By Naveen Yeri, SVP and Head of Enterprise Analytics and Data Science, Wells Fargo India and Philippines
Effective data-driven decision-making is key to an organisation’s growth and operations
strategy. Mushrooming data, multiplied by the variances in the behavior of decision-makers, emphasises the need for two strong pillars—Business intelligence (BI) and Artificial Intelligence (AI)—to work together to support decision-making.
What is AI and BI? How can they be synergized to deliver outcomes?
Business Intelligence combines business analytics, data mining, visualisation, tools, and
infrastructure, transforming data into actionable insights to make intelligent data-driven
decisions. However, it often stops at the ‘what’ and ‘why’, providing descriptive or diagnostic views in hindsight.
Artificial Intelligence supports situational decision-making, creating a more diagnostic,
predictive, or prescriptive approach. AI encompasses technology and systems that mimic
human-like intelligence through continuous machine learning. AI drives decision intelligence, allowing businesses to automatically lead to decisions and actions based on machine learning and augmented analytics-based insights.
While the definitions would appear so, the two are neither mutually exclusive nor should they operate in silos.
Why are businesses not often successful in leveraging AI and BI effectively?
The exponential growth in data has led to more dashboards and more AI, built as disjointed
analyses on the same data, not complementing each other. The insights are either not in sync or the right insights don’t make it into the right hands in time. On its own, BI has matured significantly with tools that clean, visualize and mine data into insights, and AI has massively improved the ability to predict events and model behaviors. Yet, we seldom find solutions that provide an end-to-end view bringing the ‘what’, ‘why’, ‘so what’, and ‘what next’ together.
Can AI-BI converge and complement each other?
It would require AI to empower BI and for BI to accelerate the build, adoption, and explainability of AI solutions.
AI can help expand a BI dashboard by seamlessly blending structured and unstructured data to enhance insights, extract key information, and extrapolate future scenarios. It can help create intelligent dashboards that provide mobile, interactive (voice command, natural-language querying, chatbot serviced) discovery of insights at the speed of thought—almost a virtual version of Hans Rosling embedded in your dashboard interface—by leveraging Natural Language Processing, Natural Language Interface and machine learning to generate more actionable insights than observational. Another example is behavioral decision-making, where key latent factors are not readily available from a BI-only solution.
Take for example a sales representative aiming to convert a lead. Wouldn’t it be empowering for her to have observable information on the customer supplemented with a dynamic propensity score that adjusts per the conversation, leveraging speech analytics? Additionally, if the most important latent factors contributing to this change in the score are available, the actionability would change dramatically.
Similar examples can be found in contact center operations, identification of data anomalies, generating multi-lingual insights, and building on-demand visualisation.
BI can democratize AI, an example being the provisioning of real-time monitoring and suggestive degradation in the performance of AI systems. While common in tech-heavy analytic shops, this is still not a norm in the larger industry. Add explainability, to show why and how the model suggests what it does, and the adoption is likely to increase many-fold. BI makes AI understandable, hence more ‘available’ for use. Creating tools for faster and more exhaustive data visualisation via templatised, self-service data analysis, and identifying relevant pockets of data to zoom into, makes modeling easier. Integrating a BI layer into the AI system makes hindsight, insight, and foresight converge into a single interface. Definitely not as easy as it sounds, but it can be a great utility.
In the age of real-time data and cognitive insights, which the Harvard Business Review describes as ‘analytics on steroids’, architectures such as edge technology, and integrated API-first headless BI, are key to ensuring the impact of these intelligent systems is exposed to the endpoint—a consumer, a machine or another system—in real-time, rather than a trickle-down strategy. Virtual assistants, ‘intelligent’ apps on smartphones, smart homes, utilities systems, and production centers, all effortlessly leverage these concepts.
So, where do we begin, and what’s next?
In some ways, this future is already here and the question we should be asking is, how do we embrace it? Using AI and BI in conjunction with strategic goals is key to business success. According to experts, often, a lot of the business analysis and results of the AI systems are consumed by a small sub-section of the workforce. The penetration of AI in the larger workforce’s decision-making is either low or siloed, implying limited visibility into data, analytics, and insight. If organisations are to truly unleash the power that AI-BI can bring to decision intelligence, they need to find more opportunities where both can converge to serve a common purpose and at the same time make it accessible to larger workforces.