I had a call this week with Maya. (name changed for privacy)
She’s a supply chain data analyst at Nordstrom. Been in the role about a year.
A few weeks ago, her manager connected Claude Code to their database through an MCP server.
Now, dashboards that used to take a week to build manually in IBM Cognos — joining tables, wiring up data sources — take 10 minutes.
A week of work. Down to 10 minutes.
I asked how she learned to do it.
“It was my manager who did that,” she said. “He taught me how to use it.”
Then she paused.
“I want it to be the other way around — where I’m the one telling him what’s possible.”
That’s why she joined our AI Project Challenge.
And she’s not wrong to want that. Here’s why learning this matters:
- MCP connects AI directly to your existing data — no exports, no manual wrangling
- It answers questions about the data that nobody on the team could answer before
- It works even when you’re new and don’t know the systems inside out
The people who learn this first become the ones others turn to.
Build before you’re asked,
Kirill
P.S. If you want to get hands-on with exactly this — building real AI tools on real data — our next AI Engineering Sprint cohort is open 👉 check it out here