Synthetic data cuts costs and time

In today’s business context, it is now a reality that AI designed to be utilised in the office or at home is sophisticated, comprehensive, fast enough to navigate in business decision-making, and still provides an interaction to that of a human.

From fast-tracking drug development in healthcare, simulating fraudulent transactions in financial services, and fuelling autonomous vehicle tests in the automotive sector, it is already demonstrating its value across various business contexts.

Traditionally, businesses have been collecting information from various sources, storing and analysing to support business verdicts. This approach will soon be obsolete as AI creates synthetic data that closely mimics the real world.

In the retail banking sector, banks and financial institutions develop strategies to target customer segments based on their income and assets. Researchers can now create AI-powered persona bots that simulate, for example, the affluent customer segment. The bot can engage in real-time conversations with department heads day and night, discussing new products and providing recommendations along with specific insights.

At the same time, any department can easily access the bot to facilitate other operations of the bank, such as new services or strategy planning. All stakeholders now can chat with digital customers, using natural communication language as if they are real, in individuals or groups, and receive unlimited answers about lifestyles, behaviours, and preferences.

Additionally, organisations have a new way to distribute their knowledge and experiences to employees and even customers through synthetic experts. For years, professional service companies relied on traditional methods and consultants to deliver projects, which is sometimes cumbersome. In the future, synthetic experts could be trained on real-world data, in a text-based and numerical-based nature. These AI-powered experts can be leased to both large corporates to small enterprises to provide the professional service.

An application already in place for synthetic data is product testing. While data collection is getting faster and less expensive each year, there is still more room for savings in time and money by using synthetic data from AI. For instance, a healthcare company wants to find 300 female customers to perform research about three new products. Now, the company only needs to find 150 female customers and the remaining data is created by AI. This approach helps increase the sample sizes and obtain statistical significant data.

It is evident that an AI model needs quality inputs, data, and training. AI alone cannot capture the emotions, expectations, or the context that humans experience. Synthetic data also comes with inherent trade-off, as it cannot fully match the accuracy of real-world data. It can fail if the human sample is not representative. It means that AI cannot work like magic and appear from nowhere; it is built from the base of a reliable dataset.

The promise of being able to generate synthetic data, at will and at scale, is extremely attractive. Public sentiments on such data, however, are quite polarised. Companies may adopt a wait-and-see approach and are hesitant to use it for the time being. One thing for sure is that synthetic and real-world data are both needed, and the accuracy of synthetic data requires investment.

With a proper approach, synthetic data can be powerful, such as with product testing. It can boost the speed and depth of customer understanding, making it ideal for resource-intensive areas by reducing costs, saving time, and improving engagement for internal stakeholders.