Hyper Automation Comes to Telco Networks
As CSPs undergo transformation, end-to-end automation from sales lead to the network is high on their list of critical functions.
By many accounts, we are halfway along the timeline from “un-automated” to “near-complete” automation in all parts of the telco. Today, much of the activity within telcos is focused on several key areas:
- Automation of the network
- Automation of the lead-to-cash cycle for new digital products
- Improvements in customer and agent experience via automation in the contact center
- Marketing automation for campaign delivery
- Automation that will enrich new products and make them attractive to increasingly automated enterprise customers.
The diagram below considers what types of automation are needed within the network, both now and in the future. The box colors denote:
Simple Automation – These processes pose fairly simple problems where it is possible to create rules for the automation to follow.
Simple with Exceptions – In these processes, most scenarios that an automation will encounter conform to rules. But a reasonable percentage require additional human decision making, and this percentage will not reduce to zero over time or with the addition of more intelligence. For example, prioritizing alarms and generating tickets in the incident management column on the diagram is for the most part rules based; however, exceptions will require review and decisioning by humans.
Automation Needs Intelligence – Here, the process has known unknowns, and the relationship between cause and effect requires analysis or expertise. AI/ML copes well because it is possible to look at all sequences of moves and make decisions. For example, human experience of network design could theoretically be mimicked by machine learning and perhaps even coded into a digital twin of the network.
The diagram demonstrates that typical automation deployed to date are either “simple” or “simple with exceptions”. Telcos have selected low-hanging fruit with more immediate return on investment – either simple, rules-based automation such as network element provisioning and management; or automation that increases efficiency within large teams where periodic human intervention from team members is required within the process.
Automation where machine learning can begin to add value (e.g. predictive maintenance, long-term network optimization and network planning) are often not the highest pain points, require still-nascent algorithms or require multi-domain orchestration which is not yet available.
Hyper automation describe the next-generation of automation using AI, machine and deep learning to mimic more human-like decision-making. It will underpin some of the longer-term automation shown as red boxes in the diagram.
In hyper automation, bots learn and add to their knowledge base for future situations, analyzing historical data and previous employee problem resolutions, following a more complete cycle of activities in order to learn and improve.
This next generation will create and hone new and existing network automation underpinning such concepts as the “self-organizing network” with increasingly intelligent load balancing and capacity management. It will also more fully augment the network staff that remain in the NOC – allowing them to more quickly resolve critical issues or more easily solve complex problems.