AI Maturity Is Not About Intelligence. It's About Trust.
AI means different things depending on who is selling it and who is buying it, and broadband operators are still sorting out which capabilities actually belong in their operations. This conversation covers how to think about AI maturity, where the risk tolerance differs across billing, support, and network ops, and what happens when the subsidized pricing era ends.
With Larry, Georgette and Rick
August 18, 2026 · listen
Watch this episode
Watch it here, or listen in your app of choice.
Show notes
Trust Is Built One Step at a Time
An ISP doesn't wake up one morning and decide to let AI disconnect customers or reroute network traffic. Trust develops gradually.
First, we ask AI to summarize information. Then we ask it to identify patterns. Eventually, we may ask it to recommend a course of action. Only after those stages have proven themselves do we begin discussing autonomous decision making.
Every one of those steps introduces a different level of organizational trust.
Support teams may become comfortable with AI much sooner than finance. Marketing might confidently use AI to help draft campaigns while engineering refuses to let it anywhere near network configurations. None of those approaches are inherently right or wrong. They're simply different levels of acceptable risk based on the potential impact of a mistake.
AI Should Help People Make Better Decisions
One of the biggest misconceptions is that AI exists to replace people.
I don't believe that's the goal.
The real opportunity is allowing people to stop spending hours searching through thousands of subscribers, endless reports, or mountains of operational data looking for patterns they may never find on their own.
As our businesses grow, our instincts can only take us so far.
When you're serving hundreds of subscribers, experience and intuition may be enough. When you're serving tens of thousands, no one can manually recognize every trend hiding inside that data. That's where AI becomes valuable. It surfaces opportunities, highlights risks, and helps us ask better questions than we would have asked on our own.
That's the part that excites me most.
The technology isn't replacing critical thinking. It's strengthening it.
Healthy Skepticism Should Apply to Everyone
One point from our conversation really stuck with me. Human beings make mistakes every single day.
We miss follow ups. We overlook trends. We enter incorrect billing adjustments. We fail to connect related events. Yet when those things happen, we usually don't question whether humans should continue doing the job.
AI, on the other hand, can produce dozens or even hundreds of correct recommendations, make one mistake, and suddenly people conclude it can never be trusted again.
I'm not suggesting AI deserves blind trust. I'm suggesting humans deserve healthy skepticism, too.
The standard shouldn't be perfection. The standard should be measurable improvement.
Adoption Is About People More Than Technology
The technology itself isn't what concerns me most.
Organizations change much more slowly than software does.
People naturally worry about losing control. They worry about changing workflows. They worry about job displacement. They wonder who owns the decisions AI helps make and who remains accountable when something goes wrong.
Those concerns are understandable.
Successful AI adoption isn't about forcing people to use new tools. It's about helping teams understand why those tools exist, where they create value, and where human judgment will always remain essential.
Trust isn't built by deploying software.
It's built through transparency, communication, and thoughtful implementation.
The Conversation Doesn't End Here
We're still early in this journey.
The technology will continue changing. New models will emerge. Costs will shift. Best practices will evolve.
The organizations that benefit most won't necessarily be the ones chasing every new AI announcement. They'll be the ones that carefully evaluate where AI creates real value, where people should remain in the loop, and how to build lasting trust across their teams.
That's ultimately what AI maturity looks like. It doesn’t just look like bigger models or more automation. It looks like better decisions, stronger organizations, and people who feel empowered by the technology instead of threatened by it.
Listen to the Full Episode
Catch the full conversation on the Bandwidth podcast.
Available now on Spotify, Apple Podcasts, and the Bandwidth YouTube Channel.
If you enjoyed this episode, be sure to subscribe wherever you get your podcasts so you never miss a conversation.
Questions, answered.
What is the difference between AI, machine learning, and LLMs for an ISP operator?+
AI is a broad term for computers reasoning in ways that approximate human thinking. Machine learning is a subset that has been running in production systems for decades. LLMs, like ChatGPT, are a specialized type of machine learning that predicts the next word in a sequence. They are related but not interchangeable, and the distinction matters when evaluating what a vendor's tool actually does.
Where should operators start with AI adoption?+
Internal operations are the lowest-risk entry point because mistakes there do not reach subscribers. From there, the practical progression is insights first, then recommendations with a human making the final call, then action in bounded and well-logged scenarios. Do not skip the first two stages to get to autonomous action.
How should operators think about risk tolerance across different departments?+
The blast radius is what matters. Network operations carries the highest risk because errors affect all subscribers at once. Billing errors create immediate trust problems. Customer-facing support sits in the middle. Internal staff tools carry the least exposure. Risk tolerance should reflect those differences, not a single company-wide threshold.
Should AI be making customer-facing decisions before network decisions?+
The panel's position is no. Customers should not be where experiments happen. Closed network testbeds exist for a reason. A mistake in a subscriber-facing AI deployment erodes trust quickly and cannot be walked back by explaining that the AI made the call.
What happens if AI provider pricing increases dramatically after an operator has built a dependency?+
Track ROI now so any price increase can be evaluated against a real number. Explore locally hosted open models for common, repetitive workflows that do not require frontier model capability. LLM routing services can direct simpler queries to lower-cost models automatically. The goal is not to reverse course but to reduce concentrated dependency before the pricing landscape shifts.