WisdomAI Raises $23M to Tackle AI Hallucinations

WisdomAI: New Data Analytics Startup Emerges

Soham Mazumdar, formerly a co-founder at Rubrik and departing in 2023, is now leading a new data-focused venture known as WisdomAI. This company specializes in AI data analytics, designed to provide actionable business insights from a variety of data types.

Analyzing Diverse Data Sources

WisdomAI’s capabilities extend to processing structured, unstructured, and even imperfect data – often referred to as “dirty” data – meaning information that hasn’t been meticulously cleaned of errors or inconsistencies.

The ability to work with data in its native state represents a significant advancement in enterprise business intelligence software. This potential is reflected in the substantial $23 million seed round led by Coatue, with participation from Madrona, GTM Capital, Anthology Fund, and other investors.

Empowering Business Managers

Unlike traditional data analytics requiring dedicated teams to generate reports, WisdomAI allows business managers to directly query the data and explore details.

For instance, a Chief Revenue Officer (CRO) could ask, “How can I successfully close this quarter?” WisdomAI would then provide a prioritized list of pending deals, detailing the specific obstacles hindering each one, such as outstanding customer questions.

Streamlined Data Access

Mazumdar explains that this platform enables a CRO to access granular details with just a few keystrokes. This contrasts sharply with conventional methods that involve multiple personnel, including data analysts, and considerable time investment.

This is just one illustration of the types of inquiries WisdomAI is engineered to address.

Real-World Applications

An oil and gas company represents another early adopter, utilizing WisdomAI to empower its field workforce with data-driven answers regarding production. The system taps into data from diverse sources, including stored documents and telemetry readings.

While numerous business analytics tools, including emerging startups, are integrating AI-powered natural language processing, WisdomAI distinguishes itself through its founding team’s experience – all previously collaborated with Mazumdar at Rubrik.

Accuracy and Reliability

The platform’s core strength lies in its accuracy, even when dealing with messy data. It can extract insights from both structured databases and unstructured files.

Crucially, WisdomAI is designed to avoid generating inaccurate or fabricated responses – often referred to as “hallucinations.”

Mitigating AI Hallucinations

Many enterprises are focusing on improving AI application accuracy by refining the data used for model training, increasing model size, optimizing prompt engineering, and implementing techniques like retrieval-augmented generation (RAG). However, the risk of fabricated answers remains.

WisdomAI employs GenAI during the query formulation stage, but not in the answer generation process. “GenAI can be prone to hallucination. We leverage GenAI to create concise programs that can effectively query various systems,” Mazumdar clarifies.

Therefore, if WisdomAI’s model were to hallucinate, it would only result in a faulty query that fails to retrieve data, rather than fabricating the answer itself.

Customer Base and Integrations

WisdomAI’s early customer roster includes prominent companies like ConocoPhillips, Cisco, and Descope. The platform integrates with leading cloud data storage services, including Snowflake, Google’s BigQuery, Amazon’s Redshift, Databricks, and Postgres.

The system can be adapted to any data storage system by analyzing its query language through query logs and other data sources, according to Mazumdar.