By Alex Henthorn-Iwane, Senior Vice President of Marketing, Gluware
Network automation built around intent, where the network is managed through policy rather than manual device-by-device instructions, is recognized as “the way to do automation” for any modern IT organization.
But how you get there depends entirely on where you start. And the two starting points are so different that they define almost everything about how hard the journey is, how long it takes, and what it costs.
Top-Down: When You Get to Start with Intent
Greenfield networks, whether in new organizations or as new or refresh projects, have the rare advantage of starting at the top, because they have minimal vendor and OS complexity. Before a single device is configured, the question is asked: What do we want this network to do? Policies are defined. Standards are established. Intent is captured in a model, and configurations are generated from that model downward onto the devices.
The flow is clean and logical:
Intent → Policy Model → Automated Configuration → Network
Every device in the network is an expression of intent that was defined above it. There is no gap between what the system thinks the network should do and what the network is actually doing — they were created together. When requirements change, engineers update the intent at the top and the change propagates down automatically.
This is genuinely powerful. It is also quite genuinely rare. It requires either building a new network from scratch or executing a full infrastructure replacement. Most network teams get that opportunity once at most, if ever.
Ground-Up: When You Have to Start with What Exists
The vast majority of enterprises are working in the opposite direction. The network is already built. It has been running for years, sometimes decades. Thousands of devices, and possibly hundreds of vendor/device/OS version permutations, are already deployed. Each device instance carries configurations that represent years of accumulated decisions, including routing policies, security rules, access controls, traffic priorities layered on top of one another by engineers who may no longer be with the organization.
For these organizations, the automation journey has to run in reverse:
Network → Configurations → Reconstruct Policy → Translate to Intent
You start at the bottom, with the raw operating instructions on every device, and work upward, trying to understand what the network is doing and why, before you can define what it should do and automate it going forward.
This ground-up journey is harder in almost every dimension. Configurations are written in dozens of vendor-specific languages, across many vendors, device types, and software versions that span generations.
The reasoning behind many decisions has been lost to time and turnover. What you have is like a fossil record: evidence of the intent that once existed, captured in the only artifact that reliably survives the years (the config), but not intent itself.
Working upward from that evidence to a coherent intent model is painstaking. Teams must catalog every device type, interpret configurations that predate their careers, resolve conflicts between approaches applied to different parts of the network at different times, and decide what the policies should be going forward. Then they must assemble this into an intent database, then create the home-grown scripting connectors to deploy against that massively variable network. This work typically takes months before meaningful automation can begin, and often takes years before the whole network is covered. That’s a rough ride.
The reality is often far less ideal. Teams run out of project funding, staff members churn, hotter priorities rear their heads, and the journey to cover all those vendor, device, and OS variants grinds teams down. According to Gartner, most organizations achieve up to 25% automation with a loose combination of vendor-specific tools and SD-WAN controllers, and open source tools. The rest (65% on average) stays in manual mode. Sometimes, that variance leads to failures due to the inability to stitch automated and manual processes together. Oh, and don’t forget that inorganic factors like an acquisition or merger can throw a spanner in the works.
Why the Direction Matters So Much
In the top-down model, intent is authoritative from the start. The model is the source of truth, and the network is its expression. The network starts clean. Drift is detectable and correctable because there is so little to begin with.
In the ground-up model, there is no authoritative source of truth to start from. The configurations are the record. Reconstructing intent from them requires judgment, interpretation, and decisions about what to preserve, what to update, and what to discard, all before the automation can begin to take over.
This is why organizations attempting ground-up automation almost always end up with partial results, at best. The reconstruction effort is so large, and the path so long, that projects lose momentum before full coverage is achieved. Automation takes hold in some parts of the network while others remain managed manually — with all the risk and operational cost that implies.
The Same Destination, a Different Problem
Top-down and ground-up organizations are pursuing the same goal: a network managed through intent, consistently and at scale, without depending on manual intervention or individual expertise. Ultimately, the results all these organizations want from automation is also the same: operational efficiency, consistent compliance, faster MTTR, and improved performance and security.
But the top-down organization is building toward that goal from the first day. The ground-up organization is working to create the conditions for that goal to become possible, interpreting intent from the ground up, establishing a model that didn’t previously exist, and building automation on a foundation that has to be constructed as they go.
Recognizing that difference is the starting point for understanding what ground-up automation actually requires, and why approaches designed for the top-down world rarely work when applied to the ground-up reality that most enterprises actually live in.
The Gluware Solution for Ground-Up Network Automation
Gluware was built from the beginning to solve the ground-up automation challenge faced by the majority of organizations–mainstream enterprises. Over many years, across hundreds of releases, we invested more than a million engineering hours to create the ultimate ground-up intent translation layer. We call that DIAL, which stands for Device Interaction and Automation Layer.
What DIAL essentially does is discover all the operating network’s devices and their configurations across vendors, device types, and OS versions, then extract, abstract, and translate that configuration information into a working intent model.
In other words, Gluware has solved one of the hardest parts of automation, which is enabling data model-driven automation that has a full-stack vendor adapter layer to reliably read and write to a multi-vendor network layer.
What you get when you first go through this discovery and translation process is an intent model that perfectly captures the imperfect state of declarative intent as it is imperatively implemented in configs. I say “imperfect” because of the historical legacy of decisions that weren’t always made with systemic design or understanding in mind.
Once you have that starting point, Gluware offers a purpose-built app suite to enable network organizations to progress from that imperfect starting point to greater and greater levels of automation maturity.
For example, Gluware enables customers to onboard their configurations into policy, then proceed through a transformation process including automating ongoing discovery, performing drift monitoring, establishing audit and OS standardization, then model configs and enforce policy control.
This forms the foundation for automation of end-to-end processes to begin. Intent-based automated config deployments, remediations, and OS upgrade cycles. The intent data model is now in a high-fidelity state, so that it can be relied upon by other systems as an always-accurate source of truth.
Furthermore, since Gluware’s DIAL (translation layer) understands all vendor/OS semantics, it is the perfect validation and governance layer to ensure that any changes that need to roll out from a normalized intent data model can be deployed safely and with predictable results.
The truly compelling part of the story is that the journey from a configuration fossil record to a living, automation-friendly and AI-readable intent model plus a clean network, takes only days and weeks with Gluware. That’s in contrast with months and years of many engineers’ labor just to get to the starting line with manual scripting.
Why This Matters to Agentic Enablement
Every CIO wants to agentically enable their teams to automate better, move faster, and be more efficient. But one of the key problems for mainstream enterprises is that if they can’t get an intent data model working to represent the state of truth for network intent to engineers and automation software, then there is simply no point to adding an agentic interface and trying to create workflows. There will be nothing working enough to flow, AI or not. The ground-up motion for creating an intent data model is a prerequisite for agentic success.
Furthermore, unless there is a translation layer that can safely and predictably govern automated actions executed against a multi-generational, multi-vendor legacy network, no agentic action can be trusted.
Fortunately, Gluware and its DIAL technology solve both of these problems. On top of that, Gluware Titan AI offers agentic enablement that is natively tied into this governed translation layer. The result is that any mainstream enterprise, with even decades-old network infrastructure can become AI-ready and turn on safe, predictable agentic workflows.
Bring Your Whole Network to Intent-Driven Automation and AI
If your organization has struggled to get to full adoption of automation, you may need to take a ground-up approach to translating intent from your network in order to succeed. To get a briefing, consultation, and demo to help discern if that’s the right path for your network, request a demo today.
