From BI to AI:
Making Data-Driven Decisions with Agentic AI
Instructor
Kerry Back
kerry.e.back@rice.edu
J. Howard Creekmore Professor of Finance and Professor of Economics
AI agents that understand plain English, query your data, and deliver answers instantly are transforming how organizations make decisions. This eight-session course gives you hands-on experience building and using these tools — no coding background required.
You’ll use AI to wrangle data, generate charts and reports, solve optimization problems, and automate workflows. You’ll learn how to build custom AI agents that connect to databases and documents, how to secure them for enterprise use, and how to develop an implementation plan for your organization.
The AI Lab provides each participant with cloud-based access to Claude Code so you can work with AI tools without installing software or subscribing to an AI provider.
Session 1: The AI Landscape. Chatbots and agents, the major models, large language models from 30,000 feet, Claude Desktop: Chat/Cowork/Code, API calls, Claude skills and connectors, prompting techniques.
Session 2: Working with AI. Terminal commands and Python scripts, querying databases, common data operations, generating charts, Excel workbooks, PowerPoint decks, and text output.
Session 3: Decision Making with AI. Optimization, exploring tradeoffs between competing objectives, simulation.
Session 4: Introduction to Agents. From Python scripts to software apps to AI agents. System prompt, tools, and connections. Prompt-driven data retrieval, analysis, charts, and reports.
Session 5: AI Security. Enterprise data agreements, prompt injection, and data exfiltration defenses. Risks with Claude Cowork vs Claude Code. Mitigating risks with custom agents. Anonymization and containers.
Session 6: Deep Dive into Agents. Build a personal-assistant agent. Build secure agents that query databases, perform data analysis, and generate reports.
Session 7: Additional Topics. Teach AI your company’s documents, perform predictive modeling, run open-source models locally, and fine-tune small models.
Session 8: Capstone. Pitch your implementation plan and get peer feedback.
Each session is 90 minutes and includes instruction, demos, and group discussion.