Generating ground breaking returns, with generative AI

Much has been written about the potential of artificial intelligence (AI) to help companies unlock productivity and generate greater business value.

The challenge lies in scaling AI and effectively implementing its use in daily operations to benefit businesses. Developing new AI models – the algorithms used to train computers to analyse data in the same way humans do – for each new use case has proven time-consuming and costly. However, a new era of AI in business is emerging with generative AI’s foundation models. These models are trained on large amounts of unlabelled data and can be adapted to new scenarios and use cases, making them potential game-changers.

According to Ms Agnes Heftberger, general manager and technology leader of ASEANZK at International Business Machine Corporation (IBM), foundation models offer a significant increase in return on investment and a faster time to market for corporate users compared to traditional AI models.

“Foundation models amortise the initial work of AI model-building each time it is used, as the data requirements for fine-tuning additional models built on the foundation model are much lower,” she explained.

In the next two years, IBM expects foundation models to power about one-third of AI within enterprise environments. Early work with clients has shown that time to value – the time it takes for new customers to derive value from a product or service – can be up to 70 per cent faster with foundation models than with traditional AI approaches.

For instance, Moderna is working with IBM’s scientists to apply an AI foundation model named MoLFormer. Powered by IBM’s open AI platform, watsonX, MoLFormer helps scientists predict a molecule’s properties and understand the characteristics of potential mRNA medicines. This enables the design of medicines with optimal safety and performance.

“Instead of treating AI as a tactical add-on, enterprises will now be empowered to put AI to work at the strategic core of their business,” said Ms Heftberger. “The flexibility and scalability of foundation models will significantly accelerate AI adoption.”

THE INFORMATION BACKBONE FOR AI

Before businesses can fully integrate AI into their strategic core, they must establish AI-ready data architectures. Ms Heftberger likens this architecture to an information backbone for the brain. “AI requires machine learning, machine learning requires analytics, and analytics requires the right data and information architecture,” she said.

One obstacle to building this backbone is that business data is often widely distributed across various locations, including on-premise data centres, mainframes, as well as private and public clouds. To scale AI efforts, organisations need an open, hybrid architecture that can access information from both on-premise and cloud sources.

“Most companies already have all the data they can handle, but it may be stored in silos, have unknown quality, or be subject to access controls and regulation,” said Ms Heftberger. “For this reason, many of our customers start with data access and integration, which is important in feeding curated content to AI systems.”

This first step involves establishing a data fabric. This connects data from multiple locations, creates an access layer, and governs information flow across various technologies, storage systems and formats. “The data fabric allows clients to catalogue and understand all enterprise data required for model-training tasks,” said Ms Heftberger.

To scale analytics and AI workloads effectively, many organisations are also adopting data lakehouses that combine the strengths of data warehouses and data lakes. Data warehouses archive structured data for specific business intelligence purposes, while data lakes store unstructured data in its raw format for various uses.

Data lakehouses can apply the structures used in a data warehouse to the unstructured data found in a data lake. This allows users to access and utilise information for business purposes more swiftly.

IBM’s watsonX platform, which is already available, includes watsonx.data, a data store built on an open data lakehouse architecture. It enables organisations to leverage multiple query engines to run governed workloads regardless of their location, optimising resource utilisation and reducing costs.

“With built-in governance, security and automation, users can start working within minutes,” said Ms Heftberger. “Watsonx.data helps organisations to cut data storage costs and provides flexible information access with the required performance for AI demands. Through workload optimisation, organisations can reduce data warehouse costs by up to 50 per cent.”

BUILDING AN AI STRATEGY FOUNDED ON GOOD GOVERNANCE

While Ms Heftberger is bullish on AI’s potential, she believes that good governance is necessary for its successful deployment.

“Possibilities that we are only beginning to imagine will become commonplace, and new technologies will lead to entirely new types of work,” she said. “But to fully realise its potential, AI must be built on a foundation of trust and transparency.”

IBM has identified five fundamental properties for creating trustworthy AI: Explainability, fairness, robustness, transparency and privacy. The company’s development of AI technologies is guided by three core beliefs: AI should augment, not replace, human decision-making; data and insights belong to their creator; and technology must always be transparent and explainable.

“Businesses that adhere to ethical principles in developing and using AI today will be better positioned for compliance with impending regulations and potentially avoid the cost of redesigning or recreating models that were not developed with AI ethical principles and human values in mind,” said Ms Heftberger.

In line with these principles, IBM is launching watsonx.governance in October. It will help companies direct, manage and monitor their AI activities, ensuring increased responsibility, transparency and explainability in data and AI workflows.

CONSIDERATIONS FOR AI STRATEGY

In addition to advancing trustworthy AI, Ms Heftberger recommends that companies address a few other critical considerations when formulating their AI strategy.

The first is the need to create a competitive edge. “It is essential to let the business strategy guide the data strategy,” she emphasised. “To be truly impactful, AI should integrate into existing workflows and systems, automating key processes across areas such as customer service, supply chain management and cybersecurity.”

Secondly, companies should focus on scaling AI across their organisation. It is important to identify the right data sets from the beginning and build AI-ready architectures capable of harnessing them effectively.

“To successfully scale AI efforts, you need the ability to make use of all your data, wherever it resides,” said Ms Heftberger. “A hybrid cloud architecture provides the data foundation for extending AI deep into the business.”

Finally, finding the right partner on the AI journey is crucial. Said Ms Heftberger: “IBM is here to work with businesses to determine the most effective ways of implementing AI within their enterprise, and the approaches they can take to make this a reality.”

Learn more about how scaling AI across your business can create a competitive edge.