Lessons from AI experiments at Prosus

Our team began operations about six years ago, in 2018, coinciding with the group's significant investments in artificial intelligence. We operate as a federation, collaborating closely with all group companies globally, facilitating knowledge sharing and expertise exchange.

Currently, across Prosus’s group companies, there are nearly 1,000 professionals. Our AI investments are directed inward, fostering growth, acquiring talent, and enhancing tools for our companies, including those in India.

Q. Give us a few examples of AI startups in your group

One notable example is Corti, a US-based health care company that aids practitioners in integrating AI into their workflows. Their services range from coding patient notes for insurance purposes to providing guidance in emergency scenarios.

Another venture, Martian, focuses on simplifying AI model selection, crucial for navigating the multitude of available models efficiently. Ema, originally based in Bengaluru (now in San Francisco), offers enterprise assistance tools tailored to specific needs.

BandLab, a Singaporean music creation platform, exemplifies successful integration of AI into music production. These examples showcase a diverse investment spectrum, from early-stage startups to companies at market growth stages, where our assistance aids expansion.

Also read: 91 percent Indian corporate leaders believe their companies need to adopt AI to stay competitive: Report

Q. When you released your own AI assistant within Prosus, any interesting productivity lessons?

We gauge productivity through various metrics, including quantitative measures. For tasks like coding and product development, we compare output before and after implementing AI. Productivity gains, measured in time saved or task completion rates, typically range between 40 percent to 60 percent, sometimes even higher. Notably, these gains vary across tasks, with some experiencing up to a 25 percent boost.

Despite differences, even a 15 percent increase in productivity across an organisation is substantial. While these findings may not universally apply, within our tech-centric domain, such improvements are significant and underscore the transformative impact of AI on our operations.

We implemented our AI assistant differently, integrating it within team environments like Slack, fostering collective learning and collaboration. This approach not only encouraged shared discovery, but also facilitated social quality control, enhancing both individual and team productivity. With instant access to assistance, individuals could resolve queries efficiently, minimising delays and optimising organisational velocity.

Surprisingly, user engagement proved sticky, with over half utilising the assistant weekly and a quarter daily. Users reported increased freedom and independence, enabling them to tackle tasks beyond their usual scope. This bottom-up approach to experimentation revealed unexpected usage patterns, prompting a shift towards collective learning and humility in assuming outcomes.

Our continuous feedback loop ensures adaptation to emerging needs and challenges, emphasising the value of organic discovery over preconceived notions.