Navigating the operationalization of GenAI

The discussions around GenAI are shifting. For more than a year, we’ve had noise and overhyped expectations. Consumer use cases and a lack of understanding of the basics of AI prompted many misguided beliefs. Yet, enterprise adoption of GenAI is predominantly contingent on integration and governance issues. Therefore, we must reinforce that AI is not a shortcut for transformation. Instead, enterprises must continue to operationalize cloud, data, and AI. The GenAI Meets Cloud discussion with enterprise leaders provided a temperature check for where organizations are on their transformation journey.

In a roundtable in London co-hosted by Hitachi Digital Services, HFS convened executive leaders from diverse industries, including finance, insurance, pharmaceuticals, and consumer packaged goods. Participants represented Actis, Ardagh Group, Associated British Foods, BAT, Citi, Deutsche Bank, Experian, Fidelity International, HSBC, Innocent Drinks, Legal and General, Lloyds Banking Group, LSEG, Marks & Spencer, Reckitt, and UBS.

The highly engaged discussion demonstrated the nascency of GenAI adoption. All participants are still trying to figure out the basics without getting carried away by the hype. The following highlights key insights from the discussion, reinforcing the dire need for a grown-up conversation.

Beware of the hype, and stay focused on the fundamentals

HFS analysts opened the discussion by providing the outside-in market perspective and food for thought, even though this highly motivated group did not need much of a push to get going. The participants were reminded to look beyond the consumer use cases dominating the discussions and focus on integration and governance as the critical issues for driving (or inhibiting) enterprise adoption of GenAI.

While HFS sees a new S-curve of value creation through technology arbitrage with GenAI as the critical lever, the expectation of 30% to 70% productivity gains is often aspirational and not a reflection of experiences with deploying GenAI in production environments. Exhibit 1 outlines the critical issues HFS called out that enterprises need to address for adoption.

Exhibit 1: Enterprise adoption of GenAI is about integration and governance

Source: HFS Research, 2024

Cutting through the market noise: The GenAI adoption market remains nascent as enterprises try to figure out the basics

To a large degree, putting these arguments to the participants was preaching to the choir. Sanjay Patel, former CIO at Tate & Lyle, succinctly summarized the challenges for operations leaders:

I think the expectation is quite high. Here is an evolution in technology where non-technology board members and executives almost demand the technology leadership to make this capability available. So, there is certainly a pressure point. Consequently, there is a rush to get stuff done. As technology leaders or business leaders, we try to find the fastest way of getting there. The tech is there. Large language models are available. They are available free. The environment is there—the infrastructure—yet we have providers create an environment within our firewall, which not everyone has done, with that capability built in. But then it comes to getting the data you want to leverage there.

Stefana Brown, Group Technology and Data Risk Director at Legal & General, built on these points and crystallized what needs to be done to operationalize GenAI:

My responsibility is to ensure that our technology and data risk frameworks are embedded across the organization. Firstly, it is obviously defined; secondly, it is embedded, and then it is complied with. I think the biggest challenge with operationalizing GenAI is understanding where the real benefit is from generative AI. Generative AI comes with several costs, complexities, and risks. So, know where you can maximize the advantage in the operational areas and how easy it will be to incorporate the generative AI as part of your everyday process.

From a provider perspective, Rajesh Rajappan, SVP & Digital Transformation Executive at Hitachi Digital Services, pointed out that AI is ultimately all about automation:

We are focused on the convergence of IT and OT technologies. This means generative AI is starting to play a big part in further pushing the envelope, as we have already been implementing AI for automation across information and operational technologies.
With the adoption of cloud we are seeing across the board, we are seeing customers looking at generative AI as a means of automation but also looking at what is the best way to implement it—from a cost standpoint, from a governance perspective, what data is there, and, most importantly, the value that AI can provide—and so we are pretty excited to be able to help our customers playing at the convergence, as I mentioned about across IT and OT technologies.

AI developments to look out for in 2024

To close out the session, HFS analysts looked into their crystal ball to share with the audience which AI developments to look out for in 2024:

• Anticipate the rise of large multi-modal models: Large multi-modal modals (LMMs) are a stepping stone toward truly general AI assistance.
• Don’t expect a breakthrough via GPT 5 and its brethren: LLMs will remain intrinsically limited and prone to hallucinations.
• Chipsets beyond Nvidia’s GPUs will emerge: The demand for chips has spurred a shortage that no single company can manage alone. Hyperscalers’ moves to invest in chips are welcome and will help reduce chip dependency, spur innovation, and advance their LLMs.
• Say hello to small language models: Improved quantization will drive on-device integration. Proliferation will cause even more demand.
• Move beyond lip service on governance: Bring together juggernauts and startups with equal opportunities. Facilitate transactions, connect buyers with sellers, and provide a platform for businesses to market their offerings.

The Bottom Line: To progress with GenAI, we have to cut through the market noise and address the integration and governance issues that operations leaders face.

To summarize, here are three key takeaways from this immensely engaging discussion:

• GenAI adoption remains nascent. Despite the noise, especially on social media, there is very little in production. Everybody is trying to figure out the basics.
• We need to clearly differentiate between productivity and innovation, such as with business model transformation. As we have seen with the cloud, the cookie crumbles when technology and business objectives don’t align.
• And, most importantly, change is hard. Operation leaders must establish how they can drive the cultural shift to harness AI’s full potential. This also means they must answer the broader and stricter question of what kind of organization they want to be.

Buoyed by this discussion, we would love to hear your views on where the industry is and what needs to be done to progress the adoption of GenAI!