The Network Automation Parallel: Decoding AI Adoption Part I

The Network Automation Parallel: Decoding AI Adoption Part I

Companies are spending significant time and effort understanding how to utilize Artificial Intelligence (AI) to deliver value back to the business. By examining how businesses have successfully adopted network automation, this blog explores the potential for AI adoption and offers predictions about its future trajectory based on these comparisons.

This is Part I in this series: it covers the similarities network automation and AI share in their foundational need for clean data and in their ability to unlock new capabilities to deliver value.

It’s All About the Data

Network Automation

The power of network automation, particularly end-to-end workflow automation that drives business value, stems from clean data—data that is organized, accessible, accurate, and formatted for automation. Consequently, a considerable portion of large automation projects focuses on ensuring data quality through cleaning and validation.

Data orchestration is also foundational for sophisticated network automation. Data orchestration governs the flow of data across the different systems in the environment. For example, it allows a user to make an update to data in the System of Record (SoR) for that data and then allows that change to propagate to other systems automatically. That is worth repeating: the data change is made in a single system, and then the change automatically propagates to other systems that need that data. Gone are the days where users have to make the same change in more than one system. An example of this could be that a user adds a management IP address for a device in the IPAM system and that change makes its way to the network inventory system and configuration management database (CMDB) automatically via data orchestration.

The mechanics of data orchestration involve the periodic collection of data from the multiple SoRs for each data type into a single location, the Network Source of Truth (NSoT). Once data is aggregated in the NSoT, the data can be validated and then distributed to other systems that need that data. Data validation includes

• Accuracy: does the data reflect reality?
• De-duplication: remove duplicate data
• Consistent formatting and structure: the data must be formatted and structured correctly (aka “modeled”) so it’s completely clear that a given data point is both unique and formatted to be usable by the automation
• Context validation: using data of one type to validate another type of data

At any meaningful scale, you must solve the data problem before automating: not doing so will result in failed automation. 

Artificial Intelligence (AI)

Many companies aim to develop AI tailored to their specific needs by training it on their own data. Just like network automation, this AI training process relies heavily on clean data. The Harvard Business Review emphasizes key requirements for quality AI training data:

• Accuracy: Does the data accurately reflect reality?
• Absence of Duplicates: Can duplicate data skew the AI’s learning?
• Consistent Identifiers: Does each data point uniquely represent an attribute or object? Uniform data formatting and structure are required to answer this question.
• Correct Labeling: Are the descriptions for each data point accurate, allowing the AI to learn effectively?

The data requirements for network automation and AI share significant overlap! Accuracy, de-duplication, and data modeling are foundational for both.

Companies that have successfully automated have also learned the methods and strategies that produce clean data (data governance) and those that track the lineage of the data (data provenance). These methods and strategies are also essential to AI model training.

AI Predictions:

• Companies with a strong track record in comprehensive network automation will likely have a head start in leveraging AI to create value because the systems and processes to produce data for automation are very similar to those required for AI.
• Conversely, companies that successfully implement AI will find it easier to achieve comprehensive automation.
• Clean data, the strategies and methods to produce it, and data provenance will become strategic for companies because each is foundational for automation and AI.

Unimaginable New Capabilities

I was struck by something a client told me once about how network automation had impacted their organization: “We’re doing things now that we couldn’t have even imagined before we were automated.” This is a great testament to how a comprehensive network automation infrastructure can reveal new capabilities that were simply unimaginable without the context of, and experience within, an automated environment. This client’s statement provides a vivid illustration of how technology can fundamentally reshape business and value creation.

The news here is not that AI will allow for amazing new capabilities. The news is that AI, like network automation, will allow for new capabilities that are literally unimaginable outside of a comprehensive AI implementation. Artificial Intelligence holds the potential for a profound, likely even more dramatic, transformation for companies that fully integrate comprehensive AI infrastructures. 

At the foundation of that infrastructure are the systems and processes that produce clean data. Clean data is everything: it literally powers technological transformation that delivers business value.

AI Predictions 

• A properly implemented AI infrastructure will enable businesses to revolutionize their operations and generate unprecedented value.
• It is difficult to imagine how this value will be specifically created outside of such an AI infrastructure.