On-Premises Limitations, Cloud Possibilities: AI Strategies for Leaders

Taking advantage of more than 40 years of leadership in the enterprise applications space, many customers have run their businesses on the Oracle Applications Unlimited family—JD Edwards, Peoplesoft, Siebel, and Oracle E-Business Suite—or on competitors such as SAP for decades. But with the explosion of AI in recent years, many business leaders are trying to understand how to leverage AI within their legacy applications. Earlier this year, Forbes reported that 6 out of 10 large companies worldwide use generative AI. Of that group, 74% are seeing a significant return on investment, and 45% saw employee productivity double.¹ AI delivers real-world value and allows those harnessing its power to move much more quickly and efficiently. These productivity gains can be an equalizer for smaller organizations, helping them scale efficiently to take on competitors and disrupt markets. However, as AI adoption grows and these types of efficiencies become more commonplace, those who wait to adopt AI may unfortunately quickly fall behind their competitors.

Choosing a path to AI adoption

Leaders are being held accountable to their boards for adopting an AI strategy that will allow them to compete in today's AI-enabled market, and they need to move quickly to keep pace with the market and avoid disruption. While Oracle's Applications Unlimited program will remain supported until 2035, leaders must decide if adding AI to their legacy environments makes sense or if they should migrate to updated platforms with AI already embedded. For example, Oracle Fusion Cloud Applications offer a complete suite of modern best practices across finance, HR, supply chain, and CX, with more than 150 embedded AI capabilities. And Fusion Applications’ rapid pace of innovation and quarterly update cycle allow leaders to maximize their investment while executing their AI strategy within their core business processes.

Migration from legacy platforms to Oracle Fusion Applications may seem too daunting or complex for organizations with entrenched, highly customized systems, but building effective AI tools is much harder. Despite all the potential benefits of AI, many companies still struggle to execute their AI vision. A recent Wall Street Journal article found that, while many executives remain optimistic about AI, AI has been much more work than initially anticipated.² There are many AI services and tools in the marketplace. However, executing an effective AI strategy requires data science skills, development expertise, specialized infrastructure, deep business process knowledge, and high-quality trusted data. While most companies may have a few of these capabilities, most don’t have all the required capabilities to execute a successful AI strategy that delivers value.

Common AI challenges

As leaders dive deeper into their AI journeys, they’re facing many common challenges. Executives with multiple disparate systems or highly customized on-premises applications face an uphill battle to clean up their data to begin their AI journey. Assuming the data can be aggregated, training a model on historical data may seem easy at first, but business process changes, exceptions, and data set outliers can all affect AI accuracy. These factors can also lead to model drift the more the tools are used. Furthermore, AI expertise is in high demand, making it difficult to find, attract, and retain top AI talent. The deep understanding of business processes required for successful AI projects means that buy-in is required from line-of-business owners who may have differing visions for how the AI tools should operate.

Graphics processing units (GPUs) that run AI models have been impacted by supply chain shortages. GPUs require massive amounts of electricity, which can impact sustainability goals. Plus, running multiple GPUs in tandem to execute complex tasks requires lossless networking. All of these factors can lead to extremely high costs for implementing AI solutions.

In addition to the technical challenges of implementing AI, recent regulatory initiatives, such as the EU AI Act³ and Canada’s Artificial Intelligence and Data Act⁴, and existing privacy laws, such as the California Consumer Privacy Act⁵ and the EU’s General Data Protection Regulation⁶, all threaten regulatory risks if AI is implemented in a noncompliant fashion. The use of AI in hiring practices can also bring increased risk and scrutiny from regulators.⁷ As a result, IT executives are faced with the difficult challenge of executing their company’s AI strategy and delivering value within a rapidly changing landscape of technologies, talent, and regulations.

Building a successful AI strategy

Successful AI strategies cannot focus solely on technological solutions. Effective strategies will need to be multifaceted to gain the most value. Research published by the National Bureau of Economic Research indicates that the largest productivity gains in using AI will require the development of complementary processes, organizational restructuring, and adaptability as individuals use new AI tools.⁸

Effective AI strategies require a new way of thinking, such as building AI into core business processes instead of bolting AI onto existing processes. While GenAI tools can certainly provide value, organizations with highly customized legacy applications may struggle to realize transformational benefits by applying an AI tool to processes that were not originally designed with AI in mind.

It is also critical that executives focus on data privacy and security when selecting embedded AI solutions. Some AI vendors contractually obligate their users to allow anonymized data mining so the vendor can monetize company data and intellectual property. However, due to data variances, aggregating multiple sets of data generates only minimal benefits.

Additionally, it is imperative that AI strategies include a human in the loop of AI processes. This includes giving humans the ability to intentionally turn AI tools on, accept or reject recommendations, and modify AI-generated content. The NIST AI Risk Management Framework identifies that having humans in the loop of AI processes can help promote fair and equitable outcomes.⁹

Finally, AI strategies should also be flexible and future-proof. AI innovation is moving much more rapidly than the adoption of other, previous transformational technologies such as the internet and smartphones. Two years ago, generative AI exploded onto the scene, democratizing AI for the masses and allowing natural language interactions with data. A year later, tools such as retrieval-augmented generation (RAG) allowed AI to provide contextually relevant answers without needing to train custom models. Now, a new type of AI, agentic AI, promises to combine the power of machine learning, advanced analytics, and contextual generative AI to automate processes and deliver enhanced business value. The nexus of multiple AI technologies and disciplines, agentic AI will be a challenging endeavor for all except the most mature and sophisticated organizations. Unfortunately, due to this complexity, CIO magazine reports that an estimated 75% of companies that attempt to build AI agents on their own will fail.¹⁰

Therefore, with the rapid innovation and update cycles in software-as-a-service (SaaS) technologies, executives should consider a SaaS-first strategy with embedded AI to achieve quick and meaningful AI productivity gains. In short, due to the complexity of AI, don’t build what you can buy.

How Oracle can help

Among AI providers, Oracle is uniquely positioned to help executives implement effective AI strategies using built-in AI technology. Oracle continuously expands the capabilities of Fusion Applications by embedding a wide range of predictive and generative AI and AI agent functionalities directly into everyday process flows, democratizing AI so all end users can benefit without needing to be data scientists. Oracle AI Agent Studio for Fusion Applications is a design-time environment that provides tools that allow customers to create, customize, validate, and deploy GenAI features and AI agents to meet their specific needs. Unlike highly customized legacy applications, Fusion Applications’ unified data model provides consistent high-quality data, which is needed for effective AI usage. In addition, the computing power of Oracle Cloud Infrastructure (OCI) for running AI is included with Oracle Fusion Applications at no additional cost.

With the CIA as its first customer, security is in Oracle’s DNA. As a result, Oracle Fusion Applications’ AI does not allow customer data to be shared with Oracle or third-party LLM providers. With a human in the loop of AI tools and quarterly release cycles, Oracle Fusion Applications’ AI is human-centric, future-proof, and ready to be implemented out of the box.

With the rapid pace of innovation combined with the risk of falling behind, executives should look to Oracle Fusion Applications to execute their AI strategy. Oracle allows organizations to adopt modern best practices, leverage AI-powered analytics, and take advantage of agentic AI out of the box. Oracle Fusion Applications with embedded AI let organizations focus their AI efforts on organization-specific use cases. And rather than building AI tools and bolting them onto legacy applications, Oracle Fusion Applications provide leaders with a path to recognize the value of AI across their organization.

Next steps

If your organization is ready to take the next step with embedded AI in Oracle Fusion Applications, please reach out to us at any time through your Oracle Cloud Applications sales team.

For more information on AI in Fusion Applications, please refer to Oracle AI for Fusion Applications. And to find additional information on the underlying technology powering Oracle Cloud Applications, please check out our blog post series, starting with the introduction.

The author is a member of Oracle’s North American Applications Office of Technology and Innovation, which is dedicated to helping customers modernize their businesses through technical innovation. This team provides subject matter expertise and vision on AI, SaaS, platform technology, operations, and data management.