By Sagar Shankaran, Founder of CallSphere
340 scale tickets, eleven missing purchase orders and a lien deadline nobody tracked. What an overnight run hands the office manager before the gate opens at 6 a.m.
Key takeaways
Ask it that way and you get an honest answer, because every aggregates office has the same tray. Yesterday's scale tickets, three hundred and forty of them. Eleven with no purchase order or job number in the box. Four where the material code does not match the sales order, because the loader operator loaded base and the ticket says surge. Two loads out to a customer past its limit. Broker haul tickets to settle per ton per zone, half of which the drivers will dispute. The landowner royalty off shipped tons on the leased parcel. And under all that, a new job for a general contractor she has never dealt with, on which a notice clock started the day the first truck rolled.
An overnight run is an assistant that starts after the scale house closes, works unattended through the night on yesterday's tickets and paperwork, and leaves a reviewed queue of exceptions on a desk before the gate opens — not a list of chores, but a short page of decisions already teed up. That is new as of 2026: not a robot that answers a question, but one that works for hours on its own and hands back finished work.
Aggregate producers are material suppliers, and material suppliers get paid last. Every state gives you lien rights or a bond claim, and conditions them on a notice sent within a window that starts when you first delivered. California wants a preliminary notice inside twenty days. Florida wants a Notice to Owner within forty-five days of first furnishing. Texas runs on a monthly clock counted from the month you delivered. On federally funded work the Miller Act gives a supplier who did not contract with the prime ninety days from last delivery. Your state's rule belongs to your attorney, but the shape is the same: a calendar starts the day the first truck crosses your scale, and nobody watches it.
Here is the scene that actually happens. A paving contractor opens an account in June, runs 12,600 tons of base and #57 through the summer on a $186,000 job, and files in October. You had lien rights for about three weeks, never sent the notice, and are now an unsecured creditor standing behind a bank. The office manager did not fail. She had three hundred and forty tickets to get out and a scale house asking whether a truck could load.
The workaround is a spreadsheet of new accounts somebody updates when they remember, plus a service like Levelset that sends notices beautifully — if a person feeds it the job address, the owner, the general contractor and the first-furnishing date. That is the whole problem. The information is on the ticket and the credit application; what is missing is a person with an hour to move it across. Same with the rest of the tray: the fix for each exception already sits in the ticketing system, the sales order, the credit file or an emailed purchase order. Not missing, just scattered — and reconciling scattered things is the rule-following work that used to require a person and now does not.
flowchart TD
A["Scale house closes at 5:30 p.m."] --> B["Overnight run pulls the day's tickets"]
B --> C["Match each ticket to sales order, PO and job number"]
C --> D{"Does everything line up?"}
D -->|Yes| E["Invoice drafted and queued to send at 7 a.m."]
D -->|No| F["Exception written in plain English with the ticket number"]
F --> G["Notice deadlines and credit holds added to the same list"]
G --> H["6 a.m. review queue on the office manager's desk"]
The 2024 version answered when spoken to: you asked, it replied, and forty invoices meant asking forty times. What arrived this year differs in the one respect a back office cares about — it takes a goal, works on its own for hours across your files and systems, and returns finished work. Claude Cowork landed in January 2026 aimed at people who are not technical; ChatGPT Work followed on 9 July 2026 doing the same — take a goal, connect to the apps and files, run unattended, hand back a completed spreadsheet or document.
Hear it before you finish reading
Talk to a live CallSphere AI voice agent in your browser — 60 seconds, no signup.
The other half is cost, down roughly ten times from 2025, which is why checking every ticket every night stopped being an extravagance. Nobody would pay a person to verify three hundred and forty tickets overnight, and until recently nobody paid a machine either.
Pointed at your ticketing system — Command Alkon Apex, Libra, whatever the scale house prints from — and your accounting side in Sage 300, Vista or Foundation, the run is doing clerical matching. It is not deciding anything, and that is what makes it safe.
The office manager sits down at 6:05 to a single page. Three hundred and twenty-nine tickets matched cleanly and are drafted as invoices, queued but not sent. Eleven exceptions, each written like a person wrote it: "Ticket 88214, Kowalski Paving, no PO on ticket; their emailed order 4471 covers base for the Route 9 job and the truck number matches — attach?" She clicks yes eight times. Two are genuine: a ticket under a customer whose highway-work exemption certificate expired in May, and a load billed as washed #8 that the plant note says came off the dry side.
Below that: two draft preliminary notices for new job accounts, addresses and first-furnishing dates already filled from the scale tickets, waiting to be checked and sent certified. A credit-hold list for the scale house — three accounts, print it and walk it out before the gate opens. The broker haul settlement with two rate disputes flagged where the ticket zone disagrees with the rate table. And shipped tons by leased parcel feeding the landowner royalty, so quarter-end is not a two-day project.
Every line on that page is a decision, and each takes ten seconds because the work behind it is done. The morning starts with judgment, not data entry.
Take a pit doing $22 million a year at 58 days sales outstanding. Invoices go out two to three days after delivery because of exception chasing; with the queue cleared overnight they go the next morning, and cleaner invoices get disputed less.
| Assumption | Value |
| Annual sales | $22,000,000 |
| Sales per day | $60,274 |
| Days sales outstanding, before | 58 |
| Days sales outstanding, after | 54 |
| Cash released, one time | about $241,000 |
| Interest on the revolving line at 9.25% | about $22,300 a year |
| Notices sent on time that previously slipped | one $186,000 exposure a year at a 25% chance of loss = $46,500 expected |
Four days of DSO is modest and achievable, and it is worth more than the software. The notice line is an illustration, not a promise — some years you lose nothing, some years a contractor files in October. Prove it by pulling your own aging report at ninety days before you start and ninety days after.
It should never send anything to a customer on its own. Notices, invoices and collection letters get a human signature, because a preliminary notice sent to the wrong owner on the wrong parcel is a legal document with your name on it. Draft, do not send.
Still reading? Stop comparing — try CallSphere live.
CallSphere ships complete AI voice agents per industry — 14 tools for healthcare, 10 agents for real estate, 4 specialists for salons. See how it actually handles a call before you book a demo.
It should never post to the general ledger and never release a credit hold. A contractor over his limit with a check in the mail is a phone call between your credit manager and his controller — a decision with fifteen years of history behind it that no assistant can see.
It will misread handwriting. Scale tickets get written on in pencil, in a truck, in the rain, and a smudged job number comes out either wrong or flagged. Flagged is fine, wrong is not, which is why every exception has to carry the ticket number so a person can pull the paper. And keep it away from sales tax exemption decisions on public work: the certificate rules vary by state, they change, and the penalty lands on you at audit.
Yes, and that is the common case at pits. The overnight run reads exports — the ticket file, the open order report, the aging — and writes a spreadsheet and a page of exceptions. It does not need the two systems joined to compare them. Joining them properly is a better long-term project, not the price of entry.
Nothing until the paper arrives. If a yard batches tickets weekly, the run finds those exceptions a week late. The fix is not the assistant, it is getting that yard printing to the same system — and the exception list is a good way to prove what the delay costs.
No. Everything here is reading tickets and orders and writing drafts. Payment authority, bank access and posting rights stay where they are. Give it read access and somewhere to write a document, nothing more.
The opposite happens first. For two weeks the list is longer than anyone expected, because it surfaces errors that used to be absorbed silently — wrong material codes, expired certificates, missing purchase orders. That is the value, not a defect. It gets short once you fix the habits it exposes.
Pick one night this week. Export yesterday's ticket file and your open sales orders and ask for one thing: every ticket that cannot be matched to an order with a purchase order and job number, with the reason in plain English. Read it at six the next morning. If it found only what your office would have caught anyway, you lost an evening. If it found four you would not have caught until October, you know what to do next.
One last note on the daytime side. Invoices going out sooner means the phone rings sooner — a foreman querying a ticket, a controller asking for a copy, a new customer wanting an account. CallSphere builds the voice and chat agents that pick up those calls on the first ring, take the details, and route the ones that need your credit manager to your credit manager. The night run sorts the paper; the agent makes sure nobody hits voicemail asking about it.
Written by
Sagar Shankaran· Founder, CallSphere
LinkedInSagar Shankaran is the founder of CallSphere, where he builds production AI voice and chat agents deployed across healthcare, hospitality, real estate, and home services. He writes about agentic AI, LLM engineering, and shipping voice agents that handle real calls in production.
See how AI voice agents work for your industry. Live demo available -- no signup required.
Carbon-copy scale tickets, margin notes and container photos still get keyed by hand. What 2026 document readers catch, and the unbilled extras they surface.
Idle crew days and wrong merchandising cost real money. How 2026 demand models read gappy scale-ticket history and call mill quota shifts weeks ahead.
Payouts, oversells, returns aging and chargeback deadlines, prepared between 9pm and 6am. What a DTC ops manager reviews instead of rebuilds each morning.
Shorts, spoils and billback deductions researched unattended overnight, with PODs and credit memos attached, so the controller approves a queue at 7 AM.
Card batch, fourteen invoices, missed punches and delivery payouts, reconciled between last call and open. Here is what a restaurant morning looks like after.
DoorDash adjustments, short cases and off-contract invoice lines quietly cost six stores real money. What a 2026 overnight agent has finished before 6 am.
© 2026 CallSphere Inc. All rights reserved.
Made within San Francisco
Watch how CallSphere handles real customer calls, schedules appointments, and processes payments — live.
Try Live DemoBook a DemoCalculate Your ROI