Tool decision matrix — eighteen-month review of four note systems

The matrix in the diagram covers four AI-augmented note systems evaluated over an eighteen-month period across five dimensions. The full evaluation methodology and verdict is in Article №02 (Tool Decisions). This output is the raw scoring matrix.

The five evaluation dimensions:

1. Retrieval speed under time pressure. How fast can you find the relevant note when you have three minutes before a client call? Tested by timing retrievals of specific notes placed sixty days earlier. Not a benchmark; actual use.

2. Capture friction. How many steps does it take to go from “I need to capture this” to “it’s captured”? Anything above two steps reliably fails under deadline pressure. Measured by counting mandatory UI interactions.

3. AI integration depth. Does the AI augmentation actually change the output, or is it a chatbot stapled onto a notes interface? Evaluated by running the same summarisation task across all four systems on the same source material.

4. Eighteen-month staying power. The hardest dimension to fake. Is the system still in the workflow at month eighteen, or has it been replaced? Only one system passed this test.

5. Failure mode tolerance. What happens when the system fails or is unavailable? A note system that causes a single point of failure in a client-facing workflow is not a note system — it’s a liability. Tested by deliberately inducing a sync failure and timing recovery.

The scoring: The matrix uses a 1–5 scale across all dimensions for all four systems. The one that survived scored highest on dimensions 1 and 4. It did not score highest on dimension 3 — the system with the best AI integration was not the system still in use at month eighteen. That’s the finding worth noting.