AI data strategy crucial for faster return on investments: IBM exec

IBM India's AI head says most Gen AI projects fail due to weak data foundations and suggests a phased approach to unlock RoI within 18-24 months

IBM is bullish about enterprises seeing a comprehensive return on their artificial intelligence (AI) and generative AI (Gen AI) investments in 18-24 months if they adopt a proper data strategy.

 

Streamlining enterprise data, having an AI policy, reskilling people, and cultural transformation have been cited by experts as key tenets to be an AI-first organisation. All these involve a huge amount of heavy lifting, and that’s why Gen AI adoption continues to lag market expectations. AI is only as good as the data it is trained on.

 

“You have to find the right use case, the right model, and your ability to drive returns on the smaller use cases is pretty much prompt, especially for siloed use cases. But as you lay this foundation and the more elaborate and comprehensive you make it, the longer it is going to take for you to see comprehensive returns. I would say it would range anywhere between nine and 24 months,” Siddesh Naik, country leader, Data & AI Software, IBM India & South Asia, told Business Standard.

 

Almost 80 per cent of the time is spent on creating the data foundation or the data prep with only 20 per cent going into the AI part of the journey, he added.

 

The foundation layer has been the missing link for many enterprises. It should be started small and scaled efficiently to reduce complexity of data consumption.  

 

Data strategy, Naik said, is a complex task which involves data collection from multiple sources, establishing its quality, transforming it into a standardised format, enforcing policies, and data lineage.

 

He added, “Enterprises must first establish their overall data blueprint. Achieving this cannot happen overnight; each component requires substantial effort. Organisations should adopt a phased approach: first focusing on data collection and quality, followed by data pipelines and extract, transform, load (ETL) processes, then integrating a lakehouse for reporting, and eventually implementing governance and lineage mechanisms. A step-by-step approach ensures a sustainable and scalable data strategy for enterprise.”

 

A report by Boston Consulting Group (BCG) last year had said that key factors for scaling AI are largely people-and process-related. These include change management, product development, workflow optimisation, AI talent, and governance. Critical technology capabilities include data quality and management, while AI model quality and performance stand out as the top algorithm priority.

 

According to Naik, initially people were enamoured by AI and this led to massive investment into doing pilots and exploratory works. That was the hype cycle where companies jumped into AI proof of concepts (PoCs) and pilots.

 

Wipro chief technology officer (CTO) Sandhya Arun had said earlier this month that without proper policies, most PoCs will die in the laboratory or as a hobby project.

 

He added, “People will say good things but it will have no business value.”

 

“But what people started realising is that do I have a foundation of data that can really run this and help me productionise it? That is where the shift started happening — towards investing in data engineering and data foundation to create a strong base. I think the first step in all this is for the management and top leadership to understand, acknowledge, and commit investments towards building a data foundation,” said Naik.