Industry needs to distinguish between ‘Doing AI' and ‘Using AI'

Time of India

21 hours ago

According to a recent media report, healthcare conglomerate Johnson & Johnson has pivoted its AI strategy and wants to focus on only the highest-value GenAI use cases and shut down pilots that were redundant or underdelivering. This means a shift away from broad experimentation to a more focused approach so that projects that are redundant or simply not working, or where a technology other than GenAI works better are optimized.This is not unusual. In fact, it is in line with what is happening across the world. Many organizations tend to burn huge amounts of money and waste resources on experimentation. Johnson & Johnson is no exception.
Unprecedented hype around Gen AI
There is a huge hype around Gen AI across industries and sectors. The hype very often than not translates into unrealistic expectations which means that every department within large organizations and enterprises is vying to implement and showcase AI initiatives. This can lead to federated projects and initiatives across the organization, without yielding much value overall.
"This is a very practical approach to experimenting with new technology," says Jagdish Ramaswamy, Digital Transformation Advisor and former President and Chief Digital and Information Officer at Hindalco Industries Limited.
Difference between 'Doing AI' and 'Using AI'
This is a level headed approach and this is the difference between 'Doing AI' and 'Using AI' aver experts. While 'Doing AI' just needs investment in terms of tools and infrastructure along with educating the employees, on the other hand, 'Using AI' needs value measurement , change in decision making styles and most importantly, reconfiguring end-to-end process flows.
' 'Using AI' will need value measurement, change in decision making styles and most crucially - sensibly reconfiguring the end-to-end reimagined process flows. I find most of my corporate clients being fixated on the GenAI archetype - the Diagnostic, Prescriptive and Autonomous AI archetypes have to be appropriately used in conjunction," says Abhay Johorey, Managing Director, Protiviti India Member Firm and former Head of Digital Strategy at ICICI Bank.
Implementation needs to be based on use cases
Many organisations are focusing on the top 3-5 use cases with impact. Testing and learning can be a good indicator of tech capabilities.
"Many organisations are pivoting to focus on top 3-5 use cases with impact. Testings and learnings would be good to showcase capabilities and some initial tech feasibility. Beyond that Responsible AI and overall
AI governance
frameworks are discouraging the smaller pilots with the kind of intense pre-work required," says Sushma Chopra, Global Lead - Innovation CoE and Digital Evangelist at Mondelez International.
"Sensible step. Limited use cases will give the right ROI and prevent various associated overheads. Based on the learning curve, one can start adding more use cases. However, closed source models will not find space in enterprise due to data privacy issues. Open source Small LLMs have their own challenges in terms of localised training and fine tuning. But they will lead the race compared to closed source LLMs," says another tech leader, requesting anonymity.
Will efforts to centralize AI initiatives bear fruit?
Johnson & Johnson's centralization of governance and its approach to bringing all experiments under one roof-categorizing them into three buckets to track performance seems to be the most practical strategy for a number of technology experts.
"I wonder how much money has already been burned on PoCs alone. That's why it's crucial to channel all PoCs into a single program and build a pipeline. This allows us to fail fast, fail cheap, and ultimately identify the right use cases to scale," says Rajendra Deshpande, Partner, INGROUP Consulting Services and former CIO of Intelenet Global Services.
In conclusion
The bottom line is that centralizing governance, prioritizing the most impactful use cases, and eliminating the irrelevant ones is the right strategy towards implementation of a robust Gen AI strategy.