THE PRACTICE / Topic · AI Implementation

AI Implementation — Insights

What it takes to ship AI into production stacks. Custom builds, pre-built systems, security, integration — the unglamorous engineering that decides whether AI sticks.

ARTICLEOct 2026
Data pipeline diagram with a warning icon before an AI agent node
Fix the Data Before You Build the Agent

Companies race to deploy AI agents on messy, fragmented data and wonder why outputs are wrong. The model is not the problem. The data underneath is.

Frederike Falke
TOOL REVIEWMay 2026
Isometric illustration of granola spilling from a spiral notebook with purple accents, next to a bag and spoon
A Small Tool That Gets One Thing Very Right: Granola

Granola combines your live notes with call recordings to produce one clean output. No bots joining calls. No sprawling feature sets. Just the bridge between human notes and AI transcription done right.

Frederike Falke
ARTICLEMay 2026
A complex purple turbine beside a simpler grid puzzle showing a structured pattern of translucent spheres
Why Most AI Projects Succeed at Build and Fail at Adoption

We see this pattern across every industry we work in — market research, law, real estate, consulting. The build phase is complete. The system works. The handoff slides look good. Six months later, half the team has switched back to Excel.

Frederike Falke
ARTICLEApr 2026
Translucent pink and purple handshake intertwined with a glowing contact profile frame and swirling ribbons
Stop Treating Contacts Like Rows in a Database

Most CRMs turn people into data points. You lose the human connection when every interaction becomes a form to fill. Here's why conversational AI changes everything about how we manage relationships.

Frederike Falke