The Back-Office AI Gap for ISP Operators Nobody Is Talking About

AI can already cut the back-office grind for lean ISP teams: outage communications, board reports, marketing copy, and provisioning runbooks that take hours from a blank page can become usable drafts in minutes with tools like Claude and ChatGPT. The gap is not the technology - it is the guidance on how to use it.

Recently, at a broadband conference, I had the privilege of speaking on a panel about AI. I was the one talking about the operational side, how I’ve been using LLMs like Claude and ChatGPT in actual sales and marketing work. Campaigns, content, day-to-day stuff. Not theoretical. Not future-state. Right now.

After the panel, I ended up in one of those side-of-the-room conversations that turns out to be worth more than any session on the agenda.

A Fiber Provider Growing Fast, and Burning Out Faster

I was talking to an ops manager at a regional fiber provider. Really sharp guy. His company had grown from 12,000 subscribers to almost 40,000 in under three years. By every measure, a success story.

But he looked tired.

Not vacation-tired. Tired like he’d been sprinting for three years and nobody had moved the finish line any closer.

I asked him what his week looked like. He rattled it off like a grocery list: outage Sunday night, two hours writing comms while simultaneously trying to figure out what broke. Board report due Thursday that he hadn’t touched. Three new subdivisions launching next month with zero marketing materials built. And somewhere in the middle of all that, he was supposed to update their provisioning runbooks because the last person who understood them left back in January.

“I heard everything you said about AI in your session,” he told me. “I know it’s the answer to something. I just don’t know how to start. I get the ‘why.’ I don’t have the ‘how.’”

So I thought to myself, how can I help ISPs with the AI \"HOW?\"

Why Most ISP AI Conversations Miss the Point

I’d been watching the broadband industry’s AI conversation play out all year, and almost all of it was about infrastructure. Network optimization. Predictive maintenance. Automated provisioning. Big, capital-intensive, technically complex stuff that requires a completely different kind of investment and timeline.

None of that was going to help this guy when he got back to the office on Monday.

The conversation the industry actually needed to be having, and nobody really was, was about the four hours he’d spend drafting a board report starting from a blank doc. The two hours on outage comms at midnight. The onboarding documentation that had been sitting on someone’s to-do list since a key employee walked out the door.

These aren’t glamorous AI use cases. They’re not going to headline a conference keynote.

But that’s where the time actually goes for lean ISP teams that are scaling way faster than they can hire.

What Happened When I Tested AI on Real ISP Tasks

So I went home and did something simple. I pulled up Claude, typed out the scenario he described (the outage notification, the specifics, the tone he wanted) and ran a prompt.

In ten seconds, I had a draft that would’ve taken him 20 minutes or more.

Then I tried the board report scenario. Then the subdivision launch copy. Then converting rough notes into a provisioning runbook.

Every single one came back good enough that a thoughtful person could review it, make a few tweaks, and actually use it. Within minutes. Now - let me be clear here. That output is NOT good enough to ship without a HITL (Human in the Loop). You still have to check your work and refine it, but its better than starting from scratch.

That moment it all clicked for me: the gap isn’t the technology. It’s the guidance.

ISP Operators Have the Tools. They Need the Prompts.

ISP operators have access to the same AI tools as everyone else. ChatGPT. Claude. Gemini. They’re not expensive. They don’t require an IT project. They work in a browser. But nobody had built the translation layer for broadband: the specific prompts, the specific workflows, the specific use cases written for people who are actually running these companies.

So we built it. At [Sonar Software](https://sonar.software/index), we work with hundreds of ISPs every day, from Tier 2 and Tier 3 providers to WISPs and rural broadband operators managing subscriber growth, billing, network monitoring, and field operations. We know what the operational grind looks like because we build the BSS/OSS platform that sits in the middle of it.

The ISP AI Prompt Library came directly out of that conference conversation, and a lot of others like it. Here’s what’s inside:

50+ prompts written specifically for broadband operators, covering outage communications, board reporting, marketing copy, onboarding documentation, and more

Context tips for getting better outputs from any LLM (Claude, ChatGPT, Gemini, it’s tool-agnostic)

Real ISP scenarios so you can run a prompt in the next 10 minutes, not after a week of learning prompt engineering

The goal was never to sell anything. The goal was to help ISP teams get their nights back.

Start Here

I think about that ops manager a lot when I talk about this stuff. He already had everything he needed. He just needed someone to point him to the starting line.

If that sounds like your team, if you’re moving fast, doing the work of more people than you actually have, and you know there’s something useful in these AI tools but haven’t had the time to sit down and figure it out, I’d really love for this to be where you start.

It’s free. It’s practical. And if you run even one prompt from it this week, I want to hear what happened.

Get the full ungated playbook HERE.

And if you’ve got a story like the one I just told, the midnight outage, the board report that ate your whole day, the documentation that only exists in someone’s head, DM me on LinkedIn. Those conversations are exactly how this work keeps getting better.

Frequently asked questions

Which AI tools can ISP back-office teams use right now?
The same ones everyone else has: ChatGPT, Claude, and Gemini. They’re not expensive, they don’t require an IT project, and they work in a browser. The prompts in the library are tool-agnostic, so you can run them in whichever LLM you already use.

Are AI-generated drafts good enough to send without review?
No. The output is strong enough that a thoughtful person can review it, make a few tweaks, and use it, but it still needs a human in the loop (HITL). You check and refine the draft - it’s just better than starting from a blank page.

What back-office tasks does the ISP AI Prompt Library cover?
It includes 50+ prompts for broadband operators covering outage communications, board reporting, marketing copy for new subdivision launches, and onboarding and provisioning documentation, plus context tips and real ISP scenarios you can run in about 10 minutes.

____________________________________________________________

*Lindsay Kelley* is VP of Sales & Marketing at Sonar Software, a BSS/OSS platform built for Tier 2 and Tier 3 broadband operators. She writes about AI, marketing & sales operations, and what it actually looks like to scale a lean team.

[View the ISP-Prompt-Library](https://sonar.software/isp-back-office-ai-playbook-sonar-software)