Add two analytics widgets to ChatGPT (and API org dashboards), defaulting to the rolling last 7 days (configurable): Top 10 Missed Questions — prompts that received either no answer or an explicit “I don’t know / can’t help” response. Top 10 Most Asked Questions — the most frequently occurring user questions in the window. These insights help admins/trainers quickly spot knowledge gaps and prioritize improvements. Problem Hard to see which questions users repeatedly ask (to create canned answers, docs, fine-tunes, tools). Hard to see where ChatGPT fails or abstains (to add knowledge, adjust policies, or craft fallbacks). Teams waste time sifting logs and miss obvious wins without this visibility. Goals & Outcomes Faster gap closure: Identify weak spots and add targeted content/tools. Higher resolution rates: Reduce “no answer” moments with better coverage. Improved CX: Standardize responses to commonly asked questions. Definitions Missed Question (MQ): Any user message that (a) received no model reply, or (b) received a reply matching an abstention pattern (e.g., “I don’t know,” “I can’t help with that,” safety refusal). Question Canonicalization: Normalize semantically similar prompts into a single canonical cluster/representative. Most Asked Question (MAQ): Highest-frequency canonical question cluster within the selected window (defaults to last 7 days, rolling).