DareData Engineering Partner: “There’s a bubble in AI. Services are being delivered without proper expertise”

In a summer that was relatively cool for business in Portugal, NOS made a move to acquire expertise in generative Artificial Intelligence (AI). At the end of July, the telecommunications operator acquired a 20% stake in the Portuguese tech company DareData Engineering, specializing in data infrastructure development and machine learning projects.

‍

Jornal Económico (JE) got to know this company, founded in 2019 with the aim of intersecting data science and data engineering. The startup, composed of scientists and engineers, works to develop technology that has a positive business impact, rooted in reality. Why? Before the pandemic, the founders encountered projects that were “practically disconnected from the company’s database structures and applications.”

‍

“At the time, there were many companies offering these kinds of services, but they were always things with little impact on the business. It was like, ‘I’ll create an algorithm for a company.’ But then, it’s almost like a picture that gets hung on the wall, and people look at it and say, ‘How nice,’ but it never created value,” recounts partner Ivo Bernardo, who welcomed us to their headquarters in the Picoas area of Lisbon.

‍

Although he is one of the three partners, Ivo Bernardo joined DareData Engineering three or four months after the company was founded. It was a kind of spin-off, led by engineers Nuno Brás and Sam Hopkins, who were part of the Lisbon Data Science Academy—a school for retraining and training data science professionals—and were linked to James, a startup later sold to Google.

‍

Both had their own independent projects until they decided to combine their computer engineering and statistical expertise to improve their AI knowledge and create this new company, which united the “two schools”—still in the pre-ChatGPT era. NOS was one of their first clients.

‍

“If companies keep creating chatbots here and there, in five years, the systems won’t talk to each other, and they won’t be able to connect.”

‍

The first conversation, in a partnership logic, took place around six or seven months ago, and only afterward did the investment opportunity arise, although nothing in the company’s strategy pointed toward M&A or unicorn growth. From the partners’ perspective, that kind of planning is distracting, and “everyone ends up losing.” They even rejected additional service contracts that would have forced them to “turn on the hiring faucet,” as they believed it wouldn’t be sustainable.

‍

“One of the initial problems we detected was that most AI solutions being developed solve very specific problems: a chatbot for customer support, a chatbot to read instructions from a massive PDF, a chatbot for marketing… We started to fall into a bit of what happened with database systems 15-20 years ago, where it’s difficult to transfer data back and forth. It’s a mess, and all companies are feeling this difficulty,” explains the DareData partner.

‍

“What we think will happen with AI systems, if they’re built with this chatbot-here-chatbot-there approach, is that in five years, we’ll have exactly the same problem: the systems won’t talk to each other, and they won’t be able to connect. For companies moving toward a unified data and customer vision, it’s a complete setback,” he warns.

‍

In the technology expert’s opinion, the solution lies in developing AI systems that integrate with companies’ database systems and legacy systems. “It’s tough work,” he acknowledges. Even tougher when AI is (excessively) hyped.

‍

Ivo Bernardo has no doubt that the market has somewhat inflated expectations about the potential of this technology. “I think there’s an AI bubble in the sense that many people are delivering AI services without knowing how to do it, which is typically a good indicator of a bubble. People are extremely interested, maybe even with unrealistic expectations of what can be done. To say there’s no bubble would be like burying one’s head in the sand,” he states.

‍

He even admits that the peak of this bubble may have passed. “However, even with a bubble, I think when interest in AI starts to wane, that’s when the truly important case studies that deliver value will emerge. That’s where we can be positioned, and that’s where companies with deep technology knowledge will succeed,” he says.

‍

“We’ve always tried to differentiate ourselves by being demanding in our hiring. Typically, a data science professional can leverage their knowledge by also understanding data engineering. From the moment they know both, their personal value rises by 50-60%. Because of this type of talent, we deliver faster compared to companies whose strategy ends up being hiring a lot of junior staff and training them to create high turnover,” assures Ivo Bernardo.

‍

Regarding NOS, DareData’s next step is to create an AI system that acts as a service router for the various AI systems within the company. The roadmap isn’t finalized yet, but the first versions of this product should be launched in about two years. The investment in development is planned for three years.

‍

Asked about the goal of generating five million euros in revenue by 2024, after 3.1 million euros last year, Ivo Bernardo told JE that they are “recalibrating the business volume a little” and admit “without any problem that they may fall slightly short of that figure.” “We only had the business from a service perspective—revenue, revenue, revenue—and we want a more solution-oriented component,” the partner argues. Their goal remains to double the team to 100 people over the next five years.