# AI Copilot Quarantine Decision Framework: Live Call Readiness > A scored decision tool that tells managers exactly which reps earn copilot access during live calls, what rules apply at each phase, and how to run the post-call debrief that makes the whole system work. Free framework for sales managers. Topic: [AI Assessment, Scoring & Assurance](https://www.peerlab.ai/library/topic/ai-assessment-scoring-assurance/). Published by PeerLab in the [Sales Enablement Resource Library](https://www.peerlab.ai/library/). Source: https://www.peerlab.ai/library/ai-copilot-quarantine-decision-framework-live-call-readiness/ ## How to use it Score each rep against the mastery threshold checklist before assigning a copilot mode. Run calls under the rules for that mode, then use the divergence log template in every post-call debrief until the rep crosses the threshold or you decide to pull them back. ## What's inside - Mastery threshold checklist with 12 scored criteria across four competency areas - Scoring bands that map directly to three copilot modes: off, minimised, and on - Four-phase deployment map covering pre-call, live-below-threshold, live-above-threshold, and post-call - Specific behavioural rules for each phase, including what the rep may and may not do - Post-call divergence log template for reps to record every point where they deviated from a copilot suggestion - Structured debrief questions managers use to turn each divergence into a coaching moment - Re-assessment trigger criteria so managers know when to re-score a rep - Honest "when this framework breaks" notes for edge cases ## What This Framework Does AI copilots on live calls create a dependency risk. Reps who cannot yet filter suggestions intelligently become prompt-followers rather than salespeople. Reps who can filter intelligently often improve faster without the sidebar than with it. This framework gives managers a defensible, scored answer to three questions: 1. Which reps should have the copilot switched off, minimised, or on during live calls? 2. What rules apply at each phase of the call cycle for each mode? 3. How do you run the post-call debrief so the divergence between rep decisions and copilot suggestions becomes real learning? --- ## Part 1: Mastery Threshold Checklist Score each criterion 0, 1, or 2. 0 = not demonstrated. 1 = inconsistent. 2 = consistent across recent calls. ### Competency Area A: Discovery and Qualification | Criterion | 0 | 1 | 2 | Score | |---|---|---|---|---| | A1. Can identify the real buying trigger without being prompted | | | | | | A2. Holds silence after a question without filling it | | | | | | A3. Adjusts qualification depth based on live signals, not a fixed script | | | | | ### Competency Area B: Objection Handling | Criterion | 0 | 1 | 2 | Score | |---|---|---|---|---| | B1. Distinguishes a real objection from a deflection in the moment | | | | | | B2. Responds to the emotion before the content of an objection | | | | | | B3. Does not default to proof points before understanding the objection fully | | | | | ### Competency Area C: Conversational Control | Criterion | 0 | 1 | 2 | Score | |---|---|---|---|---| | C1. Can re-anchor the conversation after a prospect tangent | | | | | | C2. Tracks multiple threads and picks the right one to return to | | | | | | C3. Reads pace and adjusts without being told | | | | | ### Competency Area D: Copilot Literacy | Criterion | 0 | 1 | 2 | Score | |---|---|---|---|---| | D1. Has reviewed at least 10 post-call divergence logs and discussed them with manager | | | | | | D2. Can articulate why a copilot suggestion would have been wrong in at least two recent examples | | | | | | D3. Does not glance at the sidebar during high-stakes pauses (observed in role-play or review) | | | | | **Maximum score: 24** ### Threshold Bands | Total Score | Copilot Mode | |---|---| | 0-12 | **Copilot ON** (rep is not yet ready to filter intelligently; full suggestions active) | | 13-18 | **Copilot MINIMISED** (sidebar available but collapsed; rep may consult between stages, not mid-exchange) | | 19-24 | **Copilot OFF** (rep works unassisted; full post-call divergence log required) | Re-score after every 15 live calls, or immediately following any call where the rep visibly over-relied on or ignored the sidebar in a way that cost the conversation. --- ## Part 2: Four-Phase Deployment Map ### Phase 1: Pre-Call (All Modes) The copilot is a preparation tool here, regardless of mode. No restriction applies. **Required actions:** - Rep reviews copilot-generated account summary and flags any data they believe is wrong or missing. - Rep writes their own call objective in one sentence before reading the copilot's suggested objective. - Rep notes two questions they would ask without any prompting, then compares them to the copilot's suggested openers. The comparison is not about who is right. It is about building the habit of having an independent view before the tool speaks. --- ### Phase 2: Live Call, Copilot ON (Score 0-12) The sidebar is visible and active. This is not a failure state. It is a developmental stage. **Rules:** - Rep may glance at suggestions, but must finish the prospect's sentence or pause before acting on any prompt. - Rep must not read a suggestion verbatim. They must translate it into their own voice. - Rep must note at least one moment per call where they consciously ignored a suggestion. If they cannot identify one, that is a coaching flag. **Failure mode to watch for:** The rep stops listening to the prospect and starts managing the sidebar. If you observe this on a recording, score D3 as 0 and re-assess. --- ### Phase 3: Live Call, Copilot MINIMISED (Score 13-18) or OFF (Score 19-24) **Minimised rules:** - Sidebar is collapsed. Rep may open it only during a scheduled pause, defined as: after discovery is complete, after presenting a proposal, or during a hold moment initiated by the prospect. - Rep may not open it mid-exchange, mid-objection, or mid-close. - Each time the rep opens the sidebar during a non-permitted moment, the manager notes it. Three instances in five calls triggers a re-score. **Off rules:** - Copilot records in the background but the sidebar is not accessible during the call. - Rep has full autonomy and full accountability. - The post-call divergence log (Part 3) is mandatory after every call. **Failure mode to watch for (both modes):** The rep compensates by over-preparing scripts and loses conversational flexibility. Minimised and off modes should increase presence, not increase scripting. If transcripts show more scripted language than in Copilot ON calls, raise it in the debrief. --- ### Phase 4: Post-Call (All Modes) The debrief is the real learning event. This is where the gap between copilot suggestions and rep decisions becomes explicit and teachable. See Part 3 for the full template. --- ## Part 3: Post-Call Divergence Log and Debrief Template The rep completes this within two hours of the call, before memory degrades. ### Divergence Log (Rep Completes) **Call date:** ____________ **Prospect/account:** ____________ **Copilot mode used:** ON / MINIMISED / OFF **Overall call outcome:** ____________ | # | Copilot suggestion | What I did instead | My reason at the time | Was I right? (Yes / No / Unsure) | |---|---|---|---|---| | 1 | | | | | | 2 | | | | | | 3 | | | | | | 4 | | | | | Add rows as needed. If the rep is in Copilot ON mode and cannot identify any divergences, that is the primary coaching topic. --- ### Manager Debrief Questions Use these in sequence. Do not skip to question 4 if the rep is still stuck on question 2. **Q1. Walk me through divergence number [X]. What was the prospect doing or saying that made you go a different direction?** This checks whether the rep was responding to the actual conversation or acting on instinct disconnected from the call. **Q2. What do you think the copilot was pattern-matching on when it made that suggestion?** This builds copilot literacy. Reps who understand why the tool suggests what it does are harder to manipulate by it. **Q3. If the copilot had been right and you had been wrong, what would that have looked like in the call?** This protects against reps who mark every divergence as "I was right" without examining it. **Q4. Which of your divergences are you least confident about? What would you need to see in the next three calls to know whether it was a good instinct?** This turns individual calls into a learning sequence rather than isolated events. **Q5 (Copilot OFF reps only). Is there anything on the transcript you wish you had caught in the moment?** This is not a trap. It normalises the honest answer: the copilot would have flagged this faster than me. That is a data point, not a verdict. --- ## When This Framework Breaks - **High-volume, low-complexity calls:** If average call length is under eight minutes and the product is transactional, the mastery threshold is less relevant. The copilot is a speed tool, not a judgment tool, and the framework applies at full weight only where judgment is actually required. - **New-to-role managers:** If the manager has not done enough live call review to score criteria A1-C3 reliably, the checklist scores will be noise. Calibrate first by reviewing the same five calls together before scoring independently. - **Copilot tools with no post-call transcript:** The divergence log requires a transcript or recording. If your tool does not produce one, this framework cannot be run as written. --- ## Re-Assessment Trigger Criteria Re-score a rep immediately (do not wait for the 15-call cycle) if: - Call recording shows rep opening the sidebar more than twice per call in Minimised mode across two consecutive calls. - Rep's divergence log shows "unsure" or "no" in the final column on more than 40% of logged divergences across three calls. - Rep's close rate drops more than 15 percentage points relative to their pre-framework baseline over a rolling 20-call period. - Rep self-reports that they feel worse at the job since the copilot was turned off. Take this seriously. It may indicate the off mode was applied before the rep had enough foundational reps to sustain confidence without the tool. ## The thinking behind this [Your AI Copilot Is Training Your Reps to Sell to the Algorithm, Not the Buyer](/blog/ai-copilots-are-coaching-reps-to-perform-for-the-tool/) is the argument this resource puts to work.