early screen scraping and OpenCV integration directly to modern enterprise computer vision AI, showing a strong grasp of technical lineage.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Matt as informed peer | Guest teaching | Guest disagreement | Matt pushing back | Why |
|---|---|---|---|---|---|---|
| UiPath's Humble Beginnings in Bucharest | 2 | 2 | 1 | 1 | Matt opens by asking about UiPath's founding story in Bucharest. Daniel gently clarifies that it didn't start in his apartment but in a tiny rented apartment, setting a collaborative and conversational tone. | |
| Pivoting to Low-Code Automation Development | 2 | 3 | 1 | 1 | Matt asks if pivoting to no-code was about making tools for business people. Daniel slightly reframes the premise, explaining that it was really aimed at technical business users rather than pure non-technical business users. | |
| The 2014 Inbound Call and BPO Inflection Point | 1 | 2 | 0 | 1 | Daniel shares the story of an unexpected inbound phone call from Chennai that introduced them to the RPA category. Matt listens and prompts Daniel to expand on design partners. | |
| Scaling Through BPOs and Computer Vision | 3 | 2 | 1 | 1 | Matt observes that UiPath was effectively an AI company from the beginning due to its computer vision approach. Daniel agrees and elaborates on combining OpenCV with low-code automation. | |
| First Venture Capital Round and Board Philosophy | 2 | 2 | 1 | 1 | Daniel discusses raising venture capital, maintaining board control, and encouraging transparent access between board members and employees. Matt provides supportive comments throughout. | |
| Relocating HQ to New York and Global Expansion | 2 | 2 | 1 | 1 | Daniel explains the decision to relocate headquarters to New York rather than London or San Francisco due to global timezones and financial service clients. Matt validates the NY ecosystem choice. | |
| Breaking Into the Japanese Market & Customer Centricity | 2 | 2 | 1 | 2 | Matt asks how customer centricity manifests and subtly probes whether customer bug fixes pulled UiPath off its core product roadmap. Daniel explains their horizontal architecture prevented custom code fragmentation. | |
| Understanding the Process Discovery Layer and Process Mining | 2 | 3 | 1 | 1 | Matt asks Daniel to define process discovery and process mining concepts for the audience. Daniel explains how system logs generate visual process graphs to identify enterprise bottlenecks. | |
| Exploring Task Mining and AI-Powered Communication Mining | 2 | 3 | 1 | 1 | Daniel explains task mining desktop agents and AI-powered communication mining derived from their Reinferred acquisition. Matt asks focused follow-up questions to clarify integration points. | |
| Examining the Automate Layer and Action Execution | 2 | 3 | 1 | 1 | Daniel details the automate and operation layers, focusing on Orchestrator's ability to scale tens of thousands of automations reliably. Matt listens attentively. | |
| Generative AI Integration and Specialized Enterprise Models | 3 | 4 | 1 | 1 | Daniel breaks down why fine-tuning encoder-decoder models like T5 outperforms pure decoder LLMs for semi-structured documents. Matt engages on the technical mechanics of generative AI integration. | |
| Cognitive Science Analogy for Domain-Specific AI Models | 3 | 4 | 2 | 1 | Daniel rejects the popular industry narrative that general LLMs will solve all enterprise problems. He uses a cognitive science tennis analogy to educate on why specialized, efficient models are required. | |
| AI Agents and Next-Generation Enterprise Automation Strategy | 3 | 4 | 2 | 1 | Daniel pushes back on LLM hype, arguing that enterprise adoption requires predictability over raw intelligence. He highlights human-supervised digital assistants over immediate fully autonomous agents. | |
| Organizational Structure of UiPath's Global AI Teams | 2 | 3 | 2 | 1 | Daniel outlines UiPath's distributed AI team structure and forcefully argues that relying purely on indirect channel partners was a huge strategic mistake for competitors. | |
| Building and Scaling Enterprise Sales Leadership | 2 | 4 | 1 | 1 | Daniel articulates the structural tension in sales leadership between operational rigor and customer-centric focus, educating Matt on why technical founders must learn to love sales. | |
| Founder Journey, Scaling as CEO, and Continuous Learning | 2 | 3 | 1 | 1 | Daniel reflects on personal CEO growth, imposter syndrome, and learning during the public company transition. Matt closes with appreciation for the founder journey. |
Statements from this episode (34)
UiPath's first customer paid $10,000 without negotiation after Dines guessed the price
“And he asked me, how much do you want for a license? And I thought, look, I had no idea. So, In a, on a, you know, on a hunch, I said, like, 10,000 dollars, and next day, basically, he wired the money without any negotiation, nothing.”
UiPath made at most $200,000 per year until 2013
“So for a few years, until like, 2000 13, We were basically an OEM SDK business. Not very lucrative. I think most, I think at most we made like 200 K per year, which is kind of enough for it. It was like five people in the company, five developers.”
Dines: Tapping technical business users was key to UiPath's success
“This is one of the pillar of our success. We tapped into a rather large population of people that are not necessarily pro devs, but they have a basic programming knowledge, and they can easily be trained into our technology, and they can deliver automation.”
Dines: Automation development is simpler than full-blown software engineering
“Because an automation is a species of application development, but it's way simpler than to deliver a full-blown application. You don't have to understand so many concepts about reusability and security and all this, you know, the whistles and the bells of tru…”
Dines: Most of UiPath's current product relies on ideas from 2011–2013
“Even today, most of our the way our product works is based a lot on the ideas, you know, that we incubated during these two years.”
Dines: Founders should not build a product in search of a market
“We build a product in the search of a market, actually, but unexpectedly we found the market, which is not, I would not recommend to anyone do this, but that, that happened.”
Dines: Automation is the only way BPOs can deliver ongoing savings
“So they initially start with some kind of business process engineering, which is pure, you know, it's working on the flowchart and understanding how you can remove unnecessary trips in the process. But after that, automation is the only game in town.”
Dines: Early UiPath automated tasks in minutes that took days in competitor software
“It would have taken days in other software, using very cumbersome process, it's days.”
Dines: Earlybird Digital East accepted common shares in UiPath seed round
“They ended up signing common shares in the end.”
Multiple VCs passed on UiPath because Dines demanded founder control
“There were a few funds that passed on us. And it was interesting, not valuation, but it was, they couldn't they couldn't accept to, leave me the control of the company.”
Dines: UiPath founder maintains a controlling stake in the company
“Even today I have a controlling stake in UiPath.”
Dines: Lacking a local market forces founders to think globally from day one
“Starting out of Romania has the advantage that Romania has no market, so we had to think globally, so, and remotely.”
San Francisco is a poor location for running a global business
“San Francisco was, it's too far away. I think for our global business, timezone doesn't help you when you do global business. It's not really so well connected globally, I think from, in, from flying perspective.”
UiPath generated more revenue in Japan than the US in 2017
“Japan, in 2017, was a country that brought us more revenue than US.”
UiPath has never written custom code for a single customer
“Even today, I don't think we have a line of code built for one customer.”
Dines: Large enterprises lack a single repository of truth for processes
“Because even if it's counterintuitive, in large enterprises, there is no single repository of truth. And knowledge about what they are doing.”
Dines: Building too large a system from day one creates software spaghetti
“Because if you start Too big of a system. Day one, you'll end up with the spaghetti and nothing works.”
Dines: UiPath Orchestrator scales to hundreds of thousands of automations
“So then we've built all this, a product called Orchestrator, That is capable of scaling 100,000 of automations, or thousands of runtimes, and make the communication between them via queues, and you have, it's a pretty sophisticated system.”
Dines: Document understanding is one of UiPath's fastest-growing businesses
“In one of our fastest growing business is this document understanding business.”
Dines: Encoder-decoder models beat decoder-only models on semi-structured documents
“We believe that encoder-decoder models are better for when you deal with semi-structured documents than decoder-only models.”
UiPath's specialized models will outperform GPT-4 on enterprise document tasks
“But in the end, we will beat, you know, any performance of GPT-IV or whatever, and they would be extremely reliable.”
Future AI systems will rely on small, dedicated models for routine tasks
“I don't think so, because I think nature is pretty smart. And if you look at how our Brain operates. We have a very good cognitive engine, but we are not using that cognitive engine to do trivial tasks. It's way too expensive. If you want to learn to swim or t…”
UiPath's DocPath and ComPath AI models are T5-based LLMs trained on company data
“Yes, they are yes, this is what I refer to. They are these T-five based LLMs, but trained with a lot of A lot of the data that we gather.”
Dines: Many UiPath enterprise customers use multiple LLMs
“We discovered that many of our customers actually use multiple LLMs.”
LLMs cannot be used autonomously in enterprise production today
“You cannot use an LLM in an enterprise production in an autonomous fashion today. It's simply impossible because of the hallucination aspect and the lack of reliability.”
Scaling compute and GPUs alone won't yield autonomous enterprise AI
“I'm not an expert in the field, but I believe that this is gonna we need Another giant leap of innovation to get there. I'm not sure just throwing compute and increasing the number of GPUs without some fundamental discovery, it's gonna create this.”
Dines: Human-Supervised AI Assistants Will Precede Autonomous Agents
“So this is why, in my opinion, the first We will get to autonomous agents, but by going through, first, we will have to have this digital assistants that work under people's supervision.”
UiPath Developing Internal Multimodal Model Named Project Foundation
“We have a dual strategy. We have something called project foundation. So we are building internally You know, this dual model model that understands both text and images that we want to put at the foundation of many of our task-based automation.”
Dines: Partners open enterprise doors, but direct sales must close deals
“In our case, partners are not so important for us for reselling the technology. Partners are important to open the door and to implement our technology. But to win the deal, it's us. It's rare that the partner can win the deal for us.”
Most corporate partnerships are marketing hype without direct leadership ties
“Most partnership are just marketing announcement hype. But if there is no personal connection, it's not will on both parties to make it happen, it's not gonna work.”
Technical founders will never understand enterprise sales to the bare metal
“One of the issues I think a technical founder has is that they will never understand sales to the bare metal.”
Dines: Operational sales rigor does not matter for early-stage startups
“Initially, for a startup, I think it's operationally doesn't matter, in my opinion. All it matters, it's built an extreme customer-centric culture.”
UiPath CEO: Enterprise founders must love sales to hire great salespeople
“Especially in enterprise business, sales is as important as the technology, and it's not enough for them to understand the basic of sales. They need to love it. Because if you don't love it, you are not gonna hire really good people.”
Pitching public investors requires a single, perfectly consistent narrative
“When you talk to private investors, you have a lot of time to explain them the business. When you talk to public investors, you have to be a way more crisp, and you should tell the same story to everyone. You cannot deviate and give to some people more informa…”