A series on adopting AI after 30 years in data engineering. The thrills, the stumbles, and the mindset shift that changes everything.
I know what "early" feels like. Novell replacing the VAX room. SQL Server on command line. Now AI. But this time, being early is humbling in a way I didn't expect.
Part 2AI "optimized" a working stored procedure and the results went wrong. Here's the framework I built to make sure AI changes never break tested code again.
Part 3After 30 years of doing things by hand, muscle memory is your biggest obstacle. The daily discipline of asking "could AI do this?" is where the real transformation happens.
Part 4AI went from chatbot to control plane. Here's how I connected it to Snowflake, Jira, GitLab, Talend, and every other system I work with — one API token at a time.
Part 5AI without business context has the same problem as offshore outsourcing — fast output, wrong results. Domain expertise is the real differentiator.
Part 6An emergency project, an unfamiliar deployment process, and zero training. Here's how documentation + AI + domain expertise got it done on the first try.
Part 7A teammate said he still had three hours left. I finished in thirty minutes while he watched. Here's why domain expertise + AI changes the math.
Part 8Two years decommissioning a database. AI compressed the final stretch into four months — finding dependencies buried three joins deep that manual reviews kept missing.
Part 9AI will over-engineer a one-character fix into an hour-long deployment if you let it. Three lessons from one week on knowing when to grab the reins.