Six Carrier Portals and the Utility's Interval Data: the Login Round That Pushes Colo Invoices to the 12th

How many browser tabs are open on your provisioning coordinator's second monitor right now? Count them honestly. Lumen Control Center. Zayo Tranzact. Cogent's customer portal. Whichever regional fibre provider serves your second site. The utility's commercial billing portal. The fuel vendor's delivery site. The fire-suppression contractor's inspection portal, which still emails a password every ninety days. The state environmental agency's reporting site, opened twice a year and never remembered.

None of those has ever been connected to anything you own. There is no export button worth using, no feed into your billing sheet, and no vendor on earth who is going to build one for a two-site operator. So a person logs in, reads the screen, and types what they saw somewhere else. That person costs you roughly twenty-six hours a month and, more expensively, they are the reason your invoices go out on the 12th instead of the 6th.

The 3rd-of-the-month round nobody put on the org chart

Here is the round as it actually exists. On the 3rd, your controller logs into the utility's commercial portal to pull last month's interval data for each meter, because metered power billing to your customers has to reconcile against what the utility actually charged you, and the portal gives you a download that expires, a session that times out after fifteen minutes, and a date picker that resets if you sneeze. She does that for each meter, saves the files, and starts building the reconciliation in Excel.

Meanwhile the provisioning coordinator does the daily version: open each carrier portal, check the status of every pending cross-connect and circuit order, note which ones moved a milestone, check open trouble tickets, and update the tracker. Forty minutes on a quiet day. During a fibre cut it is all morning, because the only place the carrier's estimated restore time lives is a page inside their portal that nobody thought to email out.

Computer use means the agent drives a screen the way a person does — opens the browser, logs in, clicks the tabs, reads what is on the page and copies out what it needs — which is the only way to reach a system that was never built to connect to anything else. That is the development, and it is why this round is now automatable when it was not in 2025.

What is genuinely different from the 2024 screen-scrapers

Robotic process automation has been sold into this problem for a decade and it broke constantly, because it was recorded against fixed screen positions. The carrier moves a button and the recording fails silently, which is worse than not having it. What changed is that the agent reads the page and works out where the order status is, the same way a new hire would on their second day. A relabelled tab or a moved column does not stop it.

It also handles the small human obstacles: the cookie banner, the "your session is about to expire" box, the date range that has to be set twice. Those are exactly the things that made the old tools brittle and exactly the things that are now unremarkable.

flowchart TD
  A["06:30 - agent opens the carrier portal list"] --> B["Logs in, reads order milestones and trouble tickets"]
  B --> C["Pulls yesterday's utility interval data per meter"]
  C --> D["Writes updates into dcTrack and the billing workbook"]
  D --> E{"Anything new, disputed, or past its due date?"}
  E -->|No| F["Posts the morning summary to the ops channel"]
  F --> A
  E -->|Yes| G["Flags the controller or coordinator to open the portal"]

Tuesday, 6:45am, before anybody arrives

The coordinator opens her laptop to a summary rather than to nine logins. Three carrier orders moved: one to firm order commitment with a date, one to completed, one slipped a week with a note about permitting. Two trouble tickets updated overnight, one of them the fibre segment that affects a customer who has already emailed twice. Yesterday's interval data for all four meters is in the workbook, and the running total says this month's demand is tracking eleven percent above last month, which in the third week of July is exactly what you would expect but is still worth knowing before the bill arrives.

The completed order matters most, and it is the one that quietly pays for the whole thing. A completed cross-connect is a billable circuit. Today, that fact gets noticed whenever someone next reconciles the tracker — which is sometimes six weeks later, which means six weeks of unbilled monthly recurring revenue you will never go back and collect.

The same round covers the seasonal items. In June, July and August it watches the coincident-peak notifications, because those handful of hours set a large share of next year's demand charges in PJM territory, and knowing on the morning of a forecast peak day is worth more than knowing on the invoice. It picks up the fire-suppression contractor's inspection report the day it is posted rather than the week before the audit. It reads the generator run hours off the maintenance portal so you can show the state you stayed inside the annual limit on maintenance and testing runs for your emergency engines.

Cash-flow arithmetic: what six days off the cycle is worth

Illustrative assumptions for an operator billing about $158,000 a month in recurring colocation, metered power and cross-connect revenue.

Add it up: about $10,300 a year in returned labour, a one-time $31,600 improvement in working capital, roughly $3,000 a year in carrying cost, and the recovered recurring revenue on circuits that were live but not invoiced. The last one is usually the biggest and is the hardest to see, because nobody produces a report of revenue they forgot to bill.

Read-only first, and the clicks it must never make

Start read-only and stay read-only longer than you think you need to. Give the agent its own named account in each portal wherever the carrier permits it, at the lowest permission level that can still see order status. Do not share your controller's login — you want the audit trail to say which account did what, and your own access review evidence has to make sense to an auditor.

Three clicks it must never make on its own. It must never accept or confirm a disconnect, because a mis-read confirmation on a carrier portal takes a customer down and the recovery is measured in days. It must never submit a new order or a rate election on the utility portal — those have contract terms behind them. And it must never approve a change or close a trouble ticket, because closing a ticket ends the carrier's clock on their own service commitment to you.

Two other honest limits. Multi-factor authentication is a real obstacle and the answer is not to weaken it — it is to use portals that support a service account with a hardware or app-based factor your operations manager controls, and to accept that one or two portals will stay manual. And read the terms of use. Some carrier portals restrict automated access; the practical route is to ask your account manager for written permission, which is a five-minute conversation that most people skip and then regret.

The one to start with on Monday

Pick the single portal that produces money, not the one that annoys you most. For nearly every colocation operator that is the carrier order-status round, because a completed order is a billable circuit and the gap between completion and invoice is pure lost revenue. Run it in parallel with your existing manual round for two weeks and compare. If the agent finds one completed cross-connect that your tracker missed, it has paid for itself for the year.

Frequently asked questions

Our utility portal has a fifteen-minute session timeout and a captcha. Does that stop it?

The timeout does not — the agent works faster than the session expires and simply logs back in. A captcha does stop it, and you should not try to defeat one. If your utility uses them, the realistic answer is that the interval pull stays a person's job and you automate the carrier side instead. Be suspicious of anyone who tells you otherwise.

What happens when a carrier redesigns their portal?

Usually nothing, because the agent is reading the page rather than following recorded coordinates. When it genuinely cannot find what it needs, the correct behaviour is to say so and flag a human, not to guess. Make that the requirement when you evaluate anything: what does it do on the morning the page has changed?

Is this safe from a SOC 2 point of view?

It is if you treat those portal accounts like any other privileged access: named accounts, least privilege, quarterly access review, and the activity logged. It is not if you hand it a shared password stored in a spreadsheet — but that was already a finding waiting to happen.

Can it also chase the carrier when an order slips?

It can draft the chase email with the circuit identifier, the original committed date and the current status, which is most of the work. Send it from a human's address. Carrier account teams respond to people they know, and that relationship is worth more than the four minutes you would save.

Last thing, from us: the portal round is one half of the morning: the other half is the phone. CallSphere builds AI voice and chat agents that answer your main line and website chat around the clock — taking the caller's company, cabinet and circuit details, booking the site visit or escort, and routing anything urgent to the on-call engineer. It will not log into Tranzact for you, but it will stop the 6:45am callers from going to voicemail while your coordinator is working through the summary.