Before You Automate Anything, Fix the Data

—

AI makes it very easy to start at the end. We see what it can automate, predict or generate, and immediately start thinking about where we can put it to work. But in network operations, AI inherits whatever reality we give it. If that reality is incomplete, outdated or scattered across systems, we’re not starting with an AI problem. We’re starting with a data problem.

There is a temptation right now to start every AI conversation with automation. What can we automate? How much manual work can we eliminate? How quickly can we deploy it? For fiber operators, I think there is a more important question to ask first: How much do we trust the data?

If pole data is wrong, a permit status is three weeks out of date, field evidence is incomplete, or a closeout package is missing documentation, AI doesn’t make those problems disappear. It can process the information faster. It can also give you the wrong answer faster—and potentially with enough confidence that someone acts on it.

That is why I would start with the data before the automation. Operators need to understand what information they have, where it came from, whether it is current, and whether it reflects the work actually happening in the field. That isn’t glamorous work, but it determines what AI will eventually be capable of doing reliably.

Once that foundation is there, the AI conversation gets much more interesting. Now you can automate repetitive work, identify gaps earlier and help people make decisions with information they can actually trust. The goal shouldn’t be to put AI everywhere. It should be to make AI useful where it matters.

I think we’re going to see a real separation between organizations that simply add AI to existing processes and those that do the harder work of making their operational data trustworthy first. The second group may appear to move more slowly at the beginning. But when they start applying AI, they’ll be building on something real. That’s a much better place to start.