madgooch - stock.adobe.com
Modular AI can help enterprises apply and understand AI
Modular AI enables customers to deploy AI as it applies to individual use cases, creating a more focused, open, and cheaper overall AI system and strategy.
SAN FRANCISCO -- As enterprises move out of the AI experimentation stage and into deployment, they are increasingly turning to AI microservices and modular AI to tackle specific business needs.
For enterprises, relying on a single vendor or service for their AI use cases has begun to make less sense than it used to.
This is partly because larger enterprises are starting to experience vendor fatigue, said Traci Gusher, principal of data and analytics at KPMG.
Vendor lock-in refers to an organization's unintentional dependence on a vendor for all of its services, typically because the vendor will prevent its services from pairing with third-party services, or limit the compatibility of third-party services.
Some major cloud and technology vendors, such as IBM, for example, have opened their platforms to accommodate outside applications, but the problem still exits for some vendors.
By piecing together AI microservices or products from open vendors, enterprises can create a modular AI system that doesn't lock them into a single vendor or service.
This method can also enable enterprises to pick products or microservices that better fit their budgets, or find ones that are more technologically advanced, Gusher said.
"The ability to be more dynamic is something organizations strive for," she said in an interview at the AI Summit conference here.
Patching together microservices or products can also help organizations better tackle specific use cases -- a critical piece to the AI deployment puzzle.
For enterprise looking to begin deploying AI, it's crucial to develop a potential use case first, Dinesh Nirmal, vice president for development for data and AI at IBM.
"For some of the use cases, you probably don't even need AI," he said in an interview at the conference. Organizations should consult with technology providers, as well as speak with their own departments, to develop a use case before buying a platform or service.
Modularizing AI by taking on one use case at a time can help enterprises ensure they understand the technology they have added and the changes they have made, said Sanjay Srivastava, senior vice president and chief digital officer of professional services and AI vendor Genpact, told SearchEnterpriseAI at the AI Summit.
Dinesh NirmalVice president for development for data and AI, IBM
"Picking the right problems to solve is the largest issue in front of clients right now," Srivastava said. "AI cannot be the answer looking for the question. We must first ask the right questions."
Using modular AI can help enterprises understand how each AI system works, as it's broken down into separate chunks, rather than as a single encompassing platform.
If you design a system as a whole, it's less explainable, Srivastava explained.
Organizations can more easily "drill down" with a system broken up into chunks, he said.
After applying AI to one use case and judging how effective it is, enterprises can then build on that use case or apply AI to other cases, Srivastava said.
"This modulization is going to drive a significant expansion of AI in the enterprise," he said.
The AI Summit was held Sept. 25-26 at the Palace of Fine Arts.