By Sagar Shankaran, Founder of CallSphere
Dig season buries one locator in one-call tickets. Splitting the queue across several agents returns 37 hours a week and answers tickets on hour two, not 40.
Key takeaways
Two hundred and fourteen. That is how many one-call tickets landed overnight in the second week of June at a mid-size regional fiber operator I have in mind — a county system, about 4,000 subscribers, 380 route miles, one full-time locator and a part-timer who also drives the bucket truck. Two hundred and fourteen tickets, each with a legal clock on it, all of them screened by one man clicking through Korterra one at a time.
Every state runs it slightly differently, but the shape is identical everywhere. An excavator calls 811. The notification center pushes the ticket to every facility owner whose plant is anywhere near the dig. You have a legal window — in most states two business days before the excavation start date — to give a positive response: no conflict, marked, or contact the excavator. Miss the window and two things happen. Your fiber has no legal protection, and if a directional bore takes out a 288-count backbone at 10am on a Thursday, you eat the splice crew, the credits, the outage report and the angry calls from every school and clinic behind that cut.
The screening itself is not hard work. It is a man in a chair opening a ticket, reading a dig description that says something like "50 ft either side of pole 4471, north side of CR 216, boring for water service," and comparing it against the fiber map in Vetro or IQGeo or 3-GIS — or, at a lot of shops, a KMZ layer and the OSP as-builts. Three outcomes: the dig is nowhere near your plant, the dig is inside the buffer and someone drives out with paint and flags, or the ticket is vague enough that somebody has to call the excavator. Four or five minutes each, honestly done.
Four minutes times 214 is over fourteen hours. In a five-day week that is most of two days spent deciding which tickets matter, before a single mark is painted.
Ticket volume in this business is not flat and every plant manager knows it. It wakes up when the frost leaves the ground, runs hard from April through October, spikes when the county lets a road project, spikes again after a wet spring pushes everyone's schedule right, and collapses in November to twenty a week while the locator finally cleans up as-built records. The people are staffed for the November number. The tickets arrive at the June number.
So the workarounds appear. He screens at night. The office manager gets pulled off billing to help in the last week before the 499-Q is due. Or — and this is the one nobody says out loud in front of an auditor — he stops screening properly and marks everything, because painting a road that has no fiber under it costs forty minutes and a can of paint, and missing one costs a backbone. Blanket marking is not caution. It is a locator burning eleven hours a week driving to jobs that did not need him, while the ticket that actually mattered sits at hour 40 of a 48-hour clock.
Hear it before you finish reading
Talk to a live CallSphere AI voice agent for IT support in your browser — 60 seconds, no signup.
The 2026 change is narrow and it is exactly the right shape for this job. Agent Teams, the research preview that arrived alongside Claude Opus 4.6, lets several agents split one large job and work at the same time, then merge what they produced into one result. That is it. No new judgment, no new cleverness — the same work, not done single file.
A queue of 214 tickets is the definition of a job that is slow only because it is serial. Ticket 190 does not depend on ticket 12. Nothing about ticket 12 informs ticket 190. The only reason ticket 190 waits until Thursday afternoon is that there is one pair of eyes and it starts at the top.
flowchart TD
A["Overnight: 214 one-call tickets land in Korterra"] --> B["Queue split into four batches by dig start date"]
B --> C1["Agent 1 screens tickets 1-54"]
B --> C2["Agent 2 screens tickets 55-108"]
B --> C3["Agent 3 screens tickets 109-161"]
B --> C4["Agent 4 screens tickets 162-214"]
C1 --> D["Merged worklist: clear vs inside the buffer"]
C2 --> D
C3 --> D
C4 --> D
D --> E["Locator drives only the flagged jobs"]
The locator opens his laptop and the night's tickets are already sorted. At the top, in date order, sit the ones where the dig description falls inside the buffer around known plant — those are his day. Under them, the clears, each with a one-line reason attached: "nearest recorded facility 340 ft south, opposite side of CR 216." Under those, a short pile the agents refused to call either way: a ticket with no pole number, a ticket whose address does not exist in the county map, an emergency ticket that needs a phone call to the excavator right now.
He reads the top pile, agrees or disagrees, and the responses go back to the notification center before nine. The clears go back the same morning they arrived instead of on hour 40. And the tickets that used to get discovered on Wednesday when the clock was nearly out are now the first thing anyone sees.
The same trick applies to the other serial paperwork in this business, which is why it is worth setting up once. The twice-yearly Broadband Data Collection filing, where somebody reconciles served locations against the fabric identifiers one address at a time. The pole attachment applications and make-ready responses that sit in a folder waiting for one engineer. The annual Lifeline recertification chase. All the same shape: a big pile of small, independent decisions, currently done single file by the one person who understands them.
Assume 214 tickets in an average dig-season week and 26 weeks of season. Assume 4 minutes of honest screening per ticket. Assume that today, out of caution, 45 percent of tickets get a field mark, and that with proper screening only the 28 percent inside the buffer do. Assume a field mark averages 42 minutes with drive time, and a locator costs $52 an hour loaded.
| Line | Today | Split across four agents |
| Screening hours per week | 14.3 | 1.9 (reviewing flags and the refused pile) |
| Tickets field-marked per week | 96 | 60 |
| Field marking hours per week | 67.2 | 42.0 |
| Total locator hours per week | 81.5 | 43.9 |
| Weekly cost at $52/hr | $4,238 | $2,283 |
| Across 26 weeks | $110,188 | $59,358 |
Call it $50,000 a season, which at that shop is the second locator they keep almost hiring and then not hiring in October. But the number that changes the calendar is different: 37 hours a week come back. That is the difference between a locator who is permanently behind and one who can ride the county road project, sit in the pre-construction meeting, and get the as-builts current before the frost — which is the actual reason tickets are hard to screen in the first place.
Still reading? Stop comparing — try CallSphere live.
See the IT support AI agent handle a real call — complete, industry-specific, and live in your browser. No signup.
Two hard rules. The agents can say "clear" only on tickets where the recorded plant is well outside the buffer and the location is unambiguous; anything vague, anything with an address the map does not know, anything marked emergency goes to the human, always. And the positive response is still legally yours. A person reviews the flagged pile and signs the day's responses. You are not automating the answer; you are automating the sorting that stands in front of the answer.
The larger limit is one every carrier already knows: this is only as good as your records. If your as-builts show a route that was rerouted around a wetland in 2019 and never redrawn, four agents will confidently clear a ticket over live fiber. Run the first month in parallel with the human doing it the old way and compare, not because the software is suspect but because your map probably is. What usually comes out of that first month is a list of map errors worth more than the labor savings.
Partly. If a location exists only on paper in a filing cabinet, nothing digital can screen against it. The honest sequence is to get the backbone and the main feeder routes into a real map layer first — that is the plant a bore actually kills — and screen against that, keeping everything else on the human pile until it is drawn.
It can, and for the clear pile most operators eventually let it, after a season of watching. Start with it drafting responses and a person releasing them in one batch each morning. That is a 90-second review, and it keeps the legal act in a human's hand while you build confidence.
Those should route straight to a phone, not into a queue. Set the rule so anything flagged emergency skips the batch entirely and rings someone. The agents are for the 200 routine tickets, which is precisely what makes the two urgent ones visible.
The screening logic is a week of somebody's attention. The map cleanup is the real project, and it is the one you were going to have to do anyway before your next round of construction.
The other thing dig season generates is phone calls: excavators calling to ask if you have marked yet, customers calling when a bore takes a drop out, contractors calling the office line at 6:50am. That inbound traffic is what CallSphere handles — voice and chat agents that answer the line, take the details, book the visit and get the message to the right person, day or night. It does not screen your locate tickets. It keeps the phone from being the reason your locator never gets through the pile.
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.
Least privilege for AI agents at a regional carrier: scoped logins, no rights in the porting queue or 911 records, and a human on every irreversible action.
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