Ozgur Akaoglu
Ozgur Akaoglu

Most companies do not lose customers all at once. They lose them in signals. A missed purchase. A change in frequency. A moment when a customer could have been brought back, but the team did not see it quickly enough. For Ozgur Akaoglu, co-founder of Hilbert, that space between what a business knows and what it actually does became the problem worth building a company around.

"I kept seeing the same pattern," Akaoglu says. "Companies had data. They had smart people. They had dashboards. But the action still came too late, or it did not come at all."

Hilbert builds AI-powered growth intelligence for retailers and e-commerce brands. Its platform predicts customer behavior, including which customers may stop buying, which ones may be ready to return, and what they are likely to spend. Then it deploys AI agents to act on those signals through the right intervention, at the right moment, through the right channel.

Akaoglu describes the system in practical terms. Hilbert is not trying to give brands another report to interpret. It is designed to operate more like an AI data professional and execution engine that works around the clock on customer data.

"The question was never just whether AI could identify a signal," he says. "The question was whether it could do something useful with that signal before the moment passed."

That way of thinking did not come from watching the AI market from a distance. Akaoglu studied computer engineering, then joined Getir as one of its early engineers, when the company was still small enough that people had to work across whatever problem was most urgent. He moved through software engineering, data engineering, and data science, building systems for personalized customer actions, order assignment, and acceptance algorithms.

Those were not theoretical challenges. They were predictive systems tied to real operations, real users, and real consequences.

"At that stage, you learn very quickly that models are only valuable if they can survive the real world," Akaoglu says. "You cannot hide behind a clean dataset when the business needs something to work at scale."

Getir gave him exposure to operational machine learning problems that had to perform under pressure. Later, he spent two years with an SF-based startup working on LLM-based product development. That experience gave him a different view of where artificial intelligence was heading and what it might be able to do beyond classification, prediction, and automation.

By the time he returned, he had seen two different sides of the same shift. One side was the demanding, messy world of operational ML at scale. The other was the emerging power of large language models applied to business problems.

"That combination felt rare," he says. "I had seen what it takes to build systems that actually run, and I had also seen how much more flexible AI was becoming. It felt like the right moment to bring those worlds together."

Hilbert began with that instinct. Akaoglu and his co-founders were already working through these problems for individual companies. Again and again, they saw businesses with strong data and real growth teams still struggling to connect signals to action. The question became harder to ignore: why was there no general system for this?

"We were in the trenches solving versions of the same problem," Akaoglu says. "At some point, you stop seeing it as a one-off client issue and start seeing the shape of the category."

The category, as Akaoglu sees it, is not another analytics layer. It is not a better dashboard. Hilbert is built around the idea that growth should not depend on disconnected teams each optimizing its own slice and hoping the combined result works. A brand's CRM, paid media, product, and retention efforts may all be active, but that does not mean they are learning from one another.

"Every company has growth activity," Akaoglu says. "Very few have a growth system."

A customer who is about to churn may still look valuable in one tool, disengaged in another, and invisible in a third. Hilbert's goal is to model customer behavior at the individual level, predict what is likely to happen next, and help the business act before the opportunity closes.

The company is now working with clients across the US and Europe, with a team spread across Istanbul, San Francisco, Barcelona, and London. Hilbert has also raised a $28M Series A led by a16z, with participation from Asylum Ventures, ScaleX Ventures, SV Angel, Valkyrie VC, TIBAS Ventures, and Gunderson Ventures. Its clients include Walmart, which adds commercial weight to a company built around production AI rather than AI theater.

Agentic AI is one of the most talked-about ideas in enterprise technology, but Akaoglu believes the market is still separating companies that can demo agents from those that can operate them inside live business environments.

"People underestimate how hard production is," he says. "It is not enough for an agent to complete a nice workflow in a controlled setting. Enterprise data is messy. Requirements change. Compliance matters. Clients do not care how elegant your architecture is if the system does not work."

That lesson has followed him from Getir to SF to Hilbert. The technical details have changed, but the principle has not. A system that creates business value has to function in the conditions businesses actually live with.

For Akaoglu, the bigger ambition is to build a platform that helps enterprises move from intuition and averages to provable, compounding growth. Most companies know more about their customers than ever. Hilbert exists because knowing is not the same as doing.

"There is a gap between insight and action," he says. "That gap is where a lot of revenue gets lost. We built Hilbert to live there."

For more information on Ozgur Akaoglu, visit the Hilbert website.