Are automation and AI the same?
Because AIOps for CSPs encompasses AI, operations automation and autonomous networks, it is easy to conflate AI and automation, but they are not intertwined.
13 Jan 2021
Are automation and AI the same?
Read our related report AIOps: From automation to autonomous networks for the full insight.
Because AIOps for CSPs encompasses AI, operations automation and autonomous networks, it is easy to conflate AI and automation, failing to recognize one is a technology domain and the other an approach to business processes and IT. AI is increasingly playing a role in the automation toolkit, either as a way to analyze and optimize processes.
Momen agrees that the hype around AI tends to equate it with automation but thinks each “will continue to generate value based on different use cases.”
CSP processes tend to be siloed even though there are often dependencies across silos. There may be some interconnectivity to address these dependencies, but the silos and processes continue to work independently. For the specific job functions within those silos “traditional software does give us significant leverage” Momen says. Such operations systems’ functions can express decades of domain expertise effectively. They typically use no AI at all to automate portions of operations process, like design, planning, provisioning or service deactivation.
There are also proven use cases for traditional operations software that are AI-like but involve no AI. “Self-organizing networks, anomaly detection, process control and governance, insight and reporting and many more use cases based on traditional software are yielding excellent results without any AI,” Momen explains.
There are areas, however, where a consensus of confidence has emerged around AI’s superiority over traditional software. Momen and others interviewed for this study say AI is the best option for processing extremely large data sets that combine heterogeneous data from multiple sources and where multi-domain or cross-functional correlation is required.
Autonomous networking use cases are driving initial interest in AIOps, inclusive of those instances where AI is not actually part of the automation architecture. CSPs know it will typically take a few steps along the automation path to increase the degree of zero-touch service they can offer and ultimately to arrive at autonomous networks.
Given the complex and often disparate state of many CSP’s operations environments, there are basic criteria for automation CSPs can use to determine where to start. Most of the hard work is practical and detail oriented. “It’s not like you plant a seed and grow an autonomous network,” says María Eugenia Armijo Marchant, Platform Implementation Expert, Telecom Argentina.
Armijo Marchant advises those approaching operations automation and AI to take several key incremental steps, including:
For a closer look at how CSPs are using AIOps to automate processes and AI to being the move toward autonomous networks, download the latest report from TM Forum: AIOps: From automation to autonomous networks.
Because AIOps for CSPs encompasses AI, operations automation and autonomous networks, it is easy to conflate AI and automation, failing to recognize one is a technology domain and the other an approach to business processes and IT. AI is increasingly playing a role in the automation toolkit, either as a way to analyze and optimize processes.
Though AI and automation are related, “they are still not intertwined,” says Mohammed Fahim Momen, general manager, OSS & customer insight, for Robi Axiata. “Rules and input-output based software engines are powerful enough to execute many basic as well as advanced tasks which are not necessarily AI by definition,” Momen says.
Momen agrees that the hype around AI tends to equate it with automation but thinks each “will continue to generate value based on different use cases.”
Automation without AI
CSP processes tend to be siloed even though there are often dependencies across silos. There may be some interconnectivity to address these dependencies, but the silos and processes continue to work independently. For the specific job functions within those silos “traditional software does give us significant leverage” Momen says. Such operations systems’ functions can express decades of domain expertise effectively. They typically use no AI at all to automate portions of operations process, like design, planning, provisioning or service deactivation.
There are also proven use cases for traditional operations software that are AI-like but involve no AI. “Self-organizing networks, anomaly detection, process control and governance, insight and reporting and many more use cases based on traditional software are yielding excellent results without any AI,” Momen explains.
When AI wins
There are areas, however, where a consensus of confidence has emerged around AI’s superiority over traditional software. Momen and others interviewed for this study say AI is the best option for processing extremely large data sets that combine heterogeneous data from multiple sources and where multi-domain or cross-functional correlation is required.
For example, “360-degree assurance of network performance and customer experience requires a unified platform and solution and there the need for AI arises,” Momen says. AI may also be a better fit than traditional software for automating responses to continuously changing network configuration and customer experience-related requirements.
Where to focus automation effort
Autonomous networking use cases are driving initial interest in AIOps, inclusive of those instances where AI is not actually part of the automation architecture. CSPs know it will typically take a few steps along the automation path to increase the degree of zero-touch service they can offer and ultimately to arrive at autonomous networks.
Given the complex and often disparate state of many CSP’s operations environments, there are basic criteria for automation CSPs can use to determine where to start. Most of the hard work is practical and detail oriented. “It’s not like you plant a seed and grow an autonomous network,” says María Eugenia Armijo Marchant, Platform Implementation Expert, Telecom Argentina.
Armijo Marchant advises those approaching operations automation and AI to take several key incremental steps, including:
- Understanding and identifying processes & operations that are well suited for automation.
- Improving these first with “simple automation.”
- Gathering data to understand the automated operations environment better.
- Testing AI-assisted automated interventions and measuring whether there are gains and improvements.
She warns peers not to take automation’s complexity lightly and urges those who seek to automate operations processes to respect the “mandatory conditions” for effective automation. “The more deterministic a process is, the better automation will fit it,” Armijo Marchant says.
For a closer look at how CSPs are using AIOps to automate processes and AI to being the move toward autonomous networks, download the latest report from TM Forum: AIOps: From automation to autonomous networks.
