“AI cannot be expected to produce impactful insights without it being set up to suit the needs” of a business”


With AI, businesses are able to identify and unearth hidden revenue streams, run highly personalised campaigns on a large scale and garner more meaningful insights, as compared to traditional resource-driven analytics and model development. Sudeshna Datta, co-founder and executive VP, Absolutdata, talks to Deepshikha Shukla, technology journalist, Electronics For You, about how AI can eliminate the need for manual intervention, while enhancing productivity and giving businesses access to massive data sets in real-time.

EB: How does an AI platform deliver scalable business improvements across the enterprise?
Given the rapid pace at which AI is penetrating industries across the world, those who haven’t implemented it risk being left far behind. Therefore, enterprises are now looking to hop on to the fast track to make themselves not just AI-ready, but AI-driven, in order to achieve massive business transformation and scalability. Not only is AI flexible in its implementation with existing systems, but it enables companies to see results across the enterprise at an instant. However, simply adding AI to the mix isn’t going to result in a competitive edge. Instead, companies need to focus on implementing AI with the three key elements — an analytical framework, context and technology (or ACT).

AI cannot be expected to produce impactful insights without it being set up to suit the needs of a business. It takes an efficient analytical framework to guide the process of building an AI solution that is tailor-made to suit a particular business and this should be programmed to solve specific problems. Though AI is self-learning, it needs context to be able to carry out its functions. It acts on the inputs that are provided by human teams or other connected systems. That is why equipping AI with business context is important for businesses to achieve their broader objectives.

Further, what brings together several elements for the successful deployment of AI is the right technology. Businesses should always try and choose AI tools that cover end-to-end processes, as opposed to those that are made for just a specific function.

Using AI tools combined with the ACT strategy ensures that they are implemented seamlessly and can be scaled across the organisation, once successfully used. This also helps to integrate and streamline business functions and processes across departments, creating a harmonious environment and delivering superior outcomes. Our approach towards helping businesses achieve this is to start with a crucial business problem and build a platform around it, which has the ability to expand and be applied to other functions across the enterprise.

EB: What is the role of the technology stack in enterprise architecture?
Tech stacks help enterprises mine existing sales data with the help of analytical frameworks and natural language processing. With this, businesses are able to identify and unearth hidden revenue streams, run highly personalised campaigns on a large scale and garner more meaningful insights, as compared to traditional resource-driven analytics and model development. This largely eliminates the need for manual intervention, while enhancing productivity and giving businesses access to massive data sets in real-time.

Companies today are in the early stages of building their intelligence stack and, by combining AI, data analytics and technology, our NAVIK AI platform enables them to build a model which gives them a sustainable competitive advantage.

EB: Why is a machine learning algorithm required for specific analytical purposes?
Machine learning (ML) algorithms differ in how they are applied to address various businesses’ problems and functions, and vary from case to case. Certain problems may be very specific in nature and would require a custom approach while others may be quite open-ended, thus requiring an approach involving trial and error.

Narrowing down to a particular ML algorithm is largely based on business decisions. The right model can help enterprises discover existing connections that are beyond the reach of human teams. Further, a precise model can help an organisation accurately identify viable opportunities while avoiding areas involving major risks.

Some of the most important factors that help businesses choose the ideal algorithm are understanding the amount and type of data they have at hand and categorising the problem. Moreover, businesses need to check the requirements pertaining to the input and output, as well as keep in mind the limitations for the business. At times, it may be hard to know, right at the start, which algorithm will work best, but often, it’s best to work iteratively.

Let’s take the example of a Fortune100 global beverage brand that was looking at improving sales in its food service division. The company needed an AI-enabled platform to build multiple tools for different use cases, starting with a beverage assortment optimisation tool.

The challenge was to identify accounts, regions or individual outlets with the highest potential sales gap for a given category, brand or SKU (stock keeping unit). A robust state-of-art recommendations engine (an ensemble of deep and wide neural networks, KNN or K-Nearest Neighbours, collaborative filtering, business rules, etc) recommended an optimised beverage assortment solution for the food service operator, with details on the expected increase in volume, revenue and incidence.

The highly scalable, fully automated self-learning solution solved one of the most burning problems in analytics, i.e., arriving at an optimal trade-off between existing and new product recommendations in a widely varied landscape. Specific to the client, this was achieved using a custom cost function in the recommendation model. This led to a 3 per cent expected incremental revenue and 4 per cent expected consumption volume growth across categories.

EB: How can an AI-driven platform empower the sales force?
Sales professionals have to make critical decisions every day—be it about deciding which prospects to reach out to, what product and service offerings to highlight, or which communication channels will work the best. Many sales people tend to make these decisions based on their intuition or by following an organisational playbook. AI emerges with a better alternative in such a scenario, making the process buyer-centric while applying its capabilities on the data. While traditional sales playbooks mostly comprise predictive selling, AI takes it to the next level by transforming playbooks from being set and static to dynamic and ‘living’. These living playbooks keep getting adjusted over time with deep learning, which uses outcomes and new data to adjust algorithms. Such flexibility for adjustments enables playbooks to continuously learn, improve, and become more accurate.

The NAVIK AI platform develops an intelligent self-learning system that continuously learns from your unique business environment and practices for better decision-making. When applied to data, AI provides weekly guidance in identifying the most promising contacts, specifying which products or services each contact is likely to purchase next, and suggesting the most effective communication channel to use to reach out to them. Product recommendations outline the likely reasons that are motivating a purchase and putting the buyer at the centre, thereby giving the sales persons talking points that resonate with their prospects.

Our product, NAVIK SalesAI, has been proven to increase sales by providing predictive data on buyers, identifying the most promising prospects, and aligning the purchaser’s needs with products as well as communication channels. On providing a customised weekly game plan of the leads and actions on priority, data-driven recommendations empower sales professionals to close more deals, faster.

EB: What are the benefits of a customer-directed AI-marketing platform?
The applications of AI and the concept of the ‘digital twin’ have been helping larger players scale the new strategy of personalisation more quickly. For years, the well-established notion among marketers has been that customer-centric approaches work far better than the rest. However, the problem now is about targeting each customer in a meaningful way, which clearly makes it imperative for a better type of personalisation and segmentation to be planned and implemented. AI-enabled customer-directed marketing supports companies to conduct marketing in a more personal manner.

Digital twins in the marketing landscape indicate how customer segmentation has logically evolved. Just like the other versions of digital twins, the digital twins in marketing are essentially a collection of data that can be leveraged to run tests on and predict results accordingly. These digital twins present a deeper, more accurate and comprehensive picture of a consumer, thereby helping brands pitch relevant messages to customers on a big scale. All this information gets delivered to the marketer through an AI bot, which also provides recommendations to the marketing team regarding what campaigns are more likely to succeed, how churn and other negative changes can be prevented, and what actions should be taken to mitigate any migratory behaviour.

AI-powered customer-directed platforms enable businesses to perform live campaign optimisation and modification, rapid personalisation, dynamic segmentation, and run multiple targeted campaigns for their marketing initiatives.

NAVIK MarketingAI powered by the NAVIK AI platform, hyper personalises offers for maximum results. For instance, a global hotel brand leveraged MarketingAI to solve the decades-old issue of ‘one-and-done’ customers, i.e., guests who stay at a hotel only once and never return. As part of this, a controlled A/B test was conducted over a period of eight weeks. It was observed that the campaigns that used MarketingAI experienced a revenue increase of 51 per cent from opened emails, 31 per cent more customer bookings, and a two-fold revenue increase from customers who did not sign up to any loyalty programme.

EB: How do Micro Modules or MicroMods in the NAVIK AI platform help to perform customised specific tasks?
The NAVIK MicroMods are in essence ‘plug-n-play’ ML algorithms built for specific analytical needs to help clients achieve scalable business improvements with their AI initiatives. Some of the features of NAVIK MicroMods are listed below.

Plug-n-play AI: It consists of a library of ready-to-use algorithms to identify the organisational requirements and deploy AI-driven solutions exactly where the enterprise needs them.

Rapid deployment: Prebuilt predictive and ML models are leveraged and used in a unique business-specific manner.

Easy to embed: They include REST APIs and integration APIs, which facilitate the plug-in process of complex solutions into existing infrastructure. NAVIK provides the ML REST API to ace the predictive ability of enterprises.

MicroMods are programmed to perform specific tasks such as behavioural prediction. Dedicated APIs understand how customers will think and act to drive improved ROI levels in both sales and marketing. Such solutions have components such as multi-dimension segmentation, a product recommendations engine, a CLTV model, a next purchase model, a churn and win-back model, store preference models, etc.

On the other hand, sales APIs are aimed at improving the leadership’s ability to predict sales, develop strategies and empower sales people through diverse sales algorithms. The parameters included are sales forecasting, lead scoring, opportunity scoring, product recommendation engines, etc. So, if there is a need for AI and ML tools to perform specific tasks, companies can license specific NAVIK MicroMods, and then customise and train the models to perform the same. MicroMods are able to perform standalone functions or can run on the NAVIK line of AI-powered solutions. The NAVIK MicroMods Library also includes marketing response APIs that learn from every campaign to improve ML algorithms for continuous ROI improvement, as well as integration APIs that help algorithms fit into existing infrastructure.


Please enter your comment!
Please enter your name here