By Sagar Shankaran, Founder of CallSphere
Household goods claims carry 30-day and 120-day federal clocks. How parallel agents assemble the evidence packets while your coordinator keeps every decision.
Key takeaways
"We tried this in 2024. It wrote a settlement letter that got the valuation backwards, my claims coordinator spent an hour fixing it, and I have not touched it since." That is a fair objection and I have heard it from four owners this year. Here is what is actually different in 2026, and why it changes a part of your business that has nothing to do with writing letters.
The change is not that the writing got better, although it did. The change is that several agents can now split one large job and work at the same time, then hand back merged results. Agent Teams arrived as a research preview alongside Claude Opus 4.6, and the same idea has spread across the products your staff already use. For most businesses this is a curiosity. For a moving company in October, it lands on the single most stubbornly serial job on your org chart.
Your claims do not arrive when the damage happens. They arrive thirty to seventy days later, which means the damage from your June and July moves lands on the claims desk in September and October — the same weeks your billing clerk is chasing the summer's receivables and your coordinators are finally taking vacation.
And you have one claims person. Maybe one and a half. She works claims the only way claims can be worked: one file at a time, start to finish, because each file is an assembly job. For a single damaged dining table she has to pull the descriptive inventory and find the item number, read the exception codes the driver wrote at origin, check whether the destination rider notes it, find the crew's origin and destination photographs, confirm whether the customer elected Full Value Protection or signed down to Released Value at sixty cents per pound per article, check whether the carton was packed by the owner, get a repair estimate from your furniture vendor, and only then write anything.
Forty minutes if everything is where it should be. Two hours if the photos are on a crew leader's phone and the destination agent has not sent the rider. Forty open files in October is two solid weeks of one person's life, and the whole time new ones arrive.
This matters more than it used to because the clocks are federal and they do not care about your peak season. Under the household goods claims rules, you must acknowledge a written claim within thirty days of receiving it, and within one hundred twenty days you must pay it, decline it in writing with your reasons, or make a settlement offer. The customer, for her part, generally has nine months from delivery to file. And for interstate moves you are required to make a neutral arbitration program available if you and the shipper cannot agree.
Hear it before you finish reading
Talk to a live CallSphere AI voice agent for home services in your browser — 60 seconds, no signup.
Miss the thirty-day acknowledgement and you have handed the shipper's attorney the opening paragraph of the complaint. Drift past a hundred and twenty days and the same thing happens with a regulator, or with the van line whose agent quality scorecard your load allocation depends on. The plain version: your claims desk is not a customer service function, it is a compliance deadline with a customer attached.
flowchart TD
A["40 open claims sitting in the October queue"] --> B["Split the batch, one file per agent"]
B --> C["Agent 1: pulls inventory, rider, exception codes"]
B --> D["Agent 2: matches origin and destination photos"]
B --> E["Agent 3: confirms valuation election and PBO status"]
B --> F["Agent 4: drafts the acknowledgement letter"]
C --> G["Merged claim file, one per claim, evidence attached"]
D --> G
E --> G
F --> G
G --> H["Claims coordinator decides: pay, offer, or decline"]
What this looks like in practice on a Wednesday in October. Your claims coordinator selects the forty open files. Instead of opening the first one, she gives one instruction: build the evidence packet for all forty, flag anything missing, and draft the thirty-day acknowledgement for the eleven that have not had one.
They work in parallel. By the time she comes back from lunch, thirty-one files are assembled: item number located on the inventory, origin exception codes transcribed, destination rider matched or marked absent, photographs pulled and paired, valuation election identified with the signed page attached, packed-by-owner status noted. Nine files come back incomplete with a specific reason each — "no destination rider received from the destination agent, requested 12 October," "crew photographs for job 8841 not uploaded, crew leader R. Alvarez."
That is the whole trick. She has not lost a single decision. What she has lost is the assembly, which was never the skilled part, and what she has gained is nine phone calls she now knows to make on Wednesday afternoon instead of discovering it on 3 November when the file finally reaches the top of the pile.
Assumptions: 3,100 jobs a year, a claims incidence of 4.5%, so about 140 claims. Sixty percent of them land between September and November. One claims coordinator at 30 hours a week on claims work. Assembly time averaging 55 minutes a file; decision, negotiation and correspondence averaging 25 minutes.
| Measure | Serial, one at a time | Batch assembled in parallel |
|---|---|---|
| Autumn claims (Sep–Nov) | 84 | 84 |
| Hours of assembly | 77 | 14 (checking the packets) |
| Hours of judgement and correspondence | 35 | 35 |
| Total coordinator hours | 112 | 49 |
| Average days to acknowledge | 19 | 4 |
| Average days to first settlement offer | 71 | 34 |
| Files at risk of passing 120 days | 6–9 | 0–1 |
Look at the bottom half of that table rather than the hours. The hours are money, but the days are the business. Settling in five weeks instead of ten changes what the customer says about you in a review, changes whether your van line's scorecard shows you as a problem agent, and — the one owners consistently underrate — reduces what you pay out, because a claim answered quickly with the inventory item and the photographs attached settles closer to the repair estimate than one answered in November after the customer has spent two months getting angrier.
Draw the line in one place and never move it: the agents assemble, the human decides. Nobody but your claims coordinator or you decides to pay, to decline, or what to offer. Nobody but a person signs a declination, because a declination is a legal position and it needs a name on it.
Three more limits worth writing down. The agents cannot judge whether a scratch on the destination photograph is the same scratch as the "SC" the driver wrote at origin — they will show you both and say they may match, which is genuinely useful and is not a finding. They cannot handle the concealed-damage argument, where the customer opens a carton in December that your crew packed in July; that is a conversation with judgement and history in it. And they should not touch anything that has an attorney or a state regulator on it, where every sentence you write becomes evidence.
Still reading? Stop comparing — try CallSphere live.
See the home services AI agent handle a real call — complete, industry-specific, and live in your browser. No signup.
Also: watch the valuation math yourself for the first month. Released Value at sixty cents a pound on a 140-pound dresser is $84 regardless of what the dresser cost, and that arithmetic being right in the packet matters more than any other single field. Check every one until you have thirty in a row correct.
Do not start with the whole queue. Pull the ten oldest open claims — the ones that make you slightly uncomfortable when you see the delivery dates — and have the batch assembled on those. You will find out two things in an afternoon: how much of your evidence is actually missing rather than merely scattered, and which crew leaders are not uploading destination photographs. Both of those findings are worth more than the time savings in year one.
Then fix the intake, because assembly is only fast if the pieces exist. If your crews are photographing at origin and destination and uploading the same day, this works. If they are not, you have just bought a very fast way of discovering you have no evidence.
Most systems store the documents. Very few assemble a defensible file out of them, and none of them read the driver's handwritten exception codes against a photograph. The 2026 difference is that the work is split across several agents running at once, so a batch of forty takes about as long as a batch of four used to. If your vendor adds this, take it from them; the point is the batch, not who sells it.
It can draft them and it should. Whether it sends them without a human reading first is your call, and my advice is to read them for the first sixty days, then let the plain acknowledgements go automatically while anything with an unusual valuation, a high-value item or a customer who has already complained twice stays in the review pile.
The hour savings will not impress you at 40 claims. The deadline protection will. At small volume the claims work is usually being done by someone whose real job is something else — often the office manager — and that is exactly the situation where a file sits until day 96. Use it for the clocks first and the hours second.
Payouts per claim may rise slightly, because well-assembled files get settled rather than argued into silence. Total cost usually falls, because you stop paying for the ones that escalated only because nobody answered. Track both for a year and judge on the combined number, not on the average check.
One closing note. The claims season is also the season your line rings with "has anyone looked at my claim yet," and every one of those calls interrupts the person who is trying to look at it. CallSphere builds AI voice and chat agents that answer your phone and website chat around the clock, take the claim number and the caller's details, and route it to the right person instead of a voicemail box. It does not decide claims. It keeps the October phone from eating the October claims desk.
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.
A 40-claim DME probe takes 16 working days one at a time and finds the gaps too late. Split four ways, the records requests go out on day two of forty-five.
A 640-line BOM scrub eats three days of your buyer's week. Here is what splitting the RFQ across several agents does to quote throughput at an EMS shop.
Engine fault reports plus driver inspection photos now catch liftgate and emissions failures nine days early. The arithmetic on one stranded loaded van.
A 250-claim PBM desk audit eats three weeks of a technician's time. Four agents splitting the pile turn it into an afternoon and a 19-claim exceptions list.
A 6,100-page open-records request takes 23 days serially. Splitting the first pass across parallel agents cuts it to 6, with every redaction signed by a clerk.
The 19-day lag between a distributor price increase and the new shelf tag quietly eats food margin. What parallel agents change, with the arithmetic.
© 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