With the arrival of the fourth industrial revolution, “the possibilities of billions of people connected by mobile devices, with unprecedented processing power, storage capacity, and access to knowledge, are unlimited”. This puts artificial intelligence (AI) and machine learning (ML) firmly into the spotlight within most Communications Service Provider (CSP) boardrooms, as executives look to the technology to help improve customer experience and reduce operating expenses.
So, where are the smart bets being placed for AI & ML to create business value?
Customer Experience
Given that CSPs are generally not hugely popular with their customers (compared to other industries when measured by Net Promoter Score) and that a number of digital companies have created a ‘new normal’ for customer experiences, it is no surprise that many CSPs are examining how AI & ML can be deployed to improve customer interactions in areas such as marketing & sales, retention and customer support.
There is no doubt that ML can help CSPs take a wide range of inputs from the complex and continuous flow of data available from both network events and customer interactions. This data can be streamed from multiple sources, capturing dynamic events from all customer channels, CRM information and the network itself in order to learn and find hidden combinations.
These insights can be used to drive the appropriate contextual action, including decisions that impact measures such as churn propensity and customer lifetime value. For example, using AI & ML will enable CSPs to generate sophisticated, segmented, personalised offers in real time. These intelligent offers can fall into many categories, such as usage stimulation, loyalty programmes, device upgrades, household engagement and customer education.
Similarly, customers expect to interact with a CSP across a variety of channels, whether its directly through agent conversation, digital self-service or user communities. This is currently the most mature use case for AI, where virtual agents, chat bots and voice assistants are deployed to help automate the answer to customer queries, or even support human agents by helping them with cross-sell and upsell or making it easier to locate the required answer.
Network Operations
Network operations automation is another area where AI & ML will undoubtedly be used and will have high impact to CSPs. The burgeoning internet of everything (IoE) introduces unprecedented scale and velocity into how the network processes events at a level that becomes unfeasible to manage with manual processes. As software defined networks (SDN) and network functions virtualisation (NFV) become the norm, the complexity of these networks will require ML to learn how best to automate and manage the orchestration of network resource and capacity, amongst other functions, to ensure uninterrupted service availability.
CSPs such as Telefonica have already transitioned from a network operations centre to a service operations centre. The goal? To “maximise capacity and solve any problems before end users even notice anything”. Telefonica aim to use data from the network to move from a scheduled maintenance model to predictive and proactive maintenance.
AT&T continue the adoption of their ‘Domain 2.0’ initiative to also transition from hardware-centric to software-centric as they realise the dynamic approach delivered by SDN NFV enables more flexibility at a lower cost. Automating network functions provides a range of benefits to the business and facilitates an improved ability to give customers what they want. “It’s like moving from devices to apps…..we recently brought back unlimited data, one of the reasons we were comfortable doing that is we know this software centric network can adapt to meet the demand.”
Having the ability to analyse network data over time allows AI & ML to predict likely failures and the confidence level that failure will occur, thereby allowing for corrective action to be determined and executed. The end goal here is to combine advanced analytics with AI and allow networks to self-heal and operate autonomously.
Fraud & Security
Security must be a key consideration in the advancement of AI & ML, particularly as IoE growth accelerates.
By applying streaming User and Entity Behaviour Analytics (UEBA) that generate cyber security scores in real time, security teams can easily prioritise alerts associated with anomalous behaviour and actively respond to truly suspicious network activity. The ability to dynamically learn and adapt in real time means that fewer false positives are generated compared to more traditional rule-based approaches.
Importantly, streaming data processes data as it is being generated rather than having latency or relying on large stores of historical log files – it allows CSPs to identify risky behaviour as it occurs rather than after the event. This is crucial to ensure customer data is protected and trust maintained.
As well as protecting themselves and their customers from cyber threats, CSPs are also beginning to use AI & ML to monitor Call Data Records (CDRs) in order to learn what behaviour deviates from the norm and respond accordingly.
Additionally, there are substantial benefits to using AI & ML to identify fraud behaviour and take corrective action. For example, International Revenue Share Fraud is characterised by large volumes of calls to a single destination, to artificially inflate traffic that terminates to international revenue share providers and is identified by examining CDRs. Instead of relying on retrospectively reviewing CDRs once the damage has been done, AI & ML can help to prevent this in real time.
Explainable AI
AI comes with many challenges, including trying to decipher what these models have learned, and thus their decision criteria. One of the major areas of exploration is explainable AI (XAI), which attempts to crack open the black box and explain how and why a model derives its decisions. Explainable AI is required in regulated environments and also to build trust amongst customers and business leaders. This is especially true if CSPs are to really allow machines to make autonomous decisions around mission critical infrastructure such as network operations or security.
At our recent FICO World 2018 in April, chess grandmaster Garry Kasparov — the man famously beaten by IBM’s Deep Blue at chess, and who has become an expert on human-machine collaboration — reinforced the importance of understanding how and why algorithms are making their decisions.
FICO’s own research into this area has produced a number of ways to crack open the black box. These include:
- Scoring algorithms that inject noise and score additional data points around an actual data record being computed, to observe what features are driving the score in that part of decision phase space.
- Models that are built to express interpretability on top of inputs of the AI model.
- Models that change the entire form of the AI to make the latent (hidden) features exposable.With this approach, we are going to rethink how to design an AI model from the ground up, with the view that we will need to explain latent features that drive outcomes.
As CSPs become more familiar with artificial intelligence and machine learning, and the benefits they can bring to streamlining operations, this will lead to freeing up staff to focus on more value-add tasks. Together, people and advanced analytics can improve service, reduce churn and keep businesses and consumers protected from criminal activity.
Edited by Neo Sesinye
Follow Neo Sesinye on Twitter
Follow IT News Africa on Twitter
