As a real-time intelligence vendor, Dataminr has made changes in its reporting strategies over the last couple of years, but it still focuses on using human experts, showing the importance of human input despite the rise of generative AI.
Dataminr got its start in 2009 and relied on human experts to monitor, review and categorize different events. While the startup was successful, it had limited capacity due to its heavy reliance on human experts. In 2018, after recognizing the need to scale with intelligent automation, the vendor began implementing AI in earnest.
“Back in 2018 … the writing was on the wall,” Joel Tetreault, chief AI officer at Dataminr, said on the latest episode of the Targeting AI podcast from AI Business. “If we really want to do this and scale to these greater heights, we’re going to have to employ some intelligent automation.”
This turn toward intelligent automation marked the beginning of Dataminr’s AI journey as it realized that it needed AI technology to process billions of pieces of information from a million public data sources across more than 150 languages. However, the company never completely removed the human experts from its reporting process.
Despite spending 2018 to 2020 using machine learning models such as a convolutional neural network (a deep learning model for processing grid-structured such as images and video), Dataminr still focused on implementing a human-involved feedback system.
“What we did was build in a process to collect data from the subject matter experts. We log all their keystrokes. So, every change or edit they would make before an alert … we would track that,” Tetrault said. “We would collect this information, use it to train the models. Then the models would get better.”
With the advent of generative AI and AI agents, Dataminr pivoted again while still keeping humans a key part of the process.
“It’s 2026, and these models can alert on 99% of what we alert on, and our human experts are actually used for very nuanced and tricky types of events that are unfolding,” Tetreault said.
