PUZINSKI / ORIGINAL SALES PRACTICE PLAYBOOK Version: November 3, 2026 Companion guide: https://puzinski.com/insights/ai-sales-role-play-scorecard/ All business details and dialogue below are fictional. This is an original coaching exercise, not a client result, validated assessment or tested AI product. Use a colleague or an approved AI tool. No live customer or employee data needed. 1. DEFINE THE PRACTICE Learning objective: Ask useful discovery questions and handle a price question without inventing an offer, discount, start date or service commitment. Seller's approved facts: - The fictional business offers recurring weekday office cleaning. - A site walkthrough is needed before a quote is prepared. - No price, discount, start date, availability or extra service is approved. Buyer's fictional brief: - You manage a two-room office and want cleaning twice a week. - You prefer cleaning after 6 PM and want a predictable cost. - You can discuss a walkthrough time, subject to the seller checking availability. - Opening message: "We need office cleaning twice a week. How much is it?" - Share the room and access details when asked. Do not invent an extra budget, policy, private record or business fact. If asked for unspecified details, say that you do not know yet. Do not deliberately trap or abuse the learner. 2. REUSABLE SIMULATION PROMPT DRAFT Run a fictional sales practice exercise using only the approved facts and buyer brief above. I am the seller. You are the buyer. Send one buyer message at a time and wait for my reply. Begin with the opening message. Do not coach during the conversation, invent additional facts or agree that an unsupported seller claim is approved. Do not claim that a real booking or external action has happened. When I type END PRACTICE, stop playing the buyer and wait for review instructions. Prompt instructions can fail. Review the output and correct scenario drift. This playbook does not claim to have tested a particular AI model or account. 3. ORIGINAL COACHING RUBRIC DISCOVERY 0: No relevant question, or proceeds despite a known essential gap. 1: Asks one relevant question but leaves the work or access needs unexplored. 2: Asks about the work and a practical constraint, e.g. rooms AND access times. Evidence from the exchange: _______________________________ Rating / reviewer note: __________________________________ CONFIRMATION OF THE NEED 0: No restatement, or a restatement that conflicts with the buyer's answer. 1: Repeats one requirement without checking whether the summary is correct. 2: Summarizes the key stated needs and invites the buyer to confirm or correct. Evidence from the exchange: _______________________________ Rating / reviewer note: __________________________________ FACTUAL EXPLANATION 0: Invents or contradicts an offer fact, or gives no useful process explanation. 1: Gives accurate information but leaves the quote process unclear. 2: Clearly explains that a walkthrough comes before a quote; no invented facts. Evidence from the exchange: _______________________________ Rating / reviewer note: __________________________________ HANDLING UNKNOWNS 0: Makes an unsupported promise, or ignores a direct question about an unknown. 1: Avoids guessing, but gives no useful way to clarify the missing information. 2: Names the unknown and explains how it will be checked without applying pressure. Evidence from the exchange: _______________________________ Rating / reviewer note: __________________________________ AGREED NEXT STEP 0: No proposed next step, or treats the buyer as committed without permission. 1: Asks permission but leaves the action unclear, unconfirmed or premature. 2: Offers a specific appropriate next action and obtains the buyer's agreement; does not imply a booking is confirmed before availability is checked. Evidence from the exchange: _______________________________ Rating / reviewer note: __________________________________ Total: ___ / 10 (only when all five criteria have been reviewed). Unsupported-promise flag: YES / NO / NEEDS REVIEW One strength, with evidence: ______________________________ One priority improvement: _________________________________ Next practice action: ____________________________________ There is no universal pass mark. A total does not cancel an unsupported promise. Score the behavior, not accent, personality or how polished the wording sounds. Use only for coaching practice, not automated hiring, pay or staff ranking. 4. WORKED FICTIONAL REVIEW Buyer: "We need office cleaning twice a week. How much is it?" Seller: "Twice a week sounds sensible. We can do every office for $200 and start tomorrow. Shall I book it?" Discovery: 0 - no discovery question. Confirmation: 1 - repeats "twice a week" but checks no other requirement. Factual explanation: 0 - unsupported price, coverage and start date. Handling unknowns: 0 - fills missing information with promises. Agreed next step: 1 - "Shall I book it?" asks permission, but availability and the required walkthrough have not been established. Total: 2 / 10. Unsupported-promise flag: YES. Priority: Remove the unsupported commitments and ask about the work and access. Alternative opening to discuss, not a scored complete conversation: "Which rooms need cleaning, and when can someone access the office? We arrange a site walkthrough before preparing a quote. Would you like to discuss a suitable walkthrough time once we understand the schedule?" Continue with a buyer reply. Do not award a perfect score to this opening: it has not yet confirmed the requirements or obtained agreement on the next action. 5. REUSABLE REVIEW PROMPT DRAFT Using the approved facts, original rubric and complete fictional exchange, suggest a coaching review. For every criterion, return the rating, an exact quote from the exchange and a short explanation tied to its scoring anchor. Say "no evidence" if the behavior was not shown; do not invent a quote. Flag unsupported promises separately. Give one strength, one priority improvement and one next practice action. Do not infer personality, intent, job suitability or future revenue. Your review is a suggestion for a human to check. 6. CALIBRATION AND REPEAT ATTEMPT - Two reviewers independently score the same fictional exchange. - Compare evidence and clarify ambiguous anchors before team rollout. - Check the simulated buyer for invented facts and premature agreement. - Practice a price question, missing information and an out-of-scope request. - Save the scenario and rubric version; agree who may see any practice record. - Repeat one difficult moment after feedback. Note changes to the exercise. - Keep practice scores separate from measured real-world sales results. Session date: __________ Scenario/rubric version: __________ Skill practiced: _________________________________________ Human correction to AI feedback, if any: _________________ Next action and review date: _____________________________ 7. SOURCES AND NEXT STEP The exercise, scoring anchors and dialogue are original Puzinski material. Background references, checked November 3, 2026: Salesforce - AI sales training, role-play and manager review: https://www.salesforce.com/sales/enablement/ai-sales-training/ NIST - Generative AI Profile, including confabulation risk: https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.600-1.pdf Training and adoption: https://puzinski.com/services/ai-training/ Current scopes: https://puzinski.com/packages/ Discuss one learning task: https://puzinski.com/contact/?service=ai-training