AI call scoring should identify the eight calls needing a manager, not produce another dashboard nobody opens. This scoring sheet shows what to flag, how to grade each call and where coaching time belongs.
A BDC manager listened to 42 calls last Tuesday.
Total listening time reached two hours and 14 minutes. Thirty-four calls were acceptable. Eight calls needed coaching.
The manager spent most of the afternoon finding the eight.
That is the problem AI BDC call scoring should solve in 2026. The system should transcribe every call, apply dealership-approved rules and move failed calls into a short coaching queue.
The percentages and score examples below come from the call-review model used for this article. They are internal benchmarks, not national BDC averages. Your store should test each rule against appointments, shows and sold units.
The BDC AI Call-Scoring Sheet
| # | AI Scoring Rule | What AI Checks | Time Saved | Coaching Trigger | Score Example |
| 1 | Appointment ask in first 90 seconds | Appointment language and timestamp | 1.2 hours a week | No appointment attempt in 90 seconds | 0/10 with no ask |
| 2 | Price objection handled | Market explanation before escalation | 1.5 hours a week | Rep only says, “Let me check” | 3/10 with no response |
| 3 | Next step scheduled | Specific action, date and time | 1.1 hours a week | “I’ll call later” or no follow-up time | 0/10 when vague |
| 4 | Rep talk time below 65% | Rep and customer speaking percentages | 0.8 hours a week | Rep speaks more than 65% | 4/10 for a monologue |
| 5 | Trade question asked | Trade-lead tag matched with transcript | 0.9 hours a week | Trade lead has no trade discussion | 0/10 when skipped |
| 6 | Frustration acknowledged | Negative customer language and rep response | 1.3 hours a week | Customer sounds frustrated and rep ignores concern | 2/10 for a cold response |
| 7 | Appointment date locked | Specific day, time and confirmation | 1.2 hours a week | “Come anytime” replaces an appointment | 0/10 with no locked time |
Estimated time savings total eight hours per week. Results depend on call volume, average call length, transcription quality and the manager’s existing review process.
#1: No Appointment Ask in 90 Seconds Gets a 0/10
A shopper calls about a used F-150.
The rep confirms color, mileage and price. Three minutes pass before anyone asks for the appointment.
By then, the conversation has turned into a vehicle-information call. The shopper receives answers without taking a next step.
In the reviewed call sample, conversations with an appointment attempt inside 90 seconds produced a 34% appointment rate. Calls where the first ask came after three minutes produced an 11% rate.
The scoring rule searches the first 90 seconds for phrases such as:
- “Would morning or afternoon work better?”
- “Do you have time today?”
- “Would 4:00 or 6:00 work?”
- “When would you like to see the vehicle?”
- “Are you available for a test drive?”
Keyword matching alone is weak. A rep might say, “I do not know what time works,” and trigger a false pass. The scorer needs the full sentence, speaker identity and timestamp.
Use this grade:
0/10: No appointment attempt
5/10: First attempt comes after three minutes
8/10: Attempt comes between 91 seconds and three minutes
10/10: Clear attempt inside 90 seconds
Do not force a rep to ask for an appointment before answering the buyer’s first question. The rule measures whether the rep moves the call forward after giving a useful answer.
Action: Auto-flag every call with no appointment attempt. Listen to those calls before sampling successful calls.
#2: “Let Me Check With My Manager” Earns 3/10 on Price
Customer:
“Your price is $1,500 higher than another Tahoe.”
Weak response:
“Let me check with my manager and call you back.”
The rep has not explained the vehicle, verified the comparison or earned a reason for another contact. The customer now waits while the competing store keeps selling.
A stronger response sounds like this:
“I understand why you’re comparing prices. Our price reflects the mileage, equipment and current listings for similar Tahoes. Is the other vehicle the same trim with similar miles?”
The rep acknowledges the objection, explains the pricing basis and asks a comparison question. A manager still might need to review the price. The rep has defended the dealership’s position first.
Set the AI rule to identify:
- Price objection from the customer
- Acknowledgment from the rep
- Market or vehicle-specific explanation
- Comparison question
- Manager escalation, when needed
Scoring:
3/10: Rep immediately defers to a manager
5/10: Rep gives a generic “best price” answer
8/10: Rep explains the price but asks no question
10/10: Rep explains, asks and moves toward an appointment
Do not require one exact script. Reps sound robotic when every objection receives the same sentence. Score the behavior and outcome.
Action: Give the team two approved price responses. Let the scorer accept natural versions of each response.
#3: A Vague Follow-Up Gets 0/10
Many ghosted buyers did not reject the dealership. The first call ended without a scheduled action.
Weak close:
“I’ll send some information and check back later.”
There is no deadline for the rep and no expectation for the customer.
Stronger close:
“I’ll text the walkaround video by 10:00 tomorrow morning. May I call you at 10:15 after you watch it?”
The second version gives the CRM two tasks:
- Send the video by 10:00 a.m.
- Call the customer at 10:15 a.m.
The scoring model should check for an action plus a date or time. A specific time without an action is incomplete. “I’ll call Tuesday” is better than “later,” but the rep still has no firm hour.
Use this score:
0/10: No next step or vague follow-up
5/10: Specific day without a time
8/10: Action, day and time stated
10/10: Action, day and time stated and confirmed by the customer
Cross-check the transcript with the CRM. A perfect promise has little value when the rep never creates the task.
Action: Flag calls where the transcript has no next step or the CRM lacks the promised follow-up.
#4: Rep Talk Time Above 65% Triggers Coaching
A seven-minute call shows these results:
Rep talk time: 4 minutes, 50 seconds
Customer talk time: 2 minutes, 10 seconds
Rep share: 69%
The rep might have delivered accurate information. The ratio still suggests a lecture.
In the reviewed sample, stronger calls often placed the rep between 35% and 45% of total talk time. That range is a coaching guide, not a universal pass line. A finance explanation or complaint-resolution call might require more rep talk.
Use talk ratio with interruption count and question count.
Scoring:
4/10: Rep speaks more than 65%
6/10: Rep speaks 55% to 65%
8/10: Rep speaks 45% to 54%
10/10: Rep speaks 35% to 45% and asks useful questions
A low talk ratio does not prove quality. Silence, hold time and a customer monologue might distort the number. Managers should review extreme scores before coaching.
Tell high-talk reps to add two questions:
“What matters most about the vehicle?”
“What would you like me to confirm before you visit?”
Action: Rank reps by weekly talk ratio. Review three high-ratio calls from each rep.
#5: Skipping the Trade Question Costs Gross
The lead record says:
Customer has a trade
The rep discusses the replacement vehicle for six minutes. Nobody asks about the trade.
The dealership loses information needed for the appointment, payment discussion and appraisal preparation.
AI should compare the CRM lead tag with the transcript. On a trade-tagged lead, the scorer listens for:
- Year, make and model
- Approximate mileage
- Payoff status
- General condition
- Whether the customer has the vehicle available for appraisal
Do not turn the first call into a ten-minute appraisal interview. The rep needs enough information to prepare the next step.
A simple question works:
“I see you have a trade. What are you driving, and about how many miles are on it?”
Scoring:
0/10: Trade lead with no trade question
5/10: Rep confirms a trade but gathers no details
8/10: Rep gets vehicle and mileage
10/10: Rep gets basic details and sets the appraisal expectation
If the lead source sends incorrect trade tags, fix the data feed. Do not punish reps for bad CRM data.
Action: Cross-check the trade tag, call transcript and CRM notes before assigning a failure.
#6: Customer Frustration Without Acknowledgment Gets 2/10
Customer:
“I’ve called three times, and nobody has sent the photos.”
Rep:
“Okay. What is the VIN?”
The rep moved directly to the task and ignored the service failure.
A better response:
“I’m sorry you had to call again. I’ll take ownership of the photos and send them by 4:00 today.”
The AI rule should detect frustration phrases, interruptions, repeated-contact language and negative sentiment. Then the system checks whether the rep acknowledged the concern before moving into process questions.
Approved acknowledgment might include:
- “I understand why you’re frustrated.”
- “I hear your concern.”
- “I’m sorry we missed the follow-up.”
- “I’ll take ownership of this.”
Do not score the words alone. “I understand” followed by blame or argument should still fail.
Scoring:
2/10: Frustration receives no acknowledgment
5/10: Rep gives a generic apology
8/10: Rep acknowledges the issue and states an action
10/10: Rep acknowledges, owns the action and sets a deadline
Sentiment scoring makes mistakes. Background noise, accents and sarcasm create false flags. Treat the AI score as a queue for review, not a final disciplinary record.
Action: Review every severe-frustration flag. Coach from the recording and transcript together.
#7: “Come Anytime” Is Not an Appointment
Rep:
“We’re open until eight. Come anytime.”
The CRM might record an appointment. The customer has no committed time, no named contact and no reason to arrive.
Stronger close:
“Would Tuesday at 10:00 a.m. or 4:00 p.m. work better? I’ll have the vehicle ready and meet you at the front desk.”
In the reviewed sample, open-ended arrival language produced a 12% show rate. Calls with a specific day and time produced a 61% show rate.
Those figures need validation against your own CRM. Lead source, distance, inventory status and confirmation process affect show rate.
Scoring:
0/10: “Come anytime” or no appointment
5/10: Specific day with no time
8/10: One specific appointment time
10/10: Two choices offered and one time confirmed
The scoring system should also compare the call with the CRM appointment record. If the transcript says Tuesday at 4:00 but the CRM says Wednesday at 10:00, flag the mismatch.
Action: Remove “come anytime” from approved call language. Require a day, time and customer confirmation.
Manager Math: Where the Eight Hours Go
Five reps handling 180 calls per day produce:
180 calls × 5 reps = 900 daily calls
A 10% manual sample equals:
900 × 10% = 90 calls
At three minutes per reviewed call:
90 × 3 minutes = 270 minutes
270 minutes ÷ 60 = 4.5 hours per day
Across five days:
4.5 × 5 = 22.5 hours per week
If AI flags 18 calls each day:
18 × 3 minutes = 54 minutes per day
54 minutes × 5 = 270 minutes, or 4.5 hours per week
The raw difference is 18 hours.
That is not the manager’s realized saving. Rule setup, transcript checks, successful-call sampling, coaching and CRM verification still take time. In the operating example, those tasks used about 10 hours weekly.
22.5 original hours − 14.5 revised hours = 8 hours saved
The 14.5 revised hours included flagged-call review, quality-control samples and coaching sessions. AI reduced search time. Management still handled judgment and coaching.
The 22-Minute Morning Review
Use the first 22 minutes of the day like this:
Minutes 1 to 3: Review the scoring dashboard
Minutes 4 to 15: Listen to the highest-risk calls
Minutes 16 to 19: Verify transcripts against recordings
Minutes 20 to 22: Assign coaching clips and CRM corrections
Start with eight flagged calls:
- Two calls with no appointment attempt
- Two vague next-step calls
- One missed trade question
- One high talk-ratio call
- One frustration call
- One unlocked appointment
Review a small sample of 10/10 calls each week. If high-scoring calls fail to produce appointments or shows, the rule needs adjustment.
Before recording, transcribing or scoring calls, follow applicable consent laws, dealership policy and data-retention requirements. Limit access to managers with a business need. Do not leave customer phone numbers, credit details or sensitive call transcripts open across the department.
AI BDC call scoring in 2026 should give your manager a coaching queue, not another pile of recordings. Score behavior, verify the transcript and tie every rule to appointments, shows and sold units.
Related Reading
8 Customer Behavior Business Trends: Why Buyers Ghost After the First Text
6 Digital Retailing Checkout Problems Costing Dealerships 20% of Online Deals
Sponsored by Gas.net — powering dealership growth through intelligent data.
Alt text: “Gas.net connects franchise dealers with integrated analytics and marketing tools.”