Spend first on whatever is actually broken. If leads are not being captured, logged or answered, that is a tool or coverage problem and software comes first. If leads are reaching people who do not ask for the appointment, software will only show you the problem more clearly. Most stores need measurement, then training, then automation.
Our Bias, and Why the Question Is Fair Anyway
Pinnacle sells training. We are not a CRM vendor, and our outsourced BDC is a staffed call center, not an AI product. We do have a partnership with ProMax, a dealership CRM, which you should know before reading our opinion on CRMs. So weigh what follows as the view of people who make a living on the training side.
The question is still a fair one, because the budget is usually one budget. A GM who signs for a new CRM in March rarely signs for training in April. The honest answer depends on where the leads are being lost, and that can be checked before any money is spent.
What Each Purchase Fixes
| Purchase | What it fixes | What it does not fix |
|---|---|---|
| New CRM | Leads scattered across inboxes, no follow-up tasks, no reporting by agent or source | Agents who skip tasks, thin notes, weak calls |
| Call tracking and recording | Not knowing how many calls came in, who answered and what was said | The conversation itself. It records the call, it does not improve it |
| AI responder (text, email, voice) | Slow first response, nights and weekends, simple questions | Trade, payment and judgment conversations, and the handoff to a person |
| Training | Discovery, asking for the appointment, objections, confirmation, CRM habits | Leads that never arrive, a desk nobody staffs, a broken routing rule |
Read the right-hand column twice. Every one of these purchases has a gap, and the gap on the technology rows is the same one: a person still has to do something well.
Tools Expose Skill Gaps Before They Close Them
Turn on call recording and the first thing you learn is what your people say. That is valuable and it is not an improvement yet.
The platforms that listen to dealership calls publish what they hear. Invoca's 2026 automotive benchmarks report says 64% of automotive businesses do not ask leads to buy or book an appointment. That is platform data from Invoca's own customers, which include OEMs and service businesses as well as dealers, and the company does not state the automotive sample size. Still, notice what kind of finding it is. The software found the gap. It could not close it.
The same holds for CRMs. Foureyes' 2026 dealer benchmarks report, drawn from more than 22,900 dealership websites during 2025, says 42.7% of qualified leads were mishandled and 15.2% were never logged to the CRM. It also says 62.8% of sales leads did not hear back from a salesperson within 24 hours of returning to a dealer's site. Foureyes sells lead-tracking software and does not publish a lead count, so treat these as one vendor's platform numbers. Every store in that data has a CRM. The system was there. The habit of logging and following up was not.
A new CRM gives a disciplined team better tools. It gives an undisciplined team a new place to leave tasks overdue. If you are choosing one, our guide to CRM software for car dealerships covers the options, and CRM habits for BDC teams covers the usage side.
What the Independent Data Says About AI
AI responders are the purchase most often pitched as a replacement for training, so the evidence deserves a careful read.
Adoption is wide, results are early
The Cox Automotive AI in Auto Retail Tracker (second quarter 2026, 483 franchise and independent dealers) found 82% of dealers using AI in some form. Coordinating customer follow-up was one of the top uses, at 40%. Among dealers using AI, only 22% reported sales and revenue growth from it so far. That is a survey of what dealers say, taken early in the adoption curve. It does not show AI fails. It shows that buying it and getting paid by it are different events.
AI answers faster, and the weak point is the handoff
Pied Piper's 2026 internet lead study sent a mystery-shop inquiry through 3,290 dealership websites. According to its February 2026 release, dealers answered a typical inquiry within 24 hours 78% of the time, often through automated AI messages. When the customer asked a more complex question that needed a person, that fell to 51%. When the AI needed human help, evaluations scored nine points lower on average and customers were twice as likely to get no personal response.
On the service side, Pied Piper's September 2026 service scheduling study (6,538 service requests) reported that roughly one in three attempted AI-to-human transfers failed.
These are mystery shops, sent in business hours, measuring response and not sales. Within those limits the pattern is consistent. Automation raised the floor on first response. The losses moved to the moment a person had to take over. That moment is a skill: reading the thread, picking up the phone, answering the real question and asking for the visit. The fuller comparison is in AI BDC vs human agents.
An Order of Operations
This is our practitioner's sequence, not a study finding.
- See it first. If you cannot count calls offered and answered, or leads received and worked, you cannot tell a tool problem from a people problem. Basic call tracking and honest CRM logging come before everything, and they are usually already paid for. A quick lead-handling audit shows what you have.
- Fix capture and routing. Leads landing in a personal inbox, a ring group that ends in voicemail, a form that goes nowhere. These are setup problems. No class repairs them.
- Train the people on the tools you have. Once leads reliably reach a person, the remaining gap is what that person does. Reading your BDC numbers shows which rate points to which skill.
- Automate around a team that works. Add AI for first response, nights and overflow once agents know how to pick up a handoff. Then train the handoff specifically.
- Replace the CRM last. Do it when the current one truly cannot do what a disciplined team needs, not because the reports look bad.
When Technology Should Come First
Training is the wrong first purchase in at least three cases.
- Nobody is there. If leads arrive at night and wait until morning, that is a coverage problem. Automation or outside coverage solves it. A better-trained day shift does not.
- You are blind. No recordings and no per-agent numbers means a trainer has nothing to work from, and you will have no way to tell whether the training did anything.
- The system fights the process. If the CRM cannot assign a task or text a customer, agents will work around it, and you will be training them on workarounds.
When the cause is unclear, train, fix or outsource separates a skills problem from a process problem and a capacity problem.
How Training and Tools Fit Together
The two are not rivals. Good training is done on the store's own systems. Pinnacle's Sales BDC Training is taught on the store's own CRM and includes CRM and workflow as a core topic, along with evaluation of live calls and internet responses. The PDS Boot Camp lists AI and automation in daily workflows among its topics. In both cases the point is the same: get people using what the store already pays for.
If you are comparing a software quote with a training quote, put both through the same test. What rate will this change, by how much, and how will we know? The BDC training ROI calculator does that arithmetic for training, and what BDC training costs lists the pricing models. The other articles are in the Sales BDC Training Resource Center.
Frequently Asked Questions
Will a new CRM fix our follow-up problem?
Only if the old CRM was the cause. If tasks were going overdue and notes were thin in the old system, they will be in the new one. Check whether your current CRM can assign, remind and report by agent. If it can, the gap is habit and management, and that is a training and coaching job.
Should we buy an AI responder before we train the BDC?
If your gap is response time at night and on weekends, an AI responder addresses that directly and can come first. If your gap is what happens once a person takes the lead, train first. Either way, plan the handoff from AI to agent, because independent mystery shops found that is where customers get dropped.
Does call tracking count as training?
No. Call tracking and recording give you the raw material: what was said, by whom, and what happened next. Someone still has to listen, score the calls against a standard and coach the agent. Without that step you have recordings and the same conversations as before.
How do we tell a tool problem from a skill problem?
Follow twenty recent leads from arrival to outcome. If they stall before a person touches them (not logged, not routed, sat overnight), it is a tool or coverage problem. If a person reached the customer and no firm appointment was offered, it is a skill problem. Most stores find some of each.
Summary
Technology fixes capture, speed and visibility, and training fixes what people do once they are on the lead. Vendor data and independent mystery shops both point the same way: tools reveal that agents are not logging, following up or asking, and AI's weak point is the handoff to a person. Get visibility, fix routing, train on the systems you already own, then automate. We sell training, so test that advice against your own twenty leads.
