DBT Labs unveils AI assistant, more tools to transform data

New capabilities

Interest in enterprise AI, fueled by the rise of generative AI, has exploded in the 18 months since OpenAI’s launch of ChatGPT marked a significant advance in the capabilities of large language models (LLMs).

Enterprises and vendors alike have long searched for ways to enable more employees to use data to inform their work as well as make experienced data users more efficient.

Natural language processing (NLP) and low-code/no-code approaches held promise but never fully delivered.

Generative AI may finally be the technology that lets more users work with data and helps experts be more productive.

LLMs have extensive vocabularies, which enables true NLP and greatly reduces the need to write code to interact with software systems. In addition, LLMs can infer intent, further allowing users to engage with technology using freeform language.

As a result, when combined with an organization’s proprietary data, generative AI enables virtually any user to generate insights that can inform business-specific decisions.

Given that potential, many data management and analytics vendors have made generative AI a prominent part of their product development plans. In particular, they have unveiled plans to develop AI assistants that enable natural language interactions and simplify tasks.

Now, DBT Labs is joining the fray.

The vendor’s DBT Assist is an AI-powered assistant that focuses on improving the productivity of data experts rather than providing conversational capabilities. Currently in beta testing, the tool automatically provides documentation and generates tests to eliminate some of the manual labor required during the development process and help improve efficiency as developers build applications.

As helpful as the tool may be, equally important is simply providing an AI assistant to remain competitive with other vendors, according to Kevin Petrie, an analyst at BARC U.S. By unveiling generative AI capabilities, whether GA like a small group of vendors have or in development like most others, Alteryx is showing customers that it is up on the latest technology.

“The GenAI copilot, while not a new concept, is critical to remaining competitive in this space,” Petrie said. “Alteryx is in beta, which reflects caution. Many vendors are taking time to incorporate user feedback on these features before going GA.”

In addition to DBT Assist — and perhaps even more significant given that AI assistants are becoming common — new governance, metadata management and low-code development experience are significant additions to the vendor’s platform, according to Henschen.

New governance features added to the DBT Semantic Layer include granular access controls and permissions, integrations with Tableau and Google Sheets and improvements to MetricFlow that help users develop and consume complex metrics with greater speed and accuracy. All but the access controls and permissions, which will sone be in preview, are now GA.

New metadata management features in DBT Explorer add improved data lineage capabilities, model performance monitoring to better match the amount of work needed to build a model with its business impact, and data quality information embedded in analytics tools to ensure that data can be trusted. All will be in beta testing soon.

The low-code development experience, now in beta testing, includes a drag-and-drop editor that generates SQL code on its own, enabling users who don’t know how to write code to collaborate during the data engineering process.

“All vendors seem to be adding GenAI assistants, so DBT Assist seems less significant to me that the combination of governance, metadata management, and democratization upgrades,” Henschen said.

Petrie likewise noted that the new governance features in the DBT Semantic Layer are important and show DBT Labs’ growth beyond its roots.

With many enterprises struggling to manage data across various departments, tools that help keep data consistent and organized are imperative.

“The semantic layer enhancements also improve DBT’s competitive position,” he said. “So many companies struggle to keep distributed consumers of distributed datasets on the same page. The semantic layer helps govern this data consumption with consistent metrics, access controls and usage patterns.”

Additional new capabilities include the following:

• A view during data pipeline development that lets teams view and verify changes to the codebase to make sure the changes meet quality standards before they are put into production (beta testing coming soon).
• Unit testing, a feature now generally available that improves the coverage of tests during development without increasing costs by validating model logic earlier in the development process than was previously available.
• DBT Cloud CLI, a tool now GA that provides developers the option of contributing to projects in DBT Cloud through their integrated development environment of choice rather than forcing them to develop withing DBT’s environment.
• The availability of DBT Cloud on Microsoft Azure (beta coming soon) plus integrations with Azure Synapse (preview) and Microsoft Fabric (GA).
• Automatic exposures, a feature scheduled for beta testing soon that enables users to automate and trace end-to-end data lineage between Tableau dashboards and DBT models to help users trust the data populating their data visualizations.
• The GA of DBT Mesh, a feature that enables organizations to manage interconnected projects across domains rather than require them to centrally manage all projects.

While each of the new features address different user needs, together they are aimed at enabling DBT Labs customers to efficiently develop data pipelines, Petrie noted.

That is significant given that data pipelines are what feed analytics and AI applications.

“BARC research shows that companies’ number one success metric for AI is the timeliness and accuracy of insights, which depend directly on the effectiveness of data pipelines,” Petrie said. “DBT’s enhancements help companies meet increasingly rigorous requirements for analytics initiatives, especially those that involve AI.”

Feedback from companies, meanwhile, served as part of the impetus for the latest DBT Cloud update, according to Luis Maldonado, vice president of product at DBT Labs.

In addition, the vendor’s own observations of trends and ideas about how users can best be helped as they transform data for analysis were part of the development process, he continued.

“We certainly listen to direct feedback but have also observed that our customers are seeing an increasing importance in aligning data across the entire organization,” Maldonado said, noting the updates to the DBT Semantic Layer. “The next step is to make it even easier for stakeholders of various technical aptitudes to connect and collaborate around their data.”

The road ahead

As DBT Labs plans future updates, generative AI will be a primary focus, according to Maldonado.

In recent research, conducted by the vendor, one-third of respondents said they are currently using generative AI. However, more than half expect to be doing so soon. To meet the needs of those customers that want to add or expand their use of generative AI, DBT Labs’ roadmap includes data transformation capabilities to feed tools such as conversational assistants trained on customers’ proprietary data.

“Many customers want … the ability to ask questions of their internal data using LLMs but recognize the operational risk of powering responses with inaccurate data,” Madonado said. “As we continue to embrace GenAI, we’re dedicated to ensuring the data that feeds into these models is accurate and trustworthy.”

Henschen, meanwhile, noted that DBT Assist has limited capabilities compared to some of the AI assistants developed by other vendors. Unlike other copilots that have conversational capabilities, it focuses largely on automated documentation.

As a result, DBT Labs would be wise to add more functionality to DBT Assist or develop another generative AI tool as a complement.

“I think [DBT Labs} is headed in the right direction on many fronts, but the functionality of DBT Assist is rather limited,” Henschen said. “I could easily see more assistants or a more broadly capable single assistant that would help with the company’s push to support a broader base of users.”

Eric Avidon is a senior news writer for TechTarget Editorial and a journalist with more than 25 years of experience. He covers analytics and data management.