Cybersecurity

Barriers to Automation: If It’s Such a Good Idea, Why Is It So Hard?

Explore the challenges of network automation, from human error to organizational barriers, and discover strategies for effective implementation.


Our last post in this series [Drivers for Network Automation : Why the Pressure to Automate is no longer Optional] went through the pressures pushing organizations toward automation: human error, configuration drift, multi-vendor sprawl, a thin bench of talent, and a business that expects networks to move at software speed. Put that case in front of almost any network team and they’ll nod along. If the case for automation is that strong, a fair question follows: why isn’t everyone already there? Because automation is genuinely hard. Because the obstacles aren’t mostly technical, they’re organizational, and they show up in the same handful of forms again and again.

Everyone rebuilds from scratch

Picture a consultant who joins their fourth company in five years, opens the automation repository on day one, and finds… a brand-new homegrown framework for pulling device inventory, built from scratch, again. Not because the previous solution was bad, but because nobody at the new company knew it existed, or trusted it, or had time to evaluate it properly. Multiply that story across an entire industry and you get thousands of teams quietly reinventing the same handful of building blocks — inventory systems, config pipelines, approval workflows — with no common reference for what a mature solution even looks like. Without that reference, teams learn what “good” means the expensive way: by shipping something fragile, watching it break, and rebuilding it from the wreckage.

A pile of scripts is not a platform

Writing your own automation in-house is almost always the first thing a team tries, and it’s almost always harder than it looks going in. The scope creeps: What started as “a script to pull configs” grows an approval step, then a rollback mechanism, then a scheduler, until it’s effectively a small application that nobody planned to build and nobody was staffed to maintain.

Just how common this is was the whole subject of a 2025 Enterprise Management Associates study, “From Scripts to Platforms,” built on interviews with a dozen enterprise network automation engineers. EMA’s research puts the number of enterprise networking teams running on homegrown scripts or software at 64%, and found that 61% of those teams spend six or more hours a week just debugging and maintaining what they’ve built. Researcher Shamus McGillicuddy frames the underlying tension as the industry’s own “80/20 rule”: commercial platforms handle roughly 80% of automation use cases well, but the remaining 20% — the genuinely unusual stuff — tends to resist anything but custom code, which is exactly the stretch that eats the most engineering time. A separate, larger EMA survey of 354 IT professionals found the downstream effect: Relatively few organizations describe their automation efforts as fully successful, with the clearest wins concentrated in a couple of measurable benefits — about a third of respondents credited automation with real operational efficiency gains, and roughly the same share pointed to reduced security risk.

The bus-factor problem

Software teams have a blunt name for this one: the “bus factor” — how many people would need to get hit by a bus before a project is in serious trouble. For a lot of homegrown automation, the honest answer is one. The pattern is depressingly familiar: a solution ships, it works, everyone’s happy — and then six months later something needs to change, the original engineer has moved to a different team or a different company, and whoever’s left is reverse-engineering a codebase with no design notes and no explanation of why anything was built the way it was. The failure wasn’t bad code; it was skipping the step where intent gets written down. Automation built without documentation isn’t really an asset — it’s a loan that comes due the moment its author leaves.

Complexity compounds like interest

Automation rarely stays the size it started at. A single helpful script picks up a second one to trigger it, then a third to check its work, and within a year you have something that behaves like a small distributed system — with all the integration points, edge cases, and failure modes that implies. Left unmanaged, that growth compounds the way debt does: each addition makes the next change a little riskier and a little slower to make safely. It’s also not just your own code; today’s landscape of automation tools is large and still consolidating, which means picking a stack is hard enough, and maintaining one over several years is harder still.

Nobody had agreed on what the words meant

Perhaps the quietest barrier of all is vocabulary. Ask five engineers to define “orchestration” and you’ll often get five overlapping but different answers, each shaped by whichever tool that person happens to use. That’s precisely the wall the Network Automation Forum ran into early on: practitioners arguing past each other, not because they disagreed about the goal, but because they were each anchored to their own tool’s terminology. The fix wasn’t a better tool — it was a shared set of names for the functions any solution needs to cover, which is exactly the framework we covered in the second post of this series. If you haven’t read that one yet, it’s worth circling back to, because it directly resolves several of the barriers above.

The barrier that isn’t technical at all

Underneath the structural problems sits a quieter, human one: automation can feel like a threat. An engineer who has spent a decade being the person who knows the network by hand can reasonably wonder what’s left for them once a pipeline does the routine work. Leadership, meanwhile, often wants a tidy business case before funding anything that touches production, which can make automation feel like it has to justify its existence twice as hard as anything else. Neither reaction is irrational — most people who resist automation have, at some point, watched a broken manual process get automated, which just made it break faster. The answer isn’t to push harder against that resistance; it’s to design and communicate well enough that people can see automation is making their job better, not replacing it.

None of these barriers are reasons to avoid automation — they’re reasons to approach it on purpose instead of by accident.

— AGN

Clearing the barriers instead of colliding with them

Look closely and nearly every barrier above traces back to the same missing ingredient: a shared vocabulary, an actual design phase, and a clear owner for what happens after the first version ships. That’s the gap AGN is built to close.

We start with an honest look at where your team and your network actually stand, design the solution against the Network Automation Framework rather than around whatever product we’d like to sell you, and carry it through deployment with the documentation and knowledge transfer that keep it from becoming next year’s mystery codebase. Whether you’re stuck on “where do we even begin” or trying to tame a pile of scripts that has outgrown the person who wrote it, AGN can help you get past these barriers instead of running headlong into them. Learn more on our automation services page [link not available yet].

Alchemy Global Networks — Vendor-neutral network automation, done by design.

Similar posts

Subscribe to The Alchemist's Notebook

Get The Alchemist's Notebook delivered straight to your inbox: the Alchemists' latest thinking on cybersecurity and network engineering for Federal, State and Local, Education, and Enterprise organizations.