Quick answer: An AI BDC should answer every inbound sales and service call, book test drives and service appointments, read stock and repair-order status and take trade-in details, then hand to a person when a call turns to price, trade value, credit data, a safety symptom, an angry customer or a request the AI has twice failed to understand. Of the 12 common BDC tasks in the table below, the 4 outbound ones carry TCPA consent rules for AI voices (FCC 24-17, 2024).
If you are searching “ai bdc car dealership”, you are usually choosing between four ways to staff the phones: an in-house business development centre (BDC), an outsourced BDC, an AI voice agent, or a hybrid. Below: a task-by-task handoff table, a comparison of the four models, a missed-call calculation built on NADA’s 2025 dealership data, and the two US rules that change the answer – the FTC Safeguards Rule (16 CFR Part 314) and the TCPA as applied to AI voices by the FCC. Dealer-specific vendors Numa and Impel are described only from their own pages, read on 8 October 2026.
What does a car dealership BDC actually do?
A dealership BDC is the team that owns the phone and the lead queue so that salespeople and service advisors do not have to. In practice it works two lanes that behave very differently.
The sales lane handles internet leads from the dealer website and third-party listing sites, inbound calls about specific vehicles in stock, test-drive appointments, trade-in enquiries, and follow-up on shoppers who went quiet. The value per call is high, and the conversation drifts quickly towards money: payment, trade value, rate.
The service lane handles appointment booking, “is my car ready?” status calls, recall and maintenance reminders, and follow-up on declined work. The calls are repetitive, which is why they suit automation. NADA’s 2025 annual profile counts 16,990 franchised light-vehicle dealers writing more than 276 million repair orders, and puts the average dealership at 16,252 repair orders and $9,687,942 of service and parts sales for the year. Service and parts were 13.3% of total dealership sales dollars in 2025 – a modest share of revenue spread across a very large number of separate customer visits.
That split is the first criterion for choosing a model. A service-lane call is usually a lookup and a booking; a sales-lane call is usually the start of a negotiation. An AI voice agent can finish the first kind and should only start the second.
Which BDC tasks should an AI voice agent handle?
The table below is the core of this page: for each common BDC task, where the AI’s job ends and the trigger that should move the call to a person. The triggers follow one rule we call the money-or-identity rule: the AI may gather, look up and book, but any turn towards a number the store has to stand behind (a price, a trade value, a payment, a repair estimate) or towards financial identity data (a Social Security number, income, a credit application) goes to a human or to a secure form.
| Task | Lane | AI handles | Hand to a human when | Rule that bites |
|---|---|---|---|---|
| Inbound call about a vehicle in stock | Sales | Confirms the unit is listed, reads listed features, books a test drive | Caller asks for an out-the-door price, a discount or a payment figure | Money-or-identity rule |
| After-hours inbound sales call | Sales | Answers within seconds, qualifies timeline and vehicle, books a showroom visit | Caller asks to negotiate or to hold a vehicle with a deposit | Money-or-identity rule |
| First outbound call to a new internet lead | Sales | Only if the lead’s consent covers artificial-voice calls to that number | No written consent record exists for that number: a person dials by hand instead | 47 CFR 64.1200(a)(2); FCC 24-17 |
| Trade-in enquiry | Sales | Collects VIN, mileage, condition notes and payoff lender name; books an appraisal | Caller wants a trade value on the phone | Money-or-identity rule |
| Financing or credit question | Sales / F&I | Explains the process and sends the store’s secure application link | Caller starts reading out a Social Security number, income or loan details | 16 CFR 314.2 (application data is customer information) |
| Stale-lead follow-up call (outbound, 30+ days quiet) | Sales | Only with written consent for artificial-voice telemarketing on record | Consent cannot be shown: a person dials | 47 CFR 64.1200(a)(2), (a)(3) |
| Service appointment booking | Service | Checks scheduler availability, books, confirms by text | Caller describes a safety symptom (brakes, steering, warning lights) or asks for a diagnosis | Advisor judgement |
| “Is my car ready?” | Service | Reads repair-order status from the DMS and gives the promised time | Status shows a delay or added work needing approval | Money-or-identity rule |
| Recall or maintenance reminder (outbound) | Service | Only with prior express consent to that wireless number; residential lines have a three-call, 30-day allowance for non-telemarketing calls | The reminder adds an offer or upsell (it becomes telemarketing) | 47 CFR 64.1200(a)(1), (a)(3)(iii) |
| Declined-work follow-up (outbound) | Service | Only with written consent for artificial-voice telemarketing on record | Customer asks for a revised estimate | 47 CFR 64.1200(a)(2); money-or-identity rule |
| Complaint or “heat case” | Either | Captures the facts and the repair-order number | Immediately – the second time a caller expresses anger, or at once if they ask for a manager | Retention judgement |
| Caller the AI cannot understand | Either | Repeats once in plain words, offers another language | After two failed attempts to capture the intent | Call-quality rule of thumb |
Four of the twelve rows are outbound calls. Three depend on consent records the AI cannot create for itself; the fourth, a reminder with no offer, needs prior express consent to a mobile and otherwise relies on the residential three-calls-in-30-days allowance. The other eight rows are inbound or in-call work, where an AI BDC earns its keep with the least legal friction: 47 CFR 64.1200(a) restricts calls a business “initiate[s]”, and FCC 24-17 applies its requirements to AI “that initiates any outbound telephone call”. The “two failed attempts” and “second expression of anger” thresholds are our own rules of thumb, not regulatory limits; set yours and write them into the agent’s instructions.
Human BDC vs outsourced BDC vs AI BDC vs hybrid
Once the task boundary is drawn, the choice of model is mostly about hours, data access and who carries the compliance work. Each model is scored on the same seven criteria.
| Criterion | In-house human BDC | Outsourced BDC | AI BDC (voice agent) | Hybrid (AI first, humans on handoff) |
|---|---|---|---|---|
| Hours covered | Staffed hours only; 24/7 single-seat cover needs 4.2 full-time seats (168 ÷ 40) | Whatever the contract buys | 24/7, up to the concurrent-call capacity the vendor provisions | 24/7 first answer; humans during staffed hours |
| Main cost driver | Wages: NADA’s 2025 average for all dealership employees is $1,619 a week | Per-contract fee | Platform fee plus integration work | Platform fee plus a smaller human team |
| Live DMS and scheduler access | Yes, with staff logins | Only if you grant outside logins | Only through an integration you configure and test | AI through integration, humans through logins |
| Trade values, prices, credit | Yes, within store policy | Only if your store delegates that authority by contract | Some dealer vendors offer trade valuation; under the money-or-identity rule, collect and hand off | Humans only |
| Outbound consent burden | A person dialling by hand is outside the artificial-voice and autodialer limbs of 64.1200(a) | Same as in-house if agents dial by hand; autodialers bring 64.1200(a) back in | Highest: each outbound AI-voice call needs the consent shown row by row above | Route consent-less numbers to humans |
| Safeguards Rule service-provider duties (16 CFR 314.4(f)) | None added | Added: select, contract, periodically assess | Added: select, contract, periodically assess | Added for the AI vendor |
| Wrong for | Stores whose calls peak outside staffed hours | Stores whose callers need live repair-order status the outsourcer cannot see | Stores that expect it to negotiate or take credit applications, or to quote trade values with no named person standing behind the number | Stores unwilling to staff the handoff – an unanswered transfer undoes the AI’s work |
Read across the bottom row and the pattern is clear: each model fails at the boundary the store did not staff. A human BDC fails after hours, an outsourced BDC fails on data it cannot see, an AI BDC fails when it is asked to commit the store to a number, and a hybrid fails when the handoff rings out. Our comparison of AI voice agents and call centres on cost and quality covers the outsourced-versus-AI trade in more depth for non-dealer businesses.
Impel’s Voice AI page lists what happens when a transfer is not picked up: “Transferred, no answer → Message captured. Task created. SMS sent.” A sound fallback – and a reminder that the human side of a hybrid still has to work the task list.
If you are weighing a general-purpose platform rather than a dealer-specific one, check what it publishes about integrations before you assume DMS access. Zian, for example, publishes CRM integrations (HubSpot, Salesforce, HighLevel, Zapier), calendar and knowledge-base connections and an API for custom workflows; its Customer Support Agent is listed as 24/7 and 30+ languages, and the platform lists live phone, SMS, email and WhatsApp, which matters for stores whose callers do not all speak English. A DMS lookup on a platform like that is an API build you would scope and test, not a checkbox.
Want an AI agent answering every call and booking while your team handles the handoffs? Zian is in partnership-application beta. Apply For Partnership
What does a missed service call cost a dealership?
Rerun this with your own numbers. The two dealership inputs come from NADA’s 2025 annual profile; the loss rate is an assumption, and we show three versions of it because no figure we could verify at source tells you yours. Pull your own unanswered-call rate from your phone system or call-tracking report and substitute it.
| Step | Input | Value | Status |
|---|---|---|---|
| A | Repair orders written by the average franchised dealership, 2025 | 16,252 | Sourced: NADA Data 2025 |
| B | Service and parts sales per customer-pay repair order, 2025 | $494 | Sourced: NADA Data 2025 |
| C | Repair orders lost to calls nobody answered or returned, as a share of A | 1% / 3% / 5% | Assumption – replace with your own |
| D = A × C | Lost repair orders a year | 162.5 / 487.6 / 812.6 | Calculated |
| E = D × B | Lost service and parts sales a year | $80,285 / $240,855 / $401,424 | Calculated |
Three cautions. $494 is sales, not gross profit, so line E overstates what the store keeps. It is the customer-pay average; NADA’s warranty figure is $551. And a lost call is not always a lost repair order – some callers ring back – which is why step C is a share of repair orders, not of calls.
If your BDC is staffed 60 hours a week (an assumption: for example, 10 hours a day, six days), the phones are unstaffed for 108 of the week’s 168 hours, or 64.3% of the week. Covering one line for all 168 hours takes 4.2 full-time seats at 40 hours each. At NADA’s 2025 average weekly earnings for dealership employees of $1,619 – an all-roles figure, not BDC pay, and before payroll tax, benefits and leave cover – one seat is $84,188 a year, 1.5 seats (the 60-hour week) is $126,282 and 4.2 seats is $353,590. Compare your line E against the cost of the extra seats, and the AI-versus-human decision becomes arithmetic rather than opinion. The same logic about response time applies on the sales side; the evidence for it is in our guide to speed to lead with AI agents.
Does the FTC Safeguards Rule apply to an AI BDC?
It can, and dealers should check before an AI agent touches customer data. The FTC Safeguards Rule (16 CFR Part 314, under the Gramm-Leach-Bliley Act) applies to “financial institutions” within the FTC’s jurisdiction. The rule’s own examples speak to dealers directly. Section 314.2(h)(2)(ii) says “An automobile dealership that, as a usual part of its business, leases automobiles on a nonoperating basis for longer than 90 days is a financial institution with respect to its leasing business”. Section 314.2(e)(2)(i)(E) lists, as a continuing customer relationship, an agreement under which you “arrange or broker a home mortgage loan, or credit to purchase a vehicle, for the consumer”. The rule also excludes entities that are not “significantly engaged” in financial activities (314.2(h)(3)(iv)), so whether a given store is in scope is a question for its counsel, not for a vendor.
If you are in scope, five provisions bear directly on an AI voice agent:
- Your phone system may be in scope. The rule’s definition of “information system” (314.2(j)) includes “telephone switching and private branch exchange systems” that contain customer information or connect to a system that does. An AI agent wired into your DMS and phone system is connected to customer information.
- The AI vendor is a service provider. Section 314.4(f) requires you to take reasonable steps to select providers capable of appropriate safeguards, to require those safeguards by contract, and to periodically assess providers based on their risk and the continued adequacy of their safeguards. An outsourced BDC triggers the same three duties.
- Encryption and access. Section 314.4(c)(3) requires encryption of customer information in transit over external networks and at rest, unless the Qualified Individual approves effective alternative compensating controls where encryption is infeasible. Section 314.4(c)(5) requires multi-factor authentication for any individual accessing any information system, unless the Qualified Individual has approved in writing reasonably equivalent or more secure controls. Section 314.4(c)(8) requires monitoring and logging of authorised users’ activity.
- Recordings and transcripts are data you must dispose of. Section 314.4(c)(6) requires secure disposal of customer information no later than two years after it was last used for the customer, unless it is necessary for business operations or other legitimate business purposes, is required by law to be kept, or targeted disposal is not reasonably feasible.
- Breach notice to the FTC. Since 13 May 2024 (314.5), section 314.4(j) requires you to notify the FTC as soon as possible and no later than 30 days after discovering a “notification event” involving the information of at least 500 consumers. A notification event is acquisition of unencrypted customer information without the authorisation of the person it relates to; unauthorised access is presumed to be acquisition unless you have reliable evidence that there was not, and could not reasonably have been, acquisition; and data counts as unencrypted if an unauthorised person accessed the encryption key.
Smaller stores get partial relief. Section 314.6 exempts financial institutions that maintain customer information on fewer than 5,000 consumers from four elements only: the written risk assessment (314.4(b)(1)), continuous monitoring or annual penetration testing and six-monthly vulnerability assessments (314.4(d)(2)), the written incident response plan (314.4(h)) and the annual written report to the board (314.4(i)). The service-provider duties, encryption and MFA still apply. The FTC’s plain-English guide to the rule walks through the remaining elements.
The practical consequence for choosing a model: the money-or-identity rule is also a data-minimisation rule. An AI agent that sends a secure application link instead of hearing a Social Security number keeps the most sensitive data out of call recordings and transcripts. A platform that offers private model deployment on your own infrastructure – Zian publishes that option – changes where the data sits, but it does not remove your Safeguards duties.
Can a dealership use an AI voice agent for outbound calls?
Usually only with consent – and this is the lane where an AI BDC carries the most legal risk. On 2 February 2024 the FCC adopted Declaratory Ruling FCC 24-17 (released 8 February 2024, effective on release), confirming that “the TCPA’s restrictions on the use of ‘artificial or prerecorded voice’ encompass current AI technologies that generate human voices.” The ruling also says the TCPA “does not allow for any carve out of technologies that purport to provide the equivalent of a live agent”, so a conversational agent that sounds human is treated the same as a prerecorded message.
Under 47 CFR 64.1200, as at 8 October 2026:
- Any artificial-voice call to a wireless number needs the called party’s prior express consent unless it is made for an emergency purpose or falls within an exemption (64.1200(a)(1)).
- Telemarketing or advertising by artificial voice to a wireless number needs prior express written consent (64.1200(a)(2)). The two carve-outs in that paragraph – tax-exempt nonprofits and HIPAA health care messages – do not fit a dealership.
- Residential lines: artificial-voice telemarketing needs prior express written consent; a commercial call that is not telemarketing needs no consent if the caller makes no more than three calls in any consecutive 30-day period and honours opt-out requests (64.1200(a)(3)(iii)).
- Written consent has a defined shape (64.1200(f)(9)): a signed agreement (electronic signatures count) that authorises the seller to deliver artificial-voice or autodialled telemarketing to a stated number, with a clear and conspicuous disclosure that signing authorises those calls and that the person is not required to sign as a condition of buying anything.
- Time and Do Not Call limits apply whoever dials. 64.1200(c) bars telephone solicitations to residential subscribers before 8 a.m. or after 9 p.m. local time and to numbers on the national Do Not Call registry, subject to that paragraph’s own exceptions.
- Every artificial-voice message must identify the business at the start by its registered name, give a callback number during or after the message (for telemarketing and exempt calls to residential lines, one that can take a do-not-call request during regular business hours), and – for telemarketing and the residential exemptions – offer an automated opt-out within two seconds of the identification (64.1200(b)(1)–(3)).
For a dealership that means a declined-work or stale-lead campaign is telemarketing and needs written consent, while a recall reminder with no offer attached is a non-telemarketing call that still needs prior express consent to a mobile number. Our explainer on whether AI cold calling is legal under the TCPA covers the consent rules beyond the dealer context. Build the consent record before you build the outbound campaign; answering inbound calls is not a call you initiate.
What dealer-specific AI vendors say they do
Two dealership-focused vendors illustrate what an AI BDC looks like when it is built around the DMS. Both descriptions are taken from each vendor’s own pages, read on 8 October 2026; we have not tested either product.
- Numa calls itself “The AI Operating System for Car Dealerships”. Its FAQ says its Voice AI “connects to your DMS in real time”, can tell a caller their car is waiting on a part, and can book an open service bay. It says most dealerships go live in 2–3 weeks and “most teams are proficient within 90 days”, and claims 90% DMS coverage across the industry. A separate Numa page, Numa for Sales, says that agent “answers every sales call, day or night”, that “It values the trade” and books test drives, and that when a caller needs a person “The call routes to the right salesperson or desk based on rules you set.”
- Impel describes Voice AI as its “voice solution for dealership service departments”. Its page says the Digital Voice Assistant checks real-time availability in your scheduler and books on the call, pulls the caller’s name, vehicle and recent history from the DMS, and works with phone systems including RingCentral, GoTo, Zoom, Dialpad, Vonage and 8×8. Impel’s Sales AI page describes a separate SMS and email product that “answers complex inquiries on inventory, financing, and trade-ins”.
Impel positions its voice product in the service drive; Numa offers voice agents for both lanes, and its sales agent values the trade on the call – the exact turn the money-or-identity rule hands to a person. That is a store policy decision, not a technical limit: if you let an agent quote trade values, decide in writing who stands behind the number. Service calls are lookups and bookings, the work an AI finishes. On the sales lane an AI agent’s job is the first answer, the qualification and the booked visit; for a general view of that booking workflow, including reminders that protect show rates, see our guide to AI appointment setting. Across the platform, Zian’s agents have set 50,769+ qualified sales appointments.
How to choose: four questions in order
- Where do your missed calls fall? Pull a month of call data split by department and hour. If most unanswered calls are service calls during the day, you have a capacity problem an AI agent solves well. If they are after hours, you have a coverage problem the 4.2-seat arithmetic makes expensive to solve with people.
- Can the agent read and write your DMS and scheduler? Without live repair-order status and real availability, the agent takes messages instead of finishing calls. Ask for a demonstration against your own DMS, not a sandbox.
- Who picks up the handoff? Name the person or queue for every “hand to a human” cell in the threshold table, with a time limit. A hybrid without this is an AI BDC with a voicemail box behind it.
- Do you have consent records for outbound? If not, start inbound-only and add outbound as written consent accumulates through your forms and credit applications.
Ready to put an AI agent on the first answer, with the handoffs drawn where your team wants them? Zian is in partnership-application beta. Apply For Partnership
Frequently asked questions
What is an AI BDC for a car dealership?
An AI BDC is an AI voice and messaging agent that does the first-contact work of a dealership’s business development centre: answering sales and service calls, booking test drives and service appointments, and following up leads. It should hand calls to people when they turn to prices, trade values, credit, safety symptoms or complaints.
What does a BDC agent do at a car dealership?
A BDC agent answers and returns calls, responds to internet leads, books sales and service appointments, confirms them, and follows up shoppers and service customers who have gone quiet. In most stores the sales lane handles leads and test drives, and the service lane handles bookings, status calls and reminders.
Can an AI voice agent book service appointments in my DMS?
Only if it is integrated with your DMS and scheduler. Dealer-specific vendors such as Numa and Impel say on their own pages that their voice products read DMS data and book into the scheduler. A general-purpose platform needs that integration built and tested through its API first.
Do I need consent to use an AI voice agent for outbound dealership calls?
Usually. FCC Declaratory Ruling 24-17 confirms that AI-generated voices are “artificial or prerecorded voice” under the TCPA. Calls to mobiles need prior express consent, and telemarketing such as declined-work offers needs prior express written consent. Only non-telemarketing calls to residential lines get a three-calls-in-30-days allowance.
Does the FTC Safeguards Rule apply to my AI BDC vendor?
If your store is a financial institution under the rule, your AI vendor is a service provider. 16 CFR 314.4(f) requires you to select capable providers, require safeguards by contract and periodically assess them. The rule’s examples include a dealership that leases vehicles for longer than 90 days.
Will an AI BDC replace my BDC team?
Not entirely. An AI agent can take the first answer on every call, around the clock, and finish lookups and bookings. People are still needed for prices, trade values, credit, complaints and every handoff the AI creates. The hybrid model in the comparison table keeps a smaller human team working behind an AI first answer.
Where every figure on this page comes from
| Figure | Who published it | Link | Date read |
|---|---|---|---|
| 16,990 franchised light-vehicle dealers; 276 million+ repair orders (2025) | National Automobile Dealers Association (NADA) | NADA Data | 8 October 2026 |
| 16,252 repair orders and $9,687,942 service and parts sales per average dealership; $494 per customer-pay RO; $551 per warranty RO (2025) | NADA | NADA Data 2025 annual financial profile (PDF) | 8 October 2026 |
| Service and parts 13.3% of total dealership sales dollars (2025) | NADA | NADA Data 2025 annual financial profile (PDF) | 8 October 2026 |
| $1,619 average weekly earnings, new light-vehicle dealership employees, total US (2025) | NADA, from Bureau of Labor Statistics data | NADA Data 2025 annual financial profile (PDF) | 8 October 2026 |
| Dealership leasing example (90 days); vehicle credit arrangement; telephone switching and PBX systems; 314.4 elements | US Federal Trade Commission (16 CFR Part 314) | eCFR, 16 CFR Part 314 | 8 October 2026 |
| FTC notice: 500 consumers, 30 days, effective 13 May 2024; 5,000-consumer exemption; 2-year disposal | US Federal Trade Commission (16 CFR 314.4(j), 314.5, 314.6, 314.4(c)(6)) | eCFR, 16 CFR Part 314; summarised in FTC Safeguards Rule: What Your Business Needs to Know | 8 October 2026 |
| AI voices are “artificial or prerecorded voice”; adopted 2 February 2024, released 8 February 2024 | Federal Communications Commission | FCC 24-17 Declaratory Ruling | 8 October 2026 |
| Three calls in 30 days (residential, non-telemarketing); two-second opt-out; 8 a.m.–9 p.m. solicitation hours; consent definitions | Federal Communications Commission (47 CFR 64.1200) | eCFR, 47 CFR 64.1200 | 8 October 2026 |
| Numa go-live 2–3 weeks; proficient within 90 days; 90% DMS coverage | Numa (NumberAI, Inc.) | numa.com | 8 October 2026 |
| Numa for Sales: answers sales calls, values the trade, routes to a salesperson | Numa (NumberAI, Inc.) | numa.com/sales | 8 October 2026 |
| Impel Sales AI: SMS and email, inventory, financing and trade-in enquiries | Impel | impel.ai/sales-ai | 8 October 2026 |
| Impel Voice AI features and supported phone systems | Impel | impel.ai/voice-ai | 8 October 2026 |
| 1% / 3% / 5% loss rate; 60 staffed hours; 2-attempt and second-anger handoff thresholds | Zian (assumptions, labelled as such) | This page | 8 October 2026 |
| 50,769+ qualified sales appointments set; 30+ languages | Zian AI | Zian AI home page | 8 October 2026 |