Before generative AI became a prominent industry trend, predictive AI was already paving the way by making forecasts about future events based on existing data. Imagine harnessing the strengths of both technologies—this is the vision of Pecan AI.
Founded eight years ago, Pecan AI has established a predictive analytics platform for enterprises and raised $116 million in funding, including a notable $66 million round in February 2022.
Introducing Predictive GenAI
Today, Pecan AI is launching Predictive GenAI, a new tool that merges the advanced capabilities of generative AI with predictive machine learning. “While we were focused on classic machine learning solutions, the generative AI revolution unfolded across the industry,” noted Zohar Bronfman, CEO and co-founder of Pecan AI. “However, generative AI struggles with making accurate predictions.”
While generative AI excels in various tasks, like content summarization and chatbot development, it isn't designed for prediction. Predictive machine learning, on the other hand, can be challenging for users. Pecan AI’s Predictive GenAI effectively combines these technologies, allowing data scientists to easily create predictive models.
Making Predictive AI Accessible to Businesses
Pecan AI aims to simplify the adoption of machine learning and AI for companies. Traditionally, data scientists have been the main users of predictive analytics tools.
Bronfman emphasized that Pecan AI is designed to democratize AI capabilities, enabling business professionals to leverage predictive insights without needing deep technical expertise.
Key Features of Predictive GenAI
1. Predictive Chat: This feature allows users to interact with a chatbot-like interface, enabling them to pose natural language queries tailored to specific business challenges. It intuitively guides users toward the most relevant predictive frameworks.
2. Predictive Notebook: Utilizing generative AI, this proprietary SQL-based notebook serves as the foundation for building predictive models. It automates the transformation of raw company data into AI-ready datasets through generated cells that handle data querying, structuring, and joining. Users can choose to run these cells automatically or adjust them using SQL for deeper involvement.
Challenges of Predictive AI in Generative Models
Users may find that while generative AI is efficient at various tasks, it falls short in making predictions. According to Bronfman, this is because the data sets generative AI relies on during training do not align with the structured format necessary for predictive modeling.
A predictive dataset must have well-defined entities in rows and clear features in columns, along with a target label. However, acquiring data in this format often necessitates extensive data engineering—a task typically handled by experienced data scientists.
Bronfman also highlighted the limitations of using vector databases for predictive modeling. Although vector embeddings can support basic predictive functions, they tend to operate with a restricted set of features and require either simple models or significant labor in feature engineering from data scientists to prepare data adequately.
Innovations in Data Preparation
While the conversational interface of Predictive GenAI may be its most visible feature, Pecan AI is also focused on innovating automated data preparation and feature engineering.
Among the advancements is an automation solution aimed at mitigating data leakage—a common issue that can skew model accuracy. Data leakage refers to using training data that should remain hidden during predictions, which can be challenging to identify without expertise.
“It’s not easy to detect leakage, especially for non-experts,” Bronfman explained. “We offer automated methods to identify and address it.”
By enhancing accessibility and accuracy, Pecan AI is set to transform how businesses harness predictive analytics, paving the way for more informed decision-making in the ever-evolving landscape of AI.
Ranking
