Rasgo launches Rasgo AI, a generative AI agent for enterprise data warehouse analytics 

AI-driven analytics platform Rasgo has announced the launch of Rasgo AI, a self-service analytics solution that integrates a GPT (generative pre-trained transformer) into enterprise data warehouse (EDW) environments. The company said that with Rasgo AI, organizations can use the power of AI/GPT to accelerate insights and optimize recommended actions securely and efficiently.

Unlike other GPT integrations that provide only natural language chat interfaces, Rasgo said its AI stands out by employing GPT for “intelligent reasoning,” which enables it to think and act like a knowledgeable business analyst for data warehouses. 

Knowledge workers often get bogged down by time-consuming, low-value tasks that hinder effective decision-making. By offloading these tasks to AI, Rasgo aims to enable these workers to focus on strategic decision-making, leading to significant gains in enterprise value.

Answering questions — and asking them

Rasgo asserts that GPT-4 enables the model to adeptly perform intricate reasoning tasks with dynamic objectives. The autonomous agent becomes capable of generating a semantic embedding of the EDW metadata, thereby educating GPT-4 about the data while retaining control of the data within the secure environment of the enterprise.

“One of our most exciting technical findings was that when provided with the right guidance, GPT-4 is not only good at answering data analysis questions but also good at asking them. Rasgo provides a metadata repository about the data to teach the AI how to make specific decisions when analyzing data so that it can iteratively improve and learn from human calibration,” Jared Parker, cofounder and CEO of Rasgo, told TechForgePulse. “By combining the chat interface with our solution for intelligent reasoning, we aim to improve … operational efficiencies of [customers’] key stakeholders while also trusting that AI is constantly analyzing data to derive key insights.”

According to the company, one of Rasgo AI’s key differentiators is AI Guardrails, which map data structures into familiar business terms, enhancing the efficiency and accuracy of data analysis while ensuring data security. The platform also analyzes enterprise data continuously to provide trusted insights, enabling business users to make data-driven decisions without needing advanced SQL skills.

Leveraging GPT-4 for intelligent business reasoning 

Parker stated that for intelligent reasoning, the platform trains GPT to replicate a data analyst’s role. This equips enterprise data teams to accelerate analysis, as opposed to building queries and dashboards from the ground up.

“We acknowledged the potential time constraints faced by humans in formulating all necessary inquiries. Our intelligent reasoning establishes an ‘always-on’ virtual team of knowledge workers, consistently identifying business prospects and vulnerabilities,” said Parker. “A simple prompt like ‘analyze trends in year-over-year sales growth by sales rep’ can yield a comprehensive presentation of pivotal insights and actionable steps.”

For human-AI collaboration, Rasgo said that its platform aids data teams by autonomously assessing tables in the data warehouse and discerning which tables are primed for intelligent reasoning and which need further refinement. 

This approach, according to the company, enables human stakeholders to channel their energies toward transforming and documenting tables that require additional manual attention to ensure governance and trust in an AI-based workflow.

The company also highlighted that its generative AI model can mechanize numerous routine, low-value tasks inherent in the data analysis lifecycle. This automation aims to guide users through the process of data discovery and analysis, all the while maintaining the oversight of a data analyst. The platform’s ultimate goal is to optimize accuracy and instill trust in organizational processes.

“The conventional data analysis process is broken. Answering a single data-driven question can take an exorbitant amount of time, involving the identification of relevant tables, writing and debugging SQL queries, creating dashboards in BI tools, and translating results into comprehensible business recommendations,” Parker explained. “Our AI engine proactively searches metadata and query history to suggest the gold-standard table; writes, tests and executes the necessary SQL query; generates the appropriate visualization; and distills the results into actionable business recommendations. Throughout this process, we ensure the data analyst remains involved, enabling human decision-making at critical junctures to optimize for accuracy and trust.”

Parker said that for Rasgo’s AI to navigate a database, the generative AI model crafts embeddings for all data warehouse metadata and user-provided instructional data. This ensures swift retrieval within what the company calls its ReAct (reason + act) AI workflow. Additionally, it autonomously maintains and refreshes these embeddings whenever new tables emerge, schemas evolve, or fresh user instructions are incorporated.

Ensuring responsible AI development

Parker asserted that responsible functioning of generative AI and achieving desired outcomes from the technology hinge on collaborative efforts between humans and AI. This entails setting explicit rules, instructions and guardrails to ensure trust and safety, particularly in the context of enterprise data.

He explained that to counter the risks of hallucination and data disparities, the company has formulated an “AI Manager” capability. This suite of tools empowers users to establish definitive guardrails and constraints on the large language model (LLM), ensuring its selection of the gold-standard table, column and metric when addressing user prompts.

The platform’s AI automates the documentation of table metadata sourced from the data warehouse environment. Simultaneously, it assigns an “AI Readiness” score to each table. This score aids data teams in distinguishing datasets primed for secure AI applications from those requiring further human intervention.

The company has built its solution around Microsoft as Rasgo’s AI API provider, integrating directly with Microsoft’s security framework.

Democratizing trusted intelligence

“LLMs like GPT are widely used for text-to-SQL translation. However, incorrect SQL can lead to flawed decisions based on inaccurate data. Our platform democratizes trusted intelligence by teaching GPT about a user’s schemas and teaching it to respect user-provided instructional data so that it can be instantly retrieved to produce accurate SQL and trusted insights,” Parker told TechForgePulse. ”In terms of data privacy and security, we have implemented “push down compute” capabilities. This means that the SQL generated by the LLM is sent directly to the organization’s cloud data warehouse environment, ensuring no raw data leaves their warehouse.”

The company recently announced its collaboration with Snowflake’s Partner Network, aiming to enhance the benefits of the Snowflake Data Cloud for mutual customers. Through this partnership, Rasgo says it is able to harness GPT for intelligent reasoning, streamlining self-service analytics. 

“Going forward, we plan to continue the momentum with this partnership and others similar, by enhancing the accessibilities it provides to customers and overall making sure the product itself can meet the needs of all organizations at all stages of their data analytics and AI journeys,” said Parker.