When Data Stops Being Code and Starts Being Conversation (Ep. 297)

2025-12-23 · Source: Data Science at Home Podcast · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Data Science & Analytics · Depth: Intermediate, extended

Summary

Mark Brocato, Head of Engineering at Tonic.ai and creator of Mockaroo, discusses the evolution of data generation from static scripting to AI-driven conversational agents. The episode explores why traditional static test data methods are becoming obsolete in the AI era, particularly for bootstrapping AI models in new companies lacking historical data. Brocato introduces Tonic Fabricate's AI agent, which replaces manual configuration with natural language interaction, allowing developers and data scientists to "negotiate" datasets. This shift aims to overcome the limitations of schema understanding and accelerate data generation, while also touching upon the security implications of agent-driven synthesis and the future of sandbox environments.

Key takeaway

For CTOs and VPs of Engineering evaluating data generation strategies, consider adopting AI-driven agents like Tonic Fabricate to overcome the limitations of traditional static test data. This approach can significantly reduce configuration overhead, accelerate data provisioning for AI model training and testing, and enhance data realism, especially when domain expertise is limited. Ensure your chosen solution offers robust privacy guarantees, such as using your own API keys with cloud services like Amazon Bedrock or Azure, to maintain compliance and data security.

Key insights

AI agents are transforming data generation from scripted pipelines to conversational negotiation, enhancing realism and accessibility.

Principles

• AI augments domain knowledge for data producers.
• Separation of data generation and consumption improves test reliability.
• AI-generated data can bootstrap models lacking real-world data.

Method

AI agents write code (JavaScript, SQL) to generate data, ensuring relational integrity by first generating parent entities and then looping to create dependent data, like purchases for users.

In practice

• Use AI agents to generate data for demo applications.
• Employ synthetic data to test models for contrived, rare scenarios.
• Upload existing datasets for AI to discover and replicate statistical properties.

Topics

• AI Agents
• Synthetic Data Generation
• Test Data Management
• Data Privacy
• Developer Workflow

Best for: CTO, VP of Engineering/Data, Director of AI/ML, Data Scientist, Machine Learning Engineer, Software Engineer

Related on AIssential

• Synthetic Data Generation for Smarter AI Workflows — Synthetic data generation expands limited real data into robust datasets for AI model training.
• Building With AI Agents Without Losing Your Architecture — Spec-driven development, using living contracts and deterministic tests, guides AI coding agents to maintain architectural integrity.
• The Top Six Test Data Management Tools for AI-first QA in 2026 — AI is transforming test data management into an automated, policy-driven system for secure, efficient data provisioning.
• The AI-First Data Engineer: 10–50x Productivity and What Changes Next — Agentic AI transforms data engineering by automating entire workflows, shifting human roles to agent operation and increasing data product creation.
• AI Builders: Building a text-to-SQL agent — Text-to-SQL agents democratize data access by translating natural language queries into structured database commands.
• The Sequence AI of the Week #887: Meta's Autodata: When Models Learn to Make Their Own Lessons — Autodata makes AI data generation an agentic, iterative process of creation, testing, and refinement.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Data Science at Home Podcast.