TigerGraph unveils GenAI assistant, introduces new CEO

Change in leadership

While the company's changes have been tumultuous at the top, the CEO changes likely represent evolution rather than troubled times given that Wu and Xu remain with TigerGraph, according to Matt Aslett, an analyst at ISG's Ventana Research.

"The previous occupants of the role remain with the company, which is a positive sign," he said. "That continuity combined with Azzawe's experience in chief executive and product management roles indicate that the company is transitioning to a new phase rather than changing direction."

Azzawe similarly said his appointment is part of a transition and not a sign of trouble.

Like many technology vendors, TigerGraph laid off staff following the COVID-19 pandemic. But while companies such as Salesforce and Microsoft each laid off more than 10,000 employees, that amounts to a small percentage of their staff.

TigerGraph downsized from 430 employees to 130, according to Azzawe.

The vendor had historically targeted an audience of academics and engineers. In addition, customers tended to be Fortune 500 companies. That audience was limited.

Now, with Azzawe as its CEO, TigerGraph is altering course somewhat and aims to make its platform easier to use so that a broader audience of users can work with it. That includes being more than just a graph database vendor -- which Azzawe acknowledged is a niche market -- and making generative AI-fueled analysis a focal point.

With me taking on the helm, we're solidifying our focus on being customer-driven and product-driven. We're a lot more efficient and a lot more focused on crossing the chasm from early adopters to ... being mainstream.

Not only will generative AI enable TigerGraph to be accessible to more users within organizations, but it also will let SMBs without teams of data engineers use the vendor's tools.

"The company contracted because of changes in industry dynamics [after the pandemic]," Azzawe said. "As part of that contraction, the board of directors started to revisit what our focus is and how we can transition from an academic engineering focus to being a more customer-driven, product-focused company."

Wu was therefore, in essence, a transitional CEO, he continued. Azzawe is now in the role to "regrow" TigerGraph.

Beyond shifting its focus to customer growth by changing its target audience, that means building the vendor's staff again with a goal of about 250 employees in two years.

"With me taking on the helm, we're solidifying our focus on being customer-driven and product-driven," Azzawe said. "We're a lot more efficient and a lot more focused on crossing the chasm from early adopters to ... being mainstream."

Assistance with AI

While Azzawe is a stabilizing presence, TigerGraph CoPilot and TigerGraph Cloud 4.0 are in beta testing and scheduled for general availability in July.

Generative AI has been the dominant trend in data management and analytics since late 2022 when OpenAI's launch of ChatGPT marked a significant improvement in large language model (LLM) capabilities.

The need to write code has long been a hindrance to widespread use of analytics within organizations, which has been stuck around 25% of employees for about two decades. In addition, having to write code to import data, integrate data, build data pipelines and develop data products has made data management a slow, cumbersome process.

Natural language processing, which converts natural language to code that computer systems can understand and then converts computer-generated responses back to natural language, has had the potential to enable new users and make experts more efficient. However, NLP has long been held back by limited vocabularies and a lack of intuition that forced users to write precise, business-specific prompts.

LLMs change that. LLMs, which are trained on vast amounts of public data, have both vocabularies as extensive as a dictionary and the interpretive skills that enable free-form natural language interactions.

As a result, many data management and analytics vendors have been integrating their tools with LLMs to develop AI assistants that enable customers to work with data without having to write code. For example, Microsoft and Spotfire were among the first to introduce AI assistants. More recently, Tableau and SAS have done so.

Now, TigerGraph has its own AI assistant under development in a move that has the potential to significantly benefit the vendor's customers by improving the productivity of existing users and lowering the barriers to entry for new users, according to Aslett.

"Many database vendors are adding copilot or digital assistant functionality to help users carry out complex tasks and lower the expertise barriers that prevent less technically skilled users from working with their products without upfront training," he said.

The capabilities being delivered with TigerGraph CoPilot should do just that, Aslett continued.

Specific features of TigerGraph CoPilot include the following:

• NLP that enables users to ask questions of their data and receive responses without requiring them to learn a new technology or coding language.
• Use of retrieval-augmented generation (RAG) to fuel the contextual relevance and accuracy of responses.
• Reliable AI that attempts to reduce the occurrence of AI hallucinations by enabling LLMs to access customers' graph databases through curated queries.
• Responsible AI with governance measures such as role-based access control and security rules already part of the TigerGraph platform to ensure safe and secure use.
• Scalability and performance by combining graph technology and RAG capabilities.

Beyond enabling NLP, one of the key benefits of TigerGraph CoPilot is that it provides documentation that shows users how it arrived at a response, according to Aslett. That allows users to check the accuracy of the response.

In addition, because graph technology is complex, providing a generative AI assistant should help TigerGraph move toward its goal of being accessible to a more widespread audience.

"Enterprises need to think differently about how they manage and analyze data using graph technology, which can present a barrier to adoption," Aslett said. "Natural language query and presentation capabilities can make it easier for users to start working with new technologies."

Platform update

While TigerGraph CoPilot marks the vendor's foray into generative AI, TigerGraph Cloud 4.0 is geared toward improving the performance and efficiency of the vendor's cloud-based platform.

After emerging from stealth in 2017 with a platform built for on-premises users, TigerGraph introduced a cloud-based version of its tools in 2019. Subsequent updates have addressed ease of use and added machine learning capabilities.

The latest version of TigerGraph's database-as-a-service platform features a new architecture that is fully cloud-native in a move aimed at enabling users to better customize their deployments as well as more efficiently use compute power as they scale data workloads.

Specific new features include the following:

• Separation of storage and compute so that users can scale storage and compute resources independently of one another, which results in more compute power to handle increasing data workloads and more storage room to house that rising amount of data.
• The introduction of separate compute workspaces that share a common, governed data store to provide users with environments for both real-time analytical processing and also real-time transactional processing workloads.
• Streamlined data ingestion so that customers can take in data from more sources with built-in connectors and a unified workflow. Included are connectors to local files, Azure Blob storage, Google Cloud Platform storage, Spark, Snowflake, Postgres, Kafka and BigQuery.

In addition, TigerGraph Cloud 4.0 comes with solution kits, which are prebuilt graph applications for specific applications such as fraud detection, supply chain management and customer 360.

Each includes graph schema, queries and a user interface that guides customers through processes such as importing proprietary data to the application, running the application and viewing the results. The intent of developing the solution kits is to reduce the time and cost it normally takes for customers to develop their own graph applications and enable users to get value from graph technology more quickly, according to TigerGraph.

Just as the significance of TigerGraph CoPilot lies in its potential to make TigerGraph easier to use, the solution kits have similar promise, according to Aslett. As a result, while features such as streamlined data ingestion and the separation of compute and storage are important, the solution kits could have the biggest impact.

"Adoption by new customers could also be accelerated by the new solution kits targeted at key use cases," Aslett said. "Other new features being delivered with TigerGraph Cloud 4 -- such as the separation of compute and storage, and isolated workspaces -- provide improved flexibility to support multiple applications."

TigerGraph's impetus for developing CoPilot and the platform update, including the solution kits, harks back to the vendor's shifting focus, according to Azzawe. Rather than continue developing capabilities aimed at a small audience of academics and engineers, the latest product developments are geared toward a broad audience.

"Everything we do should be about customers," Azzawe said. "TigerGraph is evolving from being engineering- and academic-driven to being a customer-obsessed company. Listening to our customer feedback was a key part of what we're doing now."

In addition, TigerGraph is keeping abreast of current data management and analytics trends and gearing product development toward making sure the vendor's customers have the same new capabilities as those being developed by others in the market, he maintained.

"We certainly are closely monitoring not only our competitors, but our partners as well, [such as] Snowflake and Databricks," Azzawe said. "We're seeing the trends and making sure we're staying ahead of them."

In the future

With CoPilot and the TigerGraph Cloud update scheduled for general availability in July, TigerGraph's long-term roadmap includes continuing to evolve its generative AI and platform capabilities once the features now in preview have been fine-tuned and released for full consumption, according to Azzawe.

With respect to CoPilot, the vendor plans to add capabilities that make the feature more proactive. The initial release is about being reactive, he said, such as responding to questions from users.

"The idea here is that CoPilot will move [from] reactive to proactive and become a companion for data scientists and developers so they can build richer solutions," Azzawe said.

Regarding TigerGraph Cloud, new low-code/no-code features are in the works, as are more solution kits, he continued.

Aslett, meanwhile, noted that as one of the more well-established graph database vendors, TigerGraph could benefit from efforts to increase awareness of both the company and graph technology itself.

"[TigerGraph] is now well established as one of the key providers of graph database and graph data science products and expertise, and is well positioned to benefit from steadily increasing awareness and adoption of graph technology," he said.

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.