AI-ML relevance and importance of model accuracy

Artificial Intelligence (AI) has impacted businesses in a myriad of ways – from enhancing customer experience for eCommerce players to improving efficiency of healthcare treatments. In this blog, Anurag Jain, Deputy General Manager, Cloud & Cybersecurity Services, Tata Communications, discusses the importance of model accuracy when investing in AI and machine learning. 

In today’s highly competitive industry landscape, investment in AI is not just hype or a buzzword, but it is a tool which is here to stay and continue transforming businesses to innovate. B2C organisations have pre-dominantly been the early adopters of AI and Machine Learning (ML). Global tech majors are using AI/ML to redefine the ways businesses can extend better customer experience and services to its end customers while delivering multi-bagger returns to its investors. Gradually, industry is finding ways to leverage advantages of AI/ML in B2B as well. The type of business cases is predominantly similar in both B2B and B2C, but may require different data spreads, data labels, features and volume.

Spread of Use Cases

Data

Industry 4.0 has opened plethora of avenues to capture and gather customer centric data, with software, mobile applications, IOT devices, social media and networks generating zeta bytes of data every hour. This customer data, if processed to generate accurate insights, is a potential gold mine. But if decisions are rooted on inaccurate insights, businesses may quickly bleed millions – in revenues as well as in stock evaluations.

“To overcome this conundrum, it is important that organisations gather quality data and build accurate AI/ML algorithm that accurately interpret data contextually, while at the same time, provide meaningful insights that can be harnessed by the organisation.”

Accuracy Targets

While one may argue that projects should strive to achieve highest level of accuracy for the AI/ML algorithms to generate the most accurate insight and flawless decision making.  But that is not the case always. As the accuracy of model depends on quality of data which in turn depends on several dimensions like

• Consistency and completeness of data
• Coverage of data
• Volume of data
• Relevance of data

AI/ML Low Accuracy – risk.

Though the goal is always to design a highly accurate AI/ML algorithm but as mentioned earlier accuracy of model depends on the several factors. There is an inherent risk associated with the models with low accuracy level. Impact of the risk depends on the use case and domain.

• AI model used for designing a health care solution can have no margin of error as it might be matter of life and death.
• A low accuracy recommendation engine on e-commerce site can self-learn based on changing customer behaviour and patterns.
• Shipping and logistics management companies need to model logistics and route plan for perishable and non-perishable consignments differently.
• In telecom sector, there are tough regulatory requirements (like TRAI in India) to provide seamless connectivity to subscribers. This calls for highly accurate network capacity planning AI/ML models.

“At Tata Communications, we use AI/ML is several ways to deliver excellent customer experience and to achieve financial prudence and operational efficiency goals.”

One such area where Tata Communciations is targeting to leverage AI-ML is in meeting the service delivery commitment to customers by ensuring sufficient stock of CPEs inventory is maintained and network capacity augmentation planned in a way that it doesn’t lead to any delivery delays.

An accurate AI-ML model that facilities efficient inventory management and enables time bound network augmentation will deliver multi-fold benefits as listed below –

• Deliver the services to customer on or before time, which will help retain good customer service perception and help in delivering delightful customer experience.
• Develop good understanding of the industry trends of service/product variability requirements and be ready in time for future bandwidth and device requirements.
• Ensure company can direct the operational and capital expense in a way it delivers maximum returns both in terms of customer satisfaction and return on investment.

At Tata Communications, we are committed to deliver enhanced value to our customers.

“AI/ML plays a pivotal role in the way we define and design the cutting edge and futuristic products in the field of IOT, Cloud, connectivity solutions etc. We envisage to keep leveraging AI-ML in way that it helps us deliver the best solutions and services to our existing and future customers.”

For more information on how Tata Communications uses AI and ML, read this blog here.