Contracts in Transit Hit 14 Days at Your Store Last Month. The Overnight Agent Has the Stip List by 7:45 a.m.

It is 6:10 on a Tuesday evening in late September. The last delivery of the day just rolled off the front line with a temp tag in the glovebox. The F&I manager drops the deal jacket in the tray outside the business office, turns off the light, and goes home. Nothing else happens to that deal until roughly 8:20 the next morning, when the funding clerk gets through her first cup of coffee and starts working the pile from the bottom.

That gap — delivery at 6:10 p.m., first human touch at 8:20 a.m. — is where the money leaks. Quietly, a few dollars a day per unit, on a schedule your controller argues about once a quarter.

The 6:10 p.m. delivery that nobody funds until Thursday

Contracts in transit is dealer language for a car you have already handed to a customer whose lender has not yet paid you: the vehicle is gone, the money is not in, and the flooring note on that unit is still running. Every franchise store carries a contracts-in-transit schedule in the DMS, and every controller in the country knows the feeling of watching the average age on that schedule creep from five days to nine to fourteen during a heavy month.

The reasons are never exotic. A retail installment contract went out without the co-buyer's initials on the arbitration page. The proof of insurance never came. Income verification was a pay stub older than 30 days. The buyer's name on the contract reads "Robert" and the title application reads "Bob." The lender's funding desk flags one of those, sends a kickback notice into RouteOne or Dealertrack, and the deal sits — because the notice landed in a shared inbox at 4:50 p.m. and nobody opened it until the next business day.

Meanwhile the unit is still on your floorplan line with Ally, GM Financial, Chase or NextGear, still accruing interest, and if it drifts past curtailment you are writing a principal check on a car you no longer own. During model-year changeover the schedule balloons exactly when your cash is thinnest.

What the funding clerk actually does between 8:00 and 11:30

Watch her for a morning. She opens the DMS schedule. She opens the lender portals — usually three or four, because a captive, a credit union, a bank and a subprime source all have different screens. She cross-checks each open deal against a funding status, writes on a legal pad, then starts calling and texting: the customer for the insurance binder, the sales manager for the missing signature, the lender's funding analyst for the reason code nobody spelled out.

By 11:30 she has worked maybe 18 of the 34 open deals, and the ones she has not reached are aging the fastest, because she works the schedule top-down and the oldest deals are the ugliest. Nobody is doing anything wrong. There simply is not enough of her.

The workaround every store pretends is fine: the title clerk covers funding when the funding clerk is out, the office manager "spot checks" the schedule on the 25th, and the controller does a real reconciliation at month end, when the factory financial statement is due and it is far too late to change the number.

What changed in 2026: the agent works the whole night, not the whole minute

The 2024 version of this idea was a chatbot that answered a question and stopped. The 2026 version does not stop. Claude Cowork, which launched on 12 January 2026 and expanded to web and mobile in July, and ChatGPT Work, which launched 9 July 2026, both take a goal instead of a command and then run for hours across your files and your applications, unattended, returning finished work rather than a suggestion. Anthropic's Agent Teams preview goes further and splits one job across several agents working different parts at the same time.

For a dealership back office, the practical translation is this: the work that has to happen between close of business and open — the matching, the chasing, the drafting, the filing — can happen while the building is dark. Your funding clerk stops arriving to a to-do list and starts arriving to a reviewed queue with the easy 70% already cleared and the hard 30% already explained.

flowchart TD
  A["Deal delivered 6:10 p.m."] --> B["Agent opens the contracts-in-transit schedule at 11:00 p.m."]
  B --> C["Match every open deal to the lender's funding status"]
  C --> D{"Funded, or missing a stip?"}
  D -->|Funded| E["Mark the schedule line and note the deposit date"]
  D -->|Missing stip| F["Draft the customer text and the note to the funding analyst"]
  F --> G["Exception queue on the funding clerk's desk, 7:45 a.m."]
  E --> G

Tuesday, 7:45 a.m., in the business office

She sits down. On her screen is one list, sorted by days outstanding rather than stock number. Thirty-four deals were open at close. Nineteen show a funding confirmation posted overnight, already matched to the bank deposit — she spot-checks four and clears the rest. Eleven have a named reason: three need a current pay stub, two need the insurance binder, four need a signature page re-signed, one has a name mismatch against the title application, one is waiting on a secure power of attorney the customer never returned.

For each of those eleven the agent has already drafted the outbound text to the customer in your store's voice, already drafted the note to the lender's funding analyst quoting the deal number and the exact page reference, and already flagged which of your own people has to fix it — the sales manager for signatures, the title clerk for the name mismatch. She reads, edits two of them, and sends. That is twenty minutes, not three hours.

Four deals are marked "cannot resolve without a person": a lender changed a stipulation code the agent has not seen, and two customers have gone quiet for nine days. Those four are where she spends her morning — which is exactly where a good funding clerk should spend it.

The arithmetic on one week of contracts in transit

These are illustrative assumptions for a mid-size domestic franchise store, not a case study. Put your own numbers in the same boxes.

Add the deals that blow past a lender's 30-day contract deadline and get re-contracted at a worse rate or unwound. Two a month at roughly $600 of lost reserve and rework each is another $1,200 — call it $5,000 a month of found money at a 110-car store.

How you prove it: run the average-age report on the contracts-in-transit schedule for the three months before you start and the three months after. Nothing else. If the average age does not move, the agent is not working and you should say so out loud.

Where this still needs your controller and your title clerk

Do not let an overnight agent post entries to the general ledger on its own. Matching a funding confirmation to a schedule line is a proposal; the person with the controller's login accepts it. Every store that has watched a schedule quietly forced into balance by someone trying to be helpful knows why this rule exists.

Do not let it send anything to a lender under your dealer code without a human pressing send. Funding analysts have long memories, and a wave of machine-written notes with a wrong deal number costs you more goodwill than the hours are worth.

Do not point it at title work first. Registration rules are state-specific and deadlines are statutory. Let the agent flag the title packet; let the title clerk file it.

And be honest about the failure it will not fix: if your F&I managers deliver deals with missing signatures because they are pushed to spot-deliver on the 30th, an agent will document that pattern in high resolution every single night. Some owners find that report more uncomfortable than useful. It is still the most valuable thing in the queue.

Start Monday with one schedule

Pick the contracts-in-transit schedule. Nothing else. Give the agent read access to the schedule in CDK Drive, Reynolds ERA-IGNITE, Dealertrack DMS or Tekion, plus your lender portals, and have it produce exactly one thing each morning: a sorted exception list with a drafted next action per deal. No posting, no sending. Run it two weeks with your funding clerk marking each line right or wrong, and you will know by the second Friday whether it earns the seat. Then, and only then, extend the same overnight pattern to the warranty claim rejections sitting in your OEM portal and the we-owe schedule nobody loves.

Frequently asked questions

Does this work if my DMS vendor will not open the data up?

Mostly, yes, but slower. If you cannot get a direct read, the agent can work from the scheduled reports your DMS already emails or drops nightly — the contracts-in-transit detail, the deal log, the deposit journal. Do not let a data-access fight stall this for a quarter.

Will my funding clerk think I am replacing her?

She will think it until the second week, when she notices she is no longer working the schedule top-down at 8 a.m. Frame it as it is: the agent does the matching and the first draft, she does the judgment and the relationships. Most stores find the clerk ends up covering title backup and lender relations instead — work that was always getting squeezed.

What does it cost to run overnight?

Frontier AI pricing fell roughly tenfold between 2025 and 2026, and a nightly pass over a hundred-odd open deals is small work. The real cost is two or three weeks of your controller's attention while you agree on what "resolved" means for each lender. Budget the attention, not the software.

What about month end, when the schedule doubles?

Month end is the argument for doing this, not against it. The overnight run barely changes between 34 open deals and 90; your clerk's morning changes enormously.

One last note from us. Chasing a stip means a phone call, and those calls run both ways — the customer calling back at 7:40 p.m. about the insurance binder, or asking where the plates are. CallSphere builds AI voice and chat agents that answer the line, take the message with the right deal attached, and book the follow-up, so the after-hours half of the chase does not land back on your funding clerk's desk in the morning.