I spent the last couple of years taking a close look at the application of AI/ML technology to Network Operations. That experience left me super-impressed with some of the new capabilities that come from applying AI/ML to operations data. The most well-known of these is anomaly detection using an auto-generated baseline for metrics not well-served by a threshold alert and the grouping of related alerts into a single incident. Really cool stuff.
However, when I was interviewing for my current job, an exec reminded me of the power of focusing on a customer problem. Put a stake in the ground, and watch the rest of the company self-organize around it. As someone who spent some time in the strategy world, this idea resonates strongly with me as so much strategy literature talks about this, explicitly and implicitly. The Google OKR approach described in “Measure What Matters” has a similar theme.
This brings me to the point of this post. Is AI a product or a feature? Does focusing on being the best AI company address a compelling pain point for enough customers and provide a focal point that will drive competitive differentiation? I do not believe so. When I have seen AI/ML make a significant impact, it is because there was a great deal more work done around the AI/ML, to transform an entire way of working.
When I wrote “NetOps Requires AI/ML & Rules”, it was before I started my current job and perhaps even before I interviewed. However, a little bell was already going off in my head, that AI/ML tech is not the best answer to all NetOps problems, even though it does some things better than alternative approaches. That previous post was at a basic level - rules and machine learning - a big discussion point in AI over many decades. Zoom out from there and you see a gazillion other things that make up the answer to customer problems: tagging, retaining context as you move from screen to screen, a SaaS offering that eliminates the need to install and maintain the core software, auto-discovering dependencies, IOS widgets, widget packs for standardized dashboarding across groups, quick time to value with a large library of out of the box integrations, performance objective tracking, and more.
As the late/great Clayton Christensen reminded the world not long before he left it, there is a job to be done. And that is the bottom line, helping customers get the job done, not just supplying them with technology.
My gut is the bulk of the NetOps market is today, just a little behind some of the radical transformations we may see built around AI in the future. However, when that future day comes, it will not simply be because new platforms delivered a new technology, it will be because that new technology was one of the pieces of an overall approach to solving a compelling network operations pain point, and a job to be done.