Artificial Intelligence & Machine Learning

AI and Cloud: CIOs' Take on Ethics, Adoption and Strategy

As organizations increasingly adopt artificial intelligence and cloud technologies, the race to integrate these innovations has sparked both excitement and caution among business leaders. At CIO.inc.'s Cloud and AI Innovation Summit, IT leaders discussed the transformative potential, as well as the challenges, of adopting AI and cloud. All agreed that ethical issues must be addressed and regulations are necessary for adopting AI and large language models.

In a panel titled "Aligning Cloud Strategy With AI Adoption: Are CIOs Cashing on the Opportunity?" Nitin A. Harne, senior director of enterprise IT and cloud transformation, Capgemini; Dr. Jai Ganesh, chief product officer, HARMAN International; and Satyavathi Divadari, executive director of cloud security services at Wells Fargo, who is also a CyberEdBoard member, shared their AI journeys and highlighted application areas in their respective industries.

AI has penetrated nearly every industry, and its potential for improving decision-making, driving efficiencies, boosting productivity and introducing innovation has been proven. Organizations are moving beyond the exploratory and proof-of-concept stages to adopt AI in their pursuit to become AI-led businesses.

A recent global survey by Google Cloud found that generative AI is a key driver of business transformation. By incorporating AI into their operations, companies can achieve substantial financial returns, improve efficiency and sustain growth. One key finding is that 74% of enterprises using gen AI reported ROI within the first year, and 86% of those reported increased revenue of 6% or more. Yet many organizations are still in the exploratory stage, trying to understand which business areas will benefit from AI.

See: Report: Generative AI Is a Key Driver for Business Growth

AI Journeys

Within the financial services sector, Wells Fargo is exploring the application of AI in areas such as fraud prevention, threat remediation, early threat detection and remediation.

The company is exploring how it can use gen AI models to improve the developer experience. They also want to use AI for secure development (DevSecOps), Divadari said.

Practitioners agreed that organizations must create more awareness around AI and its responsible use. Organizations also need to implement guardrails and enforce governance to prevent misuse.

"There is a lot of drive for AI, especially its ethical use," Harne said. Capgemini's learning and development department wants to introduce a certain level of maturity and awareness for the technology internally.

"AI is a vast topic, and, as an organization, we need a lot of maturity. We want to understand what can be applied to our business, and we will know that maybe six to nine months from now. We have advanced in some areas, especially those critical for a typical service provider, where we need to minimize the cycle times for operations," he said.

Dr. Ganesh is deeply involved with HARMAN's digital transformation solutions business, which offers consulting and ER&D services for customers. He also oversees AI implementation across his company.

"AI adoption has become a boardroom conversation because of the wide-ranging impact that the technology has across various business functions, use cases and processes," he said. "While no CIO wants to be left behind, they are also prudent about their AI adoption journeys and how they implement the technology for business in a responsible manner."

While there are many business use cases, enterprises are prioritizing these "on a must-have immediately to implement" basis.

AI Adoption Challenges

AI adoption is often accompanied by barriers that include ethics, cost, model accuracy and regulation.

"The ethical aspects, what we call 'responsible AI', are critical because the explainability of AI models is a challenge," Dr. Ganesh said. "This is more so with LLMs, which are extremely complex. They work on token prediction, and we are unable to explain why a model took a certain decision."

The accuracy of AI is questionable at this stage, and an incorrect decision could result in extra expenditure or business losses. "That is why having a human in the loop is important," he said.

Cost is another reason that holds organizations back from adopting AI. Convincing management for budgets becomes a challenge if a CIO cannot articulate the business benefit or define the business outcomes.

"It is an expensive affair, especially from a data mining perspective. You can be focused if you are able to define the business outcome and what business problem AI will solve," Harne said.

The accuracy of output is also a concern, and this is dependent, to a large extent, on the dataset used for training the model.

"You run all these use cases through the AI model and then you get an accuracy of 60% to 68%. So it's like tossing a coin, and there are 50-50 chances of getting a right or a wrong. AI is not useful unless you get 99.999% of accuracy," Divadari said. "The second aspect is to consider if you really need to spend so much on an LLM."

There is also a need for AI-specific regulations and frameworks. With the exception of the EU, which has its AI Act, other countries are in the process of drafting their AI bills or creating guidance for usage.

"Discussions are ongoing in the U.S. and India [regarding regulation]. The Indian government is seriously considering interventions regarding AI-driven decisions. So, we see a scenario in the future where there won't be one model for everyone or one global model that will fit all the requirements," Dr. Ganesh said. "You need to create custom models according to how the governments in each country are specifying how data should be treated, how the model should work and where human intervention is needed. That is going to be an extra layer of complexity and cost in this overall scheme of things."