Getty Images/iStockphoto

New Acceldata co-pilot targets efficiency, broader use

The data observability specialist's new generative AI-powered capabilities are designed to help experts be more efficient as well as enable non-technical users to work with data.

Acceldata on Tuesday launched the public preview of a generative AI co-pilot aimed at making the vendor's tools easier to use and enabling more people within organizations to work with data.

The co-pilot is a direct result of Acceldata's September 2023 acquisition of Bewgle and its AI-powered unstructured data observability capabilities and is scheduled for general availability during the second quarter of 2024.

Based in Campbell, Calif., Acceldata is a data observability vendor whose platform enables customers to monitor their data for quality as it moves from ingestion through the data pipeline.

By monitoring data as they prepare it for analysis, users can immediately address any problems so that data can be trusted when put into action training models and informing decisions. Previously, users had to wait to discover and fix problematic data until the modeling and analytics processes.

In addition to Acceldata, Monte Carlo and Unravel are among vendors also specializing in data observability.

Beyond its acquisition of Bewgle, Acceldata's recent activity includes raising $50 million in venture capital funding in February 2023, adding new automated remediation and low-code/no-code capabilities in March 2023, and raising another $10 million in October 2023 to bring its total funding to $105.6 million.

A list of the five pillars of data observability.
Five pillars of data observability.

New capabilities

Increased efficiency is one of the primary reasons for using generative AI co-pilots, according to Matt Aslett, an analyst at ISG's Ventana Research.

"The most popular motivation for adopting AI is efficiency, followed by innovation and business growth," he said, citing ISG's 2024 Buyer Behavior Study on Enterprise Adoption of AI. "AI-powered assistants and co-pilots are a significant area of investment for software providers and enterprises as they seek to improve efficiency."

Co-pilots are essentially generative AI-powered assistants that allow users to both converse with their software using natural language and automate tasks that take up significant time when performed manually.

Generative AI large language models (LLMs) such as ChatGPT and Google Bard can be integrated with data management and analytics platforms to enable true natural language processing (NLP). Previously, vendors developed their own NLP tools.

Those vendor-developed NLP tools, however, had limited vocabularies and required data literacy training to use.

Public LLMs have vocabularies as extensive as the Merriam-Webster Dictionary. In addition, generative AI can infer meaning. Combined, those vocabularies and the LLMs' ability to infer meaning enable true conversational interactions that eliminate much of the coding previously needed to interact with a software system.

In the case of data management and analytics vendors, they enable customers to monitor, prepare, query and analyze data using natural language rather than code.

In addition, LLMs can generate code on their own, making them useful tools for data engineers and other data experts building data pipelines as well as analysts working with the data in those pipelines.

Acceldata's new co-pilot enables the following:

  • Anomaly detection that improves the trustworthiness of data by alerting users to problems and changes related to data freshness, data profiling and data quality.
  • Cost control by learning the consumption patterns of an organization, including seasonal increases and decreases in use, to help enterprises predict and manage spending.
  • Application automation that reduces manual workloads as well as more comprehensively oversees the data lifecycle to detect problems humans might not otherwise uncover.
  • Automatic summary generation to describe data assets and organizational policies related to data management and analytics.

"Data operations provides multiple opportunities for efficiency gains through the use of AI," Aslett said.

In particular, he noted that anomaly detection, data asset and data pipeline summarization and documentation, and automation to improve performance and cost efficiency are common applications of AI for data management.

"These are naturally among the use cases being targeted by Acceldata with its AI co-pilot," Aslett said.

Beyond efficiency, a key benefit of the co-pilot will be enabling more users within customer organizations to be able to use Acceldata, according to Rohit Choudhary, the vendor's founder and CEO.

Given that LLMs reduce the need to code, they make data management and analytics tools available to more people within a given enterprise. Due in large part to the complexity of such tools, analytics use within organizations has been stagnant around a quarter of employees within organizations for about two decades.

True NLP has the potential to change that and make data almost universally accessible.

"One of the primary goals of the co-pilot is to bridge the business/technical divide and make business users a full participant in the entire process of making a data product," Choudhary said.

He noted that Acceldata's co-pilot aims to let users to interact with data throughout the data pipeline as models, dashboards, reports and other data products are built. The result is that problems can be fixed as soon as they are discovered in the pipeline rather than during the analysis and decision-making stages.

AI-powered assistants and co-pilots are a significant area of investment for software providers and enterprises as they seek to improve efficiency.
Matt AslettAnalyst, ISG's Ventana Research

"Perspectives and feedback can be applied early, thereby saving enterprises a lot of re-work," Choudhary said. "This is in stark contrast with approaches where the business user is restricted to operating in the consumption zone, increasing the costs of building the data product since issues are caught very late in the consumption stage."

Just as Acceldata is not the only data observability specialist, it is not the only data management vendor that has developed a generative AI co-pilot over the past year, Aslett noted.

Because of their potential to make data more accessible to business users while also helping data experts become more efficient, co-pilots are becoming common.

Microsoft is perhaps the most prominent software vendor to introduce AI assistants and call them co-pilots.

The tech giant first unveiled its co-pilots in 2021, revealing an assistant for joint users of Microsoft and GitHub. Since then, Microsoft has introduced numerous other co-pilots, including ones for data management and analytics tools such as Power BI and Fabric.

Among other data management and analytics vendors, Spotfire also calls its generative AI assistant a co-pilot while AWS, Google, MicroStrategy and Domo are among the many vendors that have unveiled AI assistants under a different name.

"Co-pilots are being introduced by many vendors in many software segments and AI-powered assistants are likely to be included in all enterprise software within the next few years," Aslett said.

The assistants, however, are difficult to compare given that vendors have different specialties and develop their generative AI tools to address those specialties, Aslett continued.

"While the various co-pilot offerings have similar names and overall value, they are not necessarily directly comparable as they are tailored to the specific roles and responsibilities of individual users and the functional capabilities of the associated software," he said.

The future

With its co-pilot now in public preview and general availability expected by the end of spring, Acceldata has plans to continue expanding its AI capabilities, according to Choudhary.

Data observability's aim is to improve the quality and reliability of data. With data volume rising exponentially and data complexity similarly increasing, AI can do a better job of overseeing an organization's data than even a team of humans can.

In addition, Choudhary noted that the same data pipelines that currently feed data products such as traditional AI models, dashboards and reports will be used to train generative AI models in the future.

To reduce the occurrences of AI hallucinations and make sure those generative AI models are as accurate as possible, data quality is imperative.

"Enterprises have been experimenting with non-generative AI models for the last few years, and more models will hit production in the coming years," Choudhary said. "A lot of these models will be generative AI in nature, which is where our next focus will be."

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.

Dig Deeper on Data management strategies