Startup pioneers new use case for generative AI: Building business applications

Generative AI is among the biggest artificial intelligence (AI) trends in 2022, introducing users around the world to new capabilities such as automated text and image generation.

London-based startup Qatalog is taking a different approach to emerging technology. The company uses Generative AI to help build and enable business collaboration applications. Qatalog emerged from stealth in October 2020 with a goal of helping organizations unify enterprise tools and projects. 

There is no shortage of productivity and collaboration tools for organizations. That’s not necessarily a good thing. Qatalog partnered with Cornell University in 2021 on a research report on the state of productivity. They found that workers wasted an average of 59 minutes every day trying to find information that was stuck in one application or another.

On December 6, the company announced the availability of its Qatalog 2.0 service. The service takes a generative AI approach to helping organizations build customized application workspaces to improve productivity.

“We built up a platform to build out whatever systems organizations need for their company on their own terms,” Tariq Rauf, founder and CEO of Qatalog, told VentureBeat. “What we have essentially done with AI is made that super-fast and super-quick to get started.”

Qatalog 2.0: Getting specific

From the outset, Qatalog’s goal has been to help organizations access their data in one easy-to-use workspace.

Prior versions of Qatalog took a somewhat generic approach; they were designed to help any type of organization, said Rauf. But the 2.0 version gets a lot more specific: It enables users in specific industry verticals, with specific needs, to automatically generate workspaces for their own use cases.

Rauf said that Qatalog’s service can, for example, design a workspace for a law firm that has specific sections for clients and for legal cases.

“With Qatalog 2.0 we built this highly amorphous, really fluid system that lets you express the needs of the business as objects,” said Rauf. “And all of those objects have capabilities built around them.”

In the law firm example, a section for cases could have costs, knowledge management and discussions all centralized in one place. 

“Every single instance is unique for the specific customer,” said Rauf, with “its own data models, user interface and icons that are all generated on the fly and assembled for the customer based on the prompts that they give us.”

How Qatalog uses generative AI to enhance productivity

At the core of Qatalog’s ability to build workspaces for organizations is a combination of large language models (LLMs) and customized training for generative AI.

Rauf explained that the company has built a reinforcement learning loop to train its model. It also makes use of customized prompts for the GPT-3 LLM. 

“Typically people think of models as input/output, where you throw something into the model, and you get something out,” said Rauf. “We think of it as a multi-step process.”

Qatalog’s multi-step process refines the results it gets from the models with a sequential set of queries designed to get the right level of precision. Rauf noted that organizations tend to have slightly different needs and often use different terms for how they organize components. The generative AI piece in Qatalog 2.0 supports that uniqueness.

“Our AI understands how systems are put together, how databases work and how features talk to each other,” said Rauf. “The AI uses that knowledge and the knowledge of the customer’s business using GPT and merges the two to build a new solution that didn’t exist before.”