AI Answered Every Buying Question—Without the Brand's Own Website
We tested how AI assistants answered real buying questions about two client brands. Both generated useful answers. Only one brand's website helped shape them.

Technology Executive | Engineer | Advisor
I’m Will. These are notes from twenty-five years of turning technology strategy into practical results, building capable teams, and solving the difficult problems that sit between the plan and the outcome.
The fear of AI leaking trade secrets is understandable. The bigger risk is falling behind the competitors who are already using it.
We tested how AI assistants answered real buying questions about two client brands. Both generated useful answers. Only one brand's website helped shape them.
Version 2.x rebuilds the Pipedrive CRM MCP Server around modern APIs, safer remote access, predictable responses, and a deliberately constrained path from read-only analysis to controlled action.
I had a personal need, built a real tool, then did the work to see if it was a business. The honest answer was no — so I made it open source.
Before AI could prioritize aviation prospects, we had to connect the market data, resolve the identities, and make the operating model trustworthy.
Most AI pilots end at the demo. This one became a daily sales coaching email—and made follow-up gaps harder to ignore.
My takeaways from a Node Talks panel on what businesses really need to turn AI from an interesting experiment into a practical, scalable tool.
Subscribe
No newsletter platform tricks, no funnel. Roughly once a month, when I've finished thinking about something.