By Gajen Kandiah
Aloof to the hype and sometimes derision of skeptics, the developers and architects of artificial intelligence (AI) and generative AI (GenAI) continue to scale their innovations to new heights, pushing the boundaries of possibility ever further.
Yet, the rapid advance is also confounding organizations, many of which lack the expertise and/or resources to stay on top of the latest innovations and move successfully from experimentation to production.
A recent Gartner, Inc., survey of more than 800 business leaders predicted that at least 30% of GenAI projects will be abandoned after proof of concepts by the end of 2025 due to things like, “poor data quality, inadequate risk controls, escalating costs or unclear business value.”
And Peter Bendor-Samuel, CEO and founder of Everest Group, estimates that 90% of GenAI pilots will not move into production “any time soon, and some may never do so.” He cites as reasons everything from GenAI pilot “fatigue” all the way to difficulty in identifying a return on investment.
At Hitachi, our AI research goes back decades. And in the spirit of knowledge sharing, last year we began deliberately chronicling our work in GenAI to expose our decision-making processes. We’ve also published stories of customers in almost every industry, from energy and rail, to manufacturing and finance, that have been aided by Hitachi’s AI expertise and innovation.
But the business world’s reaction to GenAI, in particular, over the past year or so, caused us to look deeper into the dynamics around adoption rates. We wondered if there were common traits among pilot failures and abandonment. We wondered if there were technical solutions to the challenges.
After innumerable customer conversations and internal discussions, it was clear the issues were more than technical. A list of best practices wasn’t going to be enough. The reasons for stalled pilots ran deeper and began earlier than when the first software tool was downloaded.
The deliberation led us to craft an internal “position paper” on AI and GenAI that would help keep our Hitachi digital companies focused on advancing internal, as well as external projects. In it, we outlined three thoughtful and overarching principles that respond to the challenges we’ve seen so many organizations face.
So, in that same spirit of knowledge sharing, I’d like to present them here and ask for your feedback.
First, our position. AI and GenAI are as transformative as any technical innovation in history; harnessing their power is critical to the success and growth of organizations across the globe. When managed deliberately, they can open a world of possibilities and positively alter the trajectory of entire industries to advance everything from climate action to the energy transition.
Next, our guiding principles. To manage deliberately, AI and GenAI require an approach based on 1) thoughtful consideration of objectives and outcomes; 2) the selection/development of AI systems that are purpose-built for your industry and your data; and 3) an intrinsic commitment to responsibleAI.
Let’s look at each a little more closely.
1. Outcome-Oriented
When scoping out AI projects, it is just as critical to document your objectives and what your optimal outcomes might look like, as it is to understand the potential challenges and opportunities. These objectives become guardrails and help maintain focus and can aid in measuring performance as well as returns on investment.
2. Purpose-Built
The amount of AI and GenAI tools, applications, and services available today is exponentially larger than even 12 months ago. The surge has made it clear that all AI is not created equal. What may work in manufacturing, may not work in retail. From our own research and development, we are keenly aware of the value of purpose-built AI and the creation of industry-specific tools and accelerators, crafted for use within specific industries and for specific types of data. Adopting a purpose-built approach will ensure that your AI and GenAI projects can be created deliberately for optimal outcomes and measurable results. (Check out R2O2 here.)
3. Responsible & Reliable
The concepts of responsible and reliable AI are inextricable – the more attention placed on creating responsible AI, the more reliable the outcomes will be. What’s this mean in practical terms? As an example, providing customers guardrails for their machine learning (ML) and large language models (LLM) can minimize the risk of prompt and model response bias, toxicity in outcomes, and even security threats. When these things can be achieved, organizations can begin to trust that their outcomes are reliably more accurate.
This article is but an abbreviated version of a more extensive paper and practice. In it, for example, we expand further on the value of data, the lifeblood of both AI and GenAI. As LLMs rely on extensive datasets to acquire context, patterns, and knowledge, the importance of carefully curated data is clear. What’s needed is a data strategy, robust data governance, and a well-defined reference architecture to scaling GenAI use cases effectively. Companies with access to proprietary, high-quality data in areas such as manufacturing, supply chain, operations, and maintenance gain a significant competitive advantage, enabling them to leverage GenAI to drive innovation and operational excellence.
Again, this article is a glimpse into our thinking, but I think it’s worth sharing. And one more thing – as you consider or are already in the throes of AI, remember, this is not a trend. The technologies in this area will continue to advance and permeate every aspect of business and society in both small and big ways long into the future. It is key to partner with companies that understand your challenges, as well as your objectives. And it is critical to engage and adopt, deliberately. The principles above will get you off to a good start or course correct along the way.
Gajen Kandiah is President and COO, Hitachi Digital, and Executive Chairman, Hitachi Digital Services.
This story first appeared as an article on Gajen Kandiah’s LinkedIn page: (4) Taking a Position on AI and GenAI | LinkedIn.
(Cover image: ©Marco Bottigelli via Getty Images.)