Update

Launching Defog Agents: AI Assistants for complex data workflows

Handle complex, multi-step workflows in SQL, Python and R with human-in-the-loop agents.

Manas Sharma, Rishabh Srivastava

Defog Staff

Oct 5, 2023

Humans + AI = Data Analysis Superpowers

Today, we are excited to announce the launch of Defog Agents. AI Agents have been an extremely promising area of inquiry in the last quarter, and we believe that they will supercharge the productivity of knowledge workers in the coming months and years.

The computational tenacity of AI mixed with the taste and judgement of humans can lead to phenomenal outcomes, and we are excited to bring Defog Agents to the hands of more data analysts and data scientists.

Handle repetitive workflows with complete control

Defog Agents are AI assistants that can handle complex and repetitive workflows in SQL, Python, and R – while always asking data scientists clarifying questions. This allows data scientists to focus on high-value “planning” tasks, rather than spending time on routine coding and data wrangling.

Decompose complex questions into discrete, executable tasks

A key capability of Defog Agents is decomposing complex, multi-part questions into executable tasks that can be addressed by appropriate AI models. For example, if asked "What customer segments have declining usage of product X?" the agent can break this down into SQL queries to segment customers, analysis of usage trends with time series models, and application of clustering algorithms. The agent can then orchestrate and execute each sub-task, returning the full set of results.

Explore multiple hypotheses in parallel

Defog Agents automate many of the tedious trial-and-error tasks required for statistical analysis, like feature engineering. The agents can test different models, parameter settings, and data slices to find the best solutions. This eliminates the need for data scientists to manually run hundreds of variations themselves. The agents identify the most important factors and relationships in data for faster insights.

Explore results in a collaborative environment

Defog Agents generate reports in dashboards, slides, and spreadsheets, formatted for easy editing and collaboration. The agents produce initial drafts of standard data science deliverables, enabling faster delivery of finished work products. Agent outputs can be edited and regenerated, providing transparency and reproducibility for technical analyses.

AI that augments human judgement and taste

We believe Defog Agents will significantly improve efficiency and productivity for data scientists. The automation of tedious tasks coupled with human-in-the-loop oversight allows data teams to have greater impact through faster, higher-quality analysis. Defog Agents are designed to augment human intelligence, not replace it.

We have trialled agents with alpha customers in healthcare, manufacturing, and finance – and have received tremendously positive feedback. Defog Agents have delivered 4x productivity gains in some cases while requiring no training. We are excited to bring this new AI assistant to the broader data science community.

Contact Us

Reach out at [email protected], or just drop us a note on Twitter! We would love to get you started.

About the authors

Manas Sharma is an engineer at Defog. He enjoys woodworking, data visualization, making music, and hiking through the woods. Prior to joining Defog, Manas was a data journalist at Reuters, telling impactful stories through data visualizations.

Rishabh Srivastava is a co-founder of Defog. Before starting Defog, he founded Loki.ai – serving more than 5 billion API requests for Asian enterprises.


← More blogs