Telecoms.com periodically invites third-party experts to share their views on the industry’s most pressing issues. In this piece, Suresh Chintada, CTO at Subex, explains why enterprise AI and augmented analytics are important technological trends for telecommunications.
The telecommunications industry is undergoing profound shifts, driving all forms of person-to-person, machine-to-machine and machine-to-person communication. The competition is fierce; Trends such as deregulation and consolidation make it crucial for operators to differentiate cost-effectively.
Like other industries, artificial intelligence (AI) has also created waves in telecommunications. The vow? Rapid differentiation through cost optimization, service agility and new business models. These are an expected given for any telecommunications company aiming to succeed in today’s digital, competitive and disruptive marketplace.
Gartner predicts that by 2024, 75 percent of companies will switch from piloting to operationalizing artificial intelligence. This equates to a fivefold increase in streaming data and analytics infrastructures.
There are already plenty of success stories about AI in telecommunications. Those who have piloted AI implementations testify to their ability to improve decision-making, profitability, and operational efficiency. It’s not hard to imagine a future where AI plays a key role in features like proactive customer service, timely fraud management, holistic business security, optimized network quality, better asset management and 360-degree partner management, to name a few.
Despite its potential, however, operators are cautious. The entire telecommunications industry is playing on managing margins, and therefore operators are wary of the high investment costs of implementing AI. Ironically, the ability to respond quickly to market changes will promote profitability – and this ability is exactly what AI and augmented analytics provide.
In the past, companies had the benefit of business intelligence tools. But in a post-pandemic world, responsiveness and resilience are key traits that separate leaders from the laggards. Achieving this begins with democratizing access to data and leveraging AI in a cross-functional way.
Build to scale – enterprise AI for telecommunications
Innovation in both operations and business models is unlimited in the telecommunications industry. A study by the McKinsey Global Institute shows that the high-tech and telecommunications industries are forerunners when it comes to implementing AI solutions. Many users also report revenue increases thanks to AI.
In telecommunications, the buzz is tangible: Operators are testing machine learning models to test efficiency and measure value. The measurable success measures are minimizing capex, maximizing revenue and network optimization. Some examples include:
- Customer Experience – AI can conduct customer experience analysis, provide personalized recommendations, and improve campaign management.
- Networking – It can detect usage irregularities, allowing communication service providers to address disruptions through timely maintenance or asset recovery.
- Earnings – AI models can help operators leverage lucrative business opportunities that leverage existing investments.
Looking ahead, telecom operators need to shift their focus from simply implementing AI to scaling it across the enterprise. AI drives a fundamental internal shift, enabling telecommunications companies to avert capex-heavy investments while delivering ongoing value that is translated into direct, tangible benefits. As it is all-inclusive, it can easily expand its capabilities across telecommunications companies and deliver superior efficiency, smarter insights, continuous improvement and new business opportunities.
Barriers to corporate AI
Holistic AI implementation, while certainly a business goal, is typically hampered by operational constraints, rigid thinking, and manual processes.
AI is inherently complex, making the right technological expertise imperative for a successful implementation. For example, AI includes several models for machine learning, deep learning, and computer vision. Business users would be unaware of which model to choose for their specific business problem.
Since they are esoteric, AI models also require niche skills. The seemingly simple task of data preparation requires knowledge of activities such as investigative data analysis, data transformation, processing of missing values, normalization, coding, etc. Therefore, lack of talent is an ever-present challenge. The availability of data researchers and data engineers is sparse, heavily impacted by cost and expertise.
Changing human thinking and organizational culture is a prerogative. Many companies are still dependent on manual computer science processes, which affect productivity. A 2020 report by Anaconda, head of data science and distributor of the Python and R programming languages, showed that nearly 45 percent of a computer scientist’s time is spent simply preparing the data for models and visualizations.
Building user trust is another primary concern. The granular aspects of black box models usually remain unknown to the actual users of AI. Inherent model complexity means that a telecommunications company’s business team does not understand the logic or how the algorithm arrives at the final result. In processes that require decision-making, this knowledge is paramount.
Transparency supports enterprise AI
Today, companies are increasingly demanding that AI models be transparent, explicable, and accountable. According to Forbes, explanatory AI is about understanding how a model arrives at specific results. It is also about understanding how decisions are made by models and how models correct their own mistakes. Without some form of ‘explanation’, the tendency to change is rigid and adoption idle.
Models need to be accurate if they are to gain user trust. Therefore, model bias is a problem. AI-based loan models that provide recommendations that are skewed to customers from a particular region, gender, or race are an example of model bias. Solving these problems involves lengthy experiments. In parallel, nearly 50 percent of the initial experiments fail, requiring consistent adjustments to the model.
The trick here is to fail quickly and go iteratively – and quickly – to the next prototype. But to experiment and fail quickly, organizations should also be able to accelerate how they select, build, implement, and test models.
Without a single, comprehensive and proven platform to perform the above activities, telecommunications companies rely on different systems, even when implementing AI, which poses integration challenges, compromises the user experience and limits the potential of enterprise AI.
Augmented analytics – democratize AI and empower data researchers for citizens
Self-service extended analytics platforms help telecommunications companies democratize AI in a simple, user-friendly and automated way. They help operators experiment quickly using a code-free setup that offers self-service options in a single, one-stop solution. Some of the primary ways in which extended analytics supports enterprise AI are as follows:
- Experiment iteratively for accurate models – There are several tools embedded in extended analysis platforms that remove some of the data researchers’ burdensome tasks. For example, Auto-ML (machine learning) supports ‘last mile optimization’ and is based on actual results that enable models to price adjust and deliver results with increasing accuracy. Here, too, the process is iterative and automated. The Auto-CASH (Combined Algorithm Selection and Hyperparameter Optimization) modules help to select the best model and the best set of hyperparameters to optimize the selected evaluation metrics (accuracy, precision, lift, etc.). These tools greatly increase user productivity.
- Accelerates data life cycles through automation – Augmented analytics provides a governing framework with workflows to manage data efficiently. Tasks such as data preparation, tuning of the hyperparameters, selection of the most suitable model, implementation of it in production and monitoring of its performance are automated end-to-end.
- Reveal model logic through explanatory AI – Extended analysis platforms support interpretation of black box models. Users can easily understand the logic behind predictions across global, regional and local levels of explanation. Machine learning de-biasing is an additional feature to combat model bias. It helps business users trust the results of the model, especially for outputs related to fraud, customer service, business insurance, revenue leakage and so on. A robust extended analysis platform also helps monitor key evaluation and performance measurements such as precision, recall, function operation, model operation, etc.
Benefits of enterprise AI
The biggest advantage is that expanded analytics platforms allow business users to become citizen data researchers. First, the intelligent automation of data management improves the efficiency and productivity of existing data researchers. Second, it helps users easily utilize AI to solve business problems without relying on exhaustive training and domain knowledge.
While these provide a massive boost to the quality of work, there are also clearly quantifiable benefits.
Companies that have adopted expanded analytics platforms report a 50 percent increase in analytics efficiency and decision security. Automated function synthesis helps data researchers roll out more accurate models, iteratively, quickly and without user bias. Insight is readily available as data processes run up to 100 times faster. Sharp visualization of these insights and patterns is a bonus. Finally, conversation analysis makes it even easier for business users to consume this insight.
When we move to the bottom line, extended analytics boosts revenue in two ways:
- Revenue through efficiency gains – Telecom operators can expect operational profitability to increase by 23 per cent. Employees become more productive and feel helped in their work; retention increases by 31 percent and so do the new-found opportunities for value-creating tasks. This includes nurturing citizen data researchers, who can then build AI models for other functions, amplifying value – and return on investment – across the company. Some companies report having increased their data research pool for citizens by almost five times thanks to expanded analytics platforms.
- Revenue from customer satisfaction On the front-end, customers who enjoy increased personalization, faster problem solving and network quality (among others) report greater satisfaction. Some users report a 35 percent year-over-year increase in customer acquisition.
Conclusion: Speed up data-to-decisions
Expectations for artificial intelligence are sky high. Augmented analytics is the crucial differentiator that will differentiate those who win big through AI investments and those who lag. Extended analytics platforms can help players accelerate the data-to-decision lifecycle, giving them a sharper edge. Specifically in telecommunications, they balance the discussion of investments and benefits by optimizing existing processes and uncovering new business opportunities, thereby minimizing costs and maximizing revenue.
Suresh is the CTO of Subex and brings with him a broad leadership, managerial and technical experience of over 27 years. Prior to Subex, he worked with companies such as Motorola, ARRIS and CommScope, where he built and scaled major global software engineering, professional services and technical support services, and serviced industrial verticals such as cable, telecommunications, mobile and wireless networks. Suresh holds a bachelor’s and master’s degree in electronics and communications engineering from Osmania University and a Post Graduate Diploma in software enterprise management from IIM, Bangalore.