Insights by NEON by Vehicle AI

The Moat Is Your Operating Playbook

William Teslik — Sat, 07 Nov 2026 12:00:00 GMT

Software access is not a durable advantage; every competitor can buy the same tools. The defensible edge is a governed operating playbook that turns dealership judgment into repeatable action, and it is built one recorded override at a time.

Every dealer can buy the same market feeds, model families and reporting tools, so the asset a competitor cannot acquire is the store's record of how it decides and what happened next.
Overrides are the raw material of the playbook; when managers reject a recommendation for the same defensible reason, the rule is missing a condition, and the dealer principal owns turning that reason into a rule.
With average store profit down 3 percent on the year and groups buying more rooftops per deal, the stores that compound their operating record will hold margin that the stores renting a dashboard will not.

Dealer Groups Should Manage Exceptions, Not Averages

William Teslik and Kent Kirkwood — Sat, 31 Oct 2026 12:00:00 GMT

Groups are buying more stores per deal while the average store earns less, which makes the group office's view of each rooftop worth more and the group average worth less. Useful group benchmarks preserve store context, surface material exceptions and give local managers room to decide while the organization learns.

The Haig Report counted dealerships acquired up 14.3 percent in the first half of 2026 against transactions up 3.2 percent, with average store profit down 3 percent to about $1.1 million; more rooftops per group, less margin per rooftop, is the case for managing exceptions.
A group average is a summary, not an explanation; two rooftops with the same aged-inventory share can be one problem and one deliberate position, and only the comparison logic tells them apart.
Standards belong to the group, decisions belong to the rooftop; the regional director sets the measures, thresholds and evidence rules, and the local general manager records the response, including the reasoned no.

The Customer Is One Relationship, Not Four Department Records

William Teslik — Sat, 24 Oct 2026 12:00:00 GMT

Only 54 percent of recent buyers go back to the selling dealer for service, and the ones who do are far more likely to buy there again. Customer lifetime value is built through clean departmental handoffs, consent-aware context and consistent follow-through across the ownership cycle, and one household shows where it leaks.

Cox Automotive found 54 percent of owners of vehicles two years old or newer returning to the selling dealer for service in 2025, down from 72 percent in 2023, while 74 percent of those who service at the dealership say they will buy there next; the handoff from delivery to first service is the most valuable one in the store.
A relationship timeline, verified purchases, service visits, declined work, open promises, preferences, each with a source and a date, beats a lifetime-value score nobody can explain.
Measure continuity before predicting value: handoff completeness, open-loop age and preference accuracy are things the general manager can inspect this week.

Every AI Recommendation Needs an Owner

William Teslik — Sat, 17 Oct 2026 12:00:00 GMT

Human review is meaningful only when the recommendation is traceable and the reviewer has authority. Responsible AI begins with operating ownership, and this governance anchor sets out who owns what.

A human in the loop who cannot see the evidence, cannot change the rule and cannot say no is not oversight, so governance starts with decision rights, not with an approval button.
Borrow the structure of the NIST AI Risk Management Framework, govern, map, measure and manage, and write it down per decision as an operating contract that names the owner.
The regulatory floor has not moved because of AI; the Safeguards Rule, ECOA, FCRA and the FTC Act still apply, and the FTC kept bringing dealer cases under existing law while the CARS Rule rose and fell.

AI Controls Should Reduce Risk Without Freezing the Deal

William Teslik — Sat, 10 Oct 2026 12:00:00 GMT

Auto lending fraud exposure reached a record $10.4 billion in the latest industry count, and the June 2024 CDK outage showed what a security event costs a store by the hour. A useful risk control explains the anomaly, routes it to the right reviewer, and records the decision without treating every unusual deal as fraud.

Point Predictive put auto lending fraud exposure at a record $10.4 billion in its 2026 report, with income and employment misrepresentation accounting for 45 percent of it and growing 21 percent in a year, so the most common risk in the file is a number that does not match a document, not a stolen identity.
An anomaly is a question, not a verdict; every alert states what differed, which records support it, how confident the comparison is, and which approved review step comes next, and the finance director decides.
Identity risk in the deal and a security event in the store are different workflows with different owners; the CDK outage of June 2024 cost affected dealers an estimated $1.02 billion, and it was not an F&I problem.

F&I Speed Without Deal Quality Is False Efficiency

William Teslik — Sat, 03 Oct 2026 12:00:00 GMT

A fast delivery is not a clean deal. The CARS Rule came and went, the Safeguards Rule's breach clock stayed, and the FTC kept settling dealer cases under existing law. F&I improves when missing information, unusual structures and ownership gaps surface at each handoff before funding is delayed.

The rules that govern the deal jacket did not arrive with the CARS Rule and did not leave with it; the FTC's $10 million Napleton settlement in 2022 and $2.6 million Coulter judgment in 2024 were brought under law that was already on the books.
Build an acceptance gate into each handoff, sales to F&I, F&I to accounting, dealership to lender, with the evidence required to pass, the role that accepts, and a queue for the exception that cannot be resolved in place.
Measure clean progress, first-pass completeness, time from exception to owner action, funding delays by root cause, not minutes in the box or vehicles over the curb; the finance director owns the gates.

The EV Service Transition Will Happen Unevenly

William Teslik — Sat, 26 Sep 2026 12:00:00 GMT

National EV share fell from a record 10.5 percent in the third quarter of 2025 to 6.4 percent a year later, while California registrations run at more than twice the national rate. EV service demand will arrive differently in every market, and local vehicle, repair, staffing and customer signals should determine when the operation changes.

The national headline reversed in a year, from a record 10.5 percent EV share in the third quarter of 2025 to 6.4 percent in the third quarter of 2026, which is exactly why a service director should plan from the local vehicle population rather than the forecast.
Regional variance is the whole story: California's zero-emission share of new registrations was 16.4 percent through September 2026 against 6.2 percent nationally, and the vehicles already sold keep arriving in the lane regardless of this quarter's sales.
Readiness is a chain, technicians, procedures, parts, diagnostic information and scheduling, so the fixed-ops director sequences the next investment against the work the shop is already referring away.

Optimize the Work Before You Optimize the People

William Teslik and Kent Kirkwood — Sat, 19 Sep 2026 12:00:00 GMT

Technician and advisor performance is often constrained by the work around them. With the Bureau of Labor Statistics projecting about 66,200 technician openings a year, the store cannot hire its way out of a workflow problem; fix sequencing, context and ownership before adding another scorecard.

The average franchised store runs about a thousand repair orders a year through each of its 16 technicians, and the Bureau of Labor Statistics projects roughly 66,200 technician openings a year through 2035, so protecting ready-to-work time is cheaper than replacing the people who have it.
Two shops with the same headcount produce different hours because one protects ready work and the other lets it dissolve into waiting for parts, authorization and clarification; measure the waiting before measuring the people.
Coach from exceptions, not surveillance: the service director reviews the job that waited after diagnosis and asks what information or decision was missing, not why the technician was slow.

The Service Lane Is a Flow System

William Teslik and Kent Kirkwood — Sat, 12 Sep 2026 12:00:00 GMT

A full schedule can still produce a broken day. The average franchised store writes 16,252 repair orders a year with 16 technicians, and service performance improves when demand, capacity, parts and customer promises are managed as one flow with one owner per exception.

The average franchised dealership wrote 16,252 repair orders in 2025 with 16 technicians, and dealerships as a group handle 12 percent fewer service visits than in 2018, so every hour a repair order spends waiting with no owner is an hour the store cannot get back.
A full calendar describes arrivals, not usable capacity; capacity is the mix of work, skills, bays, parts and promises, and it has to be read together before 7:30.
Manage the lane by exception: the vehicle waiting too long for diagnosis, the approved job with no technician, the promise that no longer matches the estimate, each with an owner and a time.

The BDC Does Not Need More Scorecards

William Teslik and Kent Kirkwood — Sat, 05 Sep 2026 12:00:00 GMT

Most sales calls carry purchase intent, and a good share of callers buy. BDC performance improves when the store agrees on three call outcomes, coaches the conversations that did not advance in a weekly cadence, assigns the next step, and verifies follow-through without creating surveillance.

Marchex found 57 percent of inbound sales calls carrying purchase intent and at least 28 percent of callers going on to buy, so a call that ends without a defined outcome is not a missed metric, it is a missed customer.
Replace the scorecard with three outcome definitions, advanced, answered, stalled, agreed by the BDC director and the general sales manager, and record one of them on every conversation.
Coach in a weekly cadence from a small sample of stalled conversations, each with the moment, the context and a coaching question, and send the ones that stalled on missing data or a broken rule to the person who owns that process.

Your Customers Are Already Explaining the Process Breaks

William Teslik — Sat, 29 Aug 2026 12:00:00 GMT

Forty-five percent of owners say they are dissatisfied with dealership service, and they are saying why in calls, chats, reviews and notes the store already holds. Customer conversations reveal recurring process failures when teams review the evidence, protect privacy, assign ownership and verify the fix quickly.

Cox Automotive found 45 percent of owners dissatisfied with dealership service in 2025 and only 54 percent of recent buyers returning to the selling dealer for service, down from 72 percent two years earlier; the explanations are already in the store's own records.
The useful unit is the process moment, a commitment made, a question repeated, two departments disagreeing, not the sentiment label, and a theme is credible only when a reviewer can inspect representative evidence.
Every confirmed theme moves to an owner with a bounded action and a recurrence check; the general manager reviews the three themes with the most evidence, not the thirty with the most volume.

Equity Is Not Intent

William Teslik — Sat, 22 Aug 2026 12:00:00 GMT

With the average new-car loan near 70 months and a third of loans running past six years, a customer's equity position is a thinner signal than it used to be. Relevance, timing, permission and a clear reason to care determine whether outreach is useful and welcome.

Experian puts the average new-vehicle loan at 69.48 months and 35.55 percent of new loans past 72 months, so the customer with real equity in year three is rarer than the equity-mining list implies.
An equity signal says the customer has options; an intent signal says the customer has a reason, and only the second justifies a conversation the sales manager would be comfortable having in the service lane.
Suppress before you prioritize: an open repair, a dispute, a decline or an opt-out should stop the outreach, whatever the equity calculation says.

Marketing Reports Should End in a Decision

William Teslik — Sat, 15 Aug 2026 12:00:00 GMT

Google Analytics models part of what it reports and attributes the rest with a 90-day window. Useful marketing analysis takes that honestly, then connects spend, shopper intent, inventory, lead handling and sold outcomes so each report supports a decision for managers.

Attribution is evidence, not a verdict; Google Analytics assigns credit with a data-driven model over a 90-day window by default, and under consent mode it estimates the behavior of users who declined cookies from users who look like them.
The useful unit of a marketing report is a business problem, which models are drawing attention but not converting, which aged units get no qualified traffic, not a channel average.
Every finding ends in a bounded action with an owner and a review date, and sold data goes back to the start of the chain so the next decision is better than the last.

A Fast Response That Misses the Question Is Still a Bad Response

William Teslik — Sat, 08 Aug 2026 12:00:00 GMT

Nearly one in five internet leads still gets no personal response, and the ones that do often get an appointment request instead of an answer. Lead response improves when the dealership reads the shopper's question, assigns ownership, and makes the next useful action clear for the customer.

Pied Piper's 2025 study found dealerships personally answering internet leads 69 percent of the time by email or text and leaving 19 percent with no personal response at all, so the floor is still unanswered leads, and the ceiling is answered questions.
Buyers are telling the industry what they want to do from a distance, 37 percent want to settle the price online but only 19 percent do, and the first reply is where that gap opens or closes.
The internet sales manager owns a short table of lead intents and the right first reply for each; speed is measured against that table, not against a stopwatch.

Stop Managing the Month from the Last Five Days

William Teslik — Sat, 01 Aug 2026 12:00:00 GMT

Stair-step decisions improve when sales pace, pipeline quality, inventory gaps and the full cost of intervention are reviewed while options remain open. A fictional store's day-20 review shows what that costs and what it saves.

Straight-line pace is a baseline, not a forecast; the weekly briefing should carry a baseline, an attainable case and a higher case, each tied to named deals and named units.
With incentives at 6.6 percent of the average transaction price in September 2026, the economics of chasing a volume threshold have to be priced against gross concessions, dealer-trade cost and the units pulled forward from next month.
The general sales manager owns the gap by day 15, which means every inventory gap and every stalled qualified deal has an owner and a decision deadline before the final five days.

Price-to-Market Is a Signal, Not a Strategy

William Teslik — Sat, 25 Jul 2026 12:00:00 GMT

Repricing every unit to one market ratio gives away margin and hides merchandising problems. A worked six-unit review shows how P2M, demand, supply and a recorded override reason belong in the same decision.

Price-to-market is an estimate built from an imperfect comparison set; show the comps, their freshness and the confidence before anyone acts on the ratio.
Price the exceptions where signals disagree, high P2M with strong demand, low P2M with no demand, not the storewide average, and put a decision date on each.
An override is operating information, so the used-car manager records one of a short list of reasons, and the store learns which reasons were right.

Aging Starts Before Day 60

William Teslik — Sat, 18 Jul 2026 12:00:00 GMT

Inventory risk appears before a unit reaches an age threshold. With average used gross near $1,600 a unit, the first price move has to happen while it is still a small one. The useful signal is an evidence-backed exception with enough time left for a manager to act.

Days on lot describes how long the store has owned a unit; the signals that predict a bad outcome, VDP views trailing the cohort, leads that ask only about price, supply rising faster than it sells, appear weeks earlier.
With the Haig Report putting average used gross per unit at $1,634 in the second quarter of 2026, a $900 price move at day 60 gives away more than half the gross, so the used-car manager owns the call at day 20, not day 60.
Every early exception needs a decision packet, what changed, which signals, which cohort, what is missing, what actions remain, and a named owner with a review date, including the decision to hold.

The Buy Is the First Inventory Decision

William Teslik and Kent Kirkwood — Sat, 11 Jul 2026 12:00:00 GMT

NADA Data 2025 puts 69 percent of the used vehicles franchised dealers retail through the trade lane. That makes appraisal discipline the biggest inventory lever a store has, and the desk manager's over-allowance the place where used gross is won or lost.

Trade-ins on new and used deals supplied 45.4 and 23.6 percent of used units retailed by franchised dealers in 2025, so the appraisal, not the auction, is where most inventory risk is accepted.
Build the number backward from a retail-ready unit, recon range, time to line, entry price, demand and exit, and record the reason whenever the desk stretches past it.
Review over-allowances weekly as a pattern, not a deal at a time; one trade cannot grade an appraiser, but twelve over-allowances on the same model line in a month can.

Your Data Does Not Need to Be Perfect Before AI Can Help

William Teslik — Sat, 04 Jul 2026 12:00:00 GMT

AI readiness begins with a traceable decision, shared definitions and accountable source data, not a multiyear promise to clean every field in the store. This guide lays out the 30 days of work, week by week, with the Safeguards Rule's data inventory and the NIST framework's mapping step doing double duty.

Perfect data is not an operating standard; traceable data is, and the test is whether a manager can get from a recommendation back to the deal, the unit, the customer or the repair order that produced it.
The Safeguards Rule already requires a dealership to know what customer information it holds and where, so the data inventory the controller builds for compliance is the same one the first AI workflow needs.
Thirty days is enough to define one decision, write its minimum viable data contract, name the owner of each source and expose uncertainty instead of hiding it; the general manager owns the decision, the controller owns the contract.

The Dealership Does Not Need Another Dashboard

William Teslik — Sat, 27 Jun 2026 12:00:00 GMT

Dashboards explain what happened. Dealerships need a disciplined way to surface material exceptions, assign decisions, and verify follow-through. This anchor piece sets out the five-part exception model the rest of the series builds on.

The average franchised store is a $77 million business run by 65 people across a dozen systems, and none of those systems answers the question that matters at 8 a.m., which is what needs a decision today.
A reviewable exception has five parts, a defined condition, supporting evidence, decision context, a named owner, and a closed loop, and a notification that lacks any of them is noise.
Start with one recurring decision that already has an owner and a known cost of delay, make that loop reliable, and extend the same discipline to the next workflow.