Organisations often end up failing to achieve the results they want from artificial intelligence (AI) because they skip important key steps before embarking on their journey.
They rush into AI without first having a plan, and a multi-faceted one at that, said Jason Hardy, Hitachi Vantara’s CTO for AI.
While IT needs to be responsible for it, AI also must involve multiple teams within the organisation, including legal, human resources, and lines of business, Hardy said.
“It takes coordinated effort across those groups,” he said in a video interview with FutureCIO.
As CTO for AI, Hardy oversees Hitachi Vantara’s AI strategy and portfolio, including leading the direction of the tech vendor's AI platform, Hitachi iQ.
Some common challenges that enterprise customers face with regards to AI are figuring out where to start and what to do with it, Hardy noted.
They are overwhelmed by a technology that seemingly can do anything, but still have to prove that it can do so with actual use cases, he said.
In fact, just 23% of Southeast Asian businesses are transformative, including the ability to create new business models and products, due to their AI deployments, according to an IDC study commissioned by SAS.
The report attributes the lack of success primarily to poor quality data and privacy concerns, and notes that organisations are struggling with a lack of specialised skills and associated costs from AI development.
A whitepaper from Lumenalta further estimates that only 8% of businesses experience “extreme success” with their AI initiatives. The study also points to data quality and security as key challenges.
Importance of finding relevant use cases
To ensure success, Hardy noted that organisations need to first identify how AI can address problems in their business, with the current state of their data infrastructure. Other key components also need to be part of the plan, including establishing an ROI (returns on investment) and the security layer.
"They just go buy more GPUs and make the investments without [first] fully realising what needs to happen."
Hitachi Vantara aims to help its customers do these, he said, adding that if a proof-of-concept does not work, they learn from it and move on to the next one, improving on it with the learnings. They do this until they achieve the targeted ROI.
“Organisations often fail because they don’t do the first steps,” Hardy said. “They just go buy more GPUs (graphics processing units) and make the investments without [first] fully realising what needs to happen. They’re just throwing whatever [they can] against the wall and see what sticks.”
At the same time, organisations cannot afford to stand back and wait to see what works best, he said. This gives their competitors first-mover advantage and they will have to catch up later, he noted.
Enter too late into the game and there also may be supply chain issues to deal with, he added.
The good news is, 42% of businesses in Asia deem AI critical to their operations, above the global average of 37%, reveals research from Hitachi Vantara. Some 71% are hiring staff with AI-relevant skillsets, while 68% are working with external expertise.
In particular, there has been growing interest on horizontal or function-specific use cases, Hardy said. For instance, companies are looking to AI to support tasks in human resources and finance.
It is driving Hitachi Vantara to prioritise its AI innovation on enterprise applications as well as vertical industries, such as energy and manufacturing.
Building products that support new deployments
And it wants to helps customers deploy AI sustainably, said Hardy, who added that the wholly-owned subsidiary of Hitachi focuses its development efforts on building AI-ready data storage products that are as efficient as possible.
For instance, some AI processes can run locally on the device rather than at the data centre, depending on the data privacy and security control requirements. The technology, hence, needs to support processing capabilities on localised resources, such as laptops.
“Every piece we put into our iQ platform is done with intention,” Hardy said. “It’s really about building to accelerate the AI outcomes and get ROI from investments.”
This is increasingly important as businesses move into the age of AI inferencing and agentic AI, he noted.
Organisations will want to be able to scale and support these workflows, including AI finetuning and training, without taking on unnecessary infrastructure costs, he added.
“Spend the same level of energy on training people and users consuming AI."
By 2027, business leaders in the region will want at least an 80% success rate on their GenAI initiatives for higher efficiency and revenue, research firm IDC predicts. By then, organisations will aim for measurable success and strategic outcomes from their AI deployments.
With 2025 the “year of AI pivot”, organisations will need to integrate AI into their business strategies and move beyond isolated pilot projects towards measurable business outcomes, said Sandra Ng, IDC’s group vice president and general manager. This will require structured approaches, governance, quality data, and scalable fit-for-purpose infrastructure, Ng said.
As AI becomes more efficient, achieving better density and control as chipmakers improve on their products, companies will not need massive GPU systems to run some of their AI workflows, Hardy noted.
Instead, they have to focus on other key factors to ensure their deployments achieve the intended results.
“Spend the same level of energy on training people and users consuming AI,” Hardy said, pointing to skillsets such as prompt engineering, as well as the need to address issues such as hallucinations and model finetuning.
“They need to realise there’s a lot more to AI than just buying GPUs,” he reiterated. “Get trained on it and train your people on it. Then start small and scale high as you see success.”
Above all, organisations have to ensure they adopt explainable and ethical AI, he said, stressing the need for responsible AI adoption as the technology becomes more widespread.
