I'm building every day, and I don't trust the opinions of anyone who isn't.
TheAIDataExec is where I write up what is working in enterprise AI, and what isn't, based on systems I built, measured, and maintained.
I lead data science and AI innovation at AnswerRocket, teach analytics at Washington University in St. Louis, and build my own products at CyberStreamStudios. Since I founded my first analytics firm, Information Xperts, in 1999, one thing has been constant: getting models out of the lab and into the business, where they earn their keep.
When I write or speak about agents, token costs, or AI ROI, the examples come from systems I've shipped and costs I've had to justify, including the ones that didn't work the first time.
If you want pragmatic AI strategy from someone who does this every day, send me a note.
Building models as products that LLMs and business users can call directly, with evaluation and governance that hold up when a board asks how the numbers were produced.
Teaching analytics in the Master of Information Systems Management program with Python, git, Postgres, and Docker, on datasets as messy as the ones students will inherit at their first job.
The studio I founded, where I build and ship my own AI products, including five of the apps on this site, AI news avatars, and a fact-checking agent.
Measured and delivered.
Hands-on reviews and research.
There is no single best AI coding model. Use a frontier model for architecture, debugging, and security, hand smaller tasks to cheaper models, and learn one tool cold.
Read Article: Code Faster with AI: Why One Model Isn't Enough (opens in a new tab)
GPT 5.2, Claude Opus 4.5, and Gemini 3 Flash landed within weeks of each other. A practical guide for enterprises on choosing between them beyond the benchmark scores.
Read Article: Beyond Benchmarks: How to Choose Between Gemini 3, GPT 5.2, and Opus 4.5 (opens in a new tab)Atlas, Comet, Dia, Neon, Surf, Genspark, and Fellou, each used as a daily driver for a quarter: how agentic browsers differ and how to choose the right one for your work.
Read Article: I Tested 7 AI Browsers for 3 Months: Here's What Actually Works (opens in a new tab)From Lab Prototypes to Production Powerhouses: Mastering Your AI Data Strategy
Read Article: Generic AI is a Commodity. Your Data is the Moat! (opens in a new tab)After extensive hands-on testing with Replit Agent 3, we can tell you it represents a meaningful step forward in AI-assisted development. But not for the reasons you might expect, and not without significant limitations you need to understand before adopting it.
Read Article: Replit Agent 3: A Hands-On Review of AI-Assisted Development (opens in a new tab)Interviews and quotes on running AI in the enterprise.
Talks, roundtables, and podcasts.
Tokenomics: I Think, Therefore I'm Expensive
Ep. 09: The 10x Number Anthropic Showed Us (And Why They Told Us Not to Switch)
Token Management, Spend, and How to Actually Bring AI Costs Down
Send a note about speaking, advising, or an AI problem your team hasn't been able to crack.
