AI in the BDC: What AI Agents Should Handle and What People Should

Every vendor at every conference now sells an “AI BDC.” Some of it is real. Some of it is an auto-responder with a new name. And the GM signing the contract gets a demo, not an answer to the question that matters: which jobs should the machine do, and which should stay with people?

The problem AI is trying to fix is real. Call-tracking data shows 28% of inbound dealership calls go unanswered, and consumer research on missed calls says 85% of those callers won’t call back. A 2026 study of dealer response times found only 13.2% of dealers respond to leads within five minutes.

The risk is just as real. A 2026 quarterly industry tracker, surveying roughly 500 dealers and 1,500 shoppers each quarter, found 82% of dealers already use AI in some form, but only 22% of those users report sales or revenue growth from it so far, and about one in three either don’t measure its impact or aren’t sure how. A 2025 study of 537 franchise dealership leaders found 74% worry about AI accuracy and errors.

So here is the split, task by task, for the GM and the BDC manager. No product pitch.

What an AI agent in the BDC actually does

Strip away the marketing and an AI agent is software that holds a real conversation by phone, text or chat. It understands the question instead of forcing a button menu, checks your inventory and calendar, follows the rules you give it, writes notes into your CRM, and hands the customer to a person when it should.

What it is not: a salesperson. It doesn’t read a spouse’s body language or desk a deal. Think of it as a BDC hire who never gets tired, is excellent at the first five minutes of a conversation, and should never be left alone with the hard parts.

In most stores it can touch:

• Inbound sales and service calls, especially after hours and on overflow
• First response to internet and chat leads
• Appointment setting, confirmations and reschedules
• Service scheduling and “is my car ready?” calls
• Follow-up with opted-in leads who went quiet

Jobs AI does better: speed, after-hours, volume, consistency

Software beats a person here, not because your people are bad, but because the work is repetitive and the clock matters.

• Speed. It answers on the first ring and replies to a web lead in seconds. Industry research says 78% of buyers purchase from the first dealer to call them back.
• After hours. Nights, Sundays, holidays. When the building is dark, it can answer questions and book a time for the morning.
• Volume. Saturday at 11 a.m., every rep is with a customer and the phones light up. Call-tracking data shows the average dealer connects with 65% of callers while top performers reach 80 to 85%. Overflow coverage is how you close that gap.
• Consistency. Same qualifying questions every time, every confirmation sent, every note logged, and never a bad Monday.
• The long tail. Following up with hundreds of opted-in old leads is work reps won’t do consistently. The same goes for conquest buyers who raise a hand through your marketing, where fast first contact matters more than charm.

Jobs people still do better

The list vendors skip. Keep these with people:

• Money and judgment. Trade values, payments, rates, approvals, “what’s your best price.” An AI can share your published price. It should never quote a payment, promise an approval or negotiate. Those belong to your desk and finance office and carry federal and state disclosure rules.
• Upset customers. A complaint, a bad service visit, a billing dispute. Straight to a manager.
• The real conversation. Finding out why they’re really shopping, handling “I need to talk to my wife,” building trust with a nervous first-time buyer.
• Relationships. A repeat customer who wants their salesperson should get that person.
• Judgment calls. Bending a rule for a loyal customer, saving a deal that’s going sideways.
• Anything legally sensitive. Consent disputes, anyone who mentions a lawyer or a regulator.

Write the split down for your team in three lines:

1. AI owns: after-hours answering, first response, appointment setting and confirmations, service scheduling, status calls, opted-in follow-up.
2. AI starts, a person finishes: hot buyers, trade questions, credit questions, anything about price beyond the published number.
3. People own: negotiation, finance, complaints, repeat customers and every decision that needs judgment.

Designing the handoff from AI to a salesperson

This is where most rollouts fail. The AI books the appointment, nobody reads the note, and the customer meets a salesperson who has never heard of them.

Handoff triggers. Write them down and make them non-negotiable:

• The customer asks for a person. Always, immediately.
• Questions about payment, trade value, credit, or price beyond what’s published.
• Frustration, repeated questions, or an answer the AI isn’t sure of.
• High intent: wants to come in today, asks about paperwork.
• A repeat customer with an assigned salesperson.

Speed rules. During business hours, a warm transfer or a callback within five minutes. After hours, the AI sets the expectation (“a salesperson will call you at 9 a.m.”) and a named rep owns that call at 9:00, not 11:30.

The handoff note. Every handoff lands with the customer’s name, vehicle, trade, timeline, what they asked, what the AI told them and the next step. The rep reads it before dialing, so the customer never repeats themselves:

“Hi Maria, it’s [your name] at [store]. Our assistant mentioned you’re looking at the Highlander and want to come in Thursday after five. I’ll have it pulled up for you. Is your husband coming too?”

Ownership. Decide before launch who gets AI-set appointments: the up system, a rotation, or the assigned salesperson for repeat customers. If you don’t, reps will fight over the good ones and ignore the rest.

The mistake that kills it: appointments nobody owns. An AI-set appointment without a named person attached is how your appointment show rate falls apart.

What to measure in the first 90 days

Get 30 days of baseline numbers before you turn anything on, or you’ll be comparing a vendor dashboard to your gut.

• Answer rate on sales and service calls, by hour of the day.
• Speed to first response on web leads, by hour.
• Appointment set rate and show rate, AI-set against people-set.
• Sold units and closed repair orders from AI-set appointments. Not “conversations handled.”
• Handoff time, from the alert to a person actually talking to the customer.
• Opt-outs and complaints. A spike means the tone or cadence is off.
• Accuracy. Listen to 20 calls and read 20 text threads every week, and log every wrong answer.

These belong next to your other sales manager KPIs. A sane rollout looks like this:

1. Days 1 to 30: after hours and overflow only. Review everything.
2. Days 31 to 60: add first response on web leads and appointment confirmations.
3. Days 61 to 90: decide on sold units, show rate and complaints. If the numbers aren’t there, change the setup or walk away.

Customer trust, consent and disclosure

This part isn’t optional, and none of it is legal advice. Run your setup past your attorney.

• Say it’s an AI. Have the agent introduce itself as an AI assistant at the start. Several states, including Maine and Utah, have rules against bots passing as humans, and more are coming. Customers forgive a bot that says it’s a bot. They don’t forgive one that pretended.
• Consent for outbound. The FCC treats AI-generated voices as artificial voices under federal telemarketing law, so outbound AI calls need the customer’s prior consent, and marketing calls need written consent. Texts go only to people who opted in.
• No purchased lists. Purchased or modeled data comes with no consent to call or text. Never let an AI dial it.
• Honor opt-outs in every channel. STOP by text, “don’t call me” on the phone, a reply by email. Best practice is to stop all marketing contact from the store.
• Recording. Some states, Pennsylvania among them, require everyone on a call to agree to recording. If the AI records, it should say so.
• Your data. Know where recordings and transcripts live, who can see them, and whether the vendor uses them to train anything.

Shoppers are already there: the same 2026 tracker found 63% plan to use AI for their next vehicle purchase. They’ll talk to your AI. They expect it to be honest.

Questions to ask before you buy

1. Does it say it’s an AI at the start of every conversation?
2. Can we write our own handoff rules, and can a customer reach a person at any time?
3. How does it know our inventory and prices, and how often does that update?
4. What does it do when it doesn’t know an answer? The right answer: it says so and hands off.
5. Will it ever call or text someone without documented consent? The only right answer is no.
6. How does it handle STOP and the Do Not Call list?
7. Can we hear every call and read every text, with notes in our CRM?
8. Can we start with after hours only, on a short term?
9. Who owns the recordings and customer data?
10. Can we call a store our size that has been live for six months?

Then ask yourself one more: who in the building owns this every day? AI without a manager watching it drifts, the same as a BDC without one. For one example of how a provider draws these lines, our page on AI employees for sales spells out what the AI handles and what it hands to people.

Quick checklist

• 30 days of baseline numbers before launch.
• A written split: AI owns, AI starts and people finish, people own.
• Handoff triggers plus a five-minute callback rule during business hours.
• Every handoff has a note and a named owner.
• The AI says it’s an AI. Outbound only with consent. No purchased lists.
• A weekly audit of 20 calls and 20 text threads.
• Judge it on sold units and show rate, not conversations handled.

Frequently asked questions

What is an AI BDC?

An AI BDC uses AI agents to handle part of the business development center’s work: answering calls, responding to leads, setting and confirming appointments, and following up with opted-in customers. The setups that work pair the AI with people who own the sale, the numbers and anything that needs judgment.

Will AI replace my BDC reps?

It changes the job more than it removes it. AI takes the repetitive first-contact work, and people take the conversations that need judgment, trust and negotiation.

Is it legal for an AI agent to call customers?

Answering inbound calls is the lower-risk use, though disclosure and recording rules still apply. Outbound AI calls fall under federal rules for artificial voices, which require the customer’s prior consent and written consent for marketing, and states add their own rules. Have your attorney approve it before any outbound AI call.

How long before we know if an AI agent is working?

Give it 90 days against a real baseline. By then answer rate, response time and show rate should have moved, and you should be able to count sold units from AI-set appointments. If you can’t count it, it isn’t working.

What to do tomorrow morning

Pull last month’s answer rate by hour and your after-hours lead count. That tells you where the gap is, and whether AI is the fix or just a cheaper way to avoid hiring the right person.

Then talk to your team before any vendor does. Share will AI replace car salespeople at your next meeting.