AI-driven conversion-rate optimization
Replace the CRO consultant with an LLM that watches every visitor, names the friction, and emits a paste-ready fix. The auto-loop ships improvements while your CRO contract is still being negotiated.
One-line workflow
$ eyepup todo --limit 5
How it works
Step 1
Profile
Every visitor gets an LLM-grounded profile — heat score, persona, blocked-by reason.
Step 2
Cluster
Profiles cluster into 3-25 friction patterns ranked by impact = users × intent × drop-off.
Step 3
Recommend
Each pattern carries a confidence-rated recommended action with the file path guess.
Step 4
Iterate
Resolved patterns archive; recurring friction flags as regression. Memory loop never repeats fixes.