Business leaders need to move a more probabilistic approach


Decision intelligence is a business application of artificial intelligence to the decision-making process in the areas of a business. Decision Intelligence is being used by organizations to optimize and enhance corporate performance. We interacted with Atul Sharma, Co-Founder and CTO, Peak, an AI-powered Decision Intelligence Company, to find out more about Decision Intelligence and the businesses’ transition to it.

TimesTech: How is decision intelligence changing decision-making?

Atul: Decision Intelligence (DI) will change the way we work and run our businesses. Perhaps the biggest change long-term will be for leaders and decision-makers themselves. As we move toward a world where Decision Intelligence is the norm, then business leaders will need to move from a deterministic management style to a more probabilistic approach. They will no longer be asking questions like, “Who made that decision, are they qualified?” instead they’ll be asking, “If the model thinks that highly unlikely, why aren’t you recommending this?”

TimesTech: How does AI help in building the right data strategy for businesses?

Atul: It’s regressive to think about a data strategy without making provisions for AI – even if your business isn’t ready for AI, yet. One day you will need to facilitate its adoption, and not considering it now puts a business in serious jeopardy of having to retrofit infrastructure further down the line.

AI is far more focused on business outcomes than traditional data technologies. Building a data strategy with it in mind helps to ensure a strategy that goes beyond data cleaning or Business Intelligence reporting and drives meaningful value for a business.

TimesTech: Do you think companies in India are ready for this transition of adopting decision intelligence?

Atul: I think Indian businesses have an advantage. A lot of our enterprises reached maturity after their counterparts in Europe and the US and, as a result, many are digital-first. They have modern tech stacks, whereas their counterparts in other countries are held back by outdated, legacy tech. Free of that hindrance, Indian companies could very well lead the world when it comes to the adoption of Decision Intelligence.

We recently released some research, which showed that Indian businesses were more advanced than those in the UK and US when it comes to adopting AI.

TimesTech: How do you see businesses transition in the intelligence era?

Atul: This might be the intelligence era, but the vast majority of companies have yet to tap into its potential. Businesses like Google and Netflix are the exception rather than the rule, most are still struggling to drive meaningful value with AI. Only around a quarter of AI models built commercially are ever actually deployed, and even fewer generate meaningful value for businesses.

Commercial AI has the last mile problem. It’s caused by a number of factors – siloed data within a business, complicated tech stacks and a lack of buy-in from end-users. But perhaps the biggest is the tendency of businesses to do “data science projects”, i.e., to start with the data and see what they can produce. The end result rarely has real utility for a business.

But we’re starting to overcome problems like this and are already seeing green shoots with platforms that simplify the tech stack needed to deploy AI and unite both technical and commercial users. This is important, if commercial teams – end-users – are part of a project from the start, then they can guide it to ensure the end product delivers on a necessary outcome.

Once we overcome this last mile, and businesses routinely encourage both technical and commercial users to work together developing AI applications, then the innovations will come thick and fast – and we’ll start to reap the benefits of the intelligence era.

TimesTech: Why does successful AI require the right data structure?

Atul: AI is a data technology. Machine learning models are trained on data and need well-organised and labeled datasets to maximise operational efficiency. It is also crucial that models – even if they are only intended to optimise one or two core functions – are trained on data that is representative of the entire business. A model that only has data from marketing, for example, would be able to optimise marketing activities, but will not be aware of any influencing factors outside of marketing. Similarly, it may optimise marketing at the expense of another department, because it has no visibility of the wider business as a whole.

So, architecture is essential to the success of any AI project and should ensure organisational data is centralised, well organised, and clearly labeled.