IBM watsonx addresses key challenges in data management, training & tuning models, and governance: Siddhesh Naik, IBM India & South Asia

In an exclusive interaction with Express Computer, Siddhesh Naik, Country Leader – Data, AI & Automation, Technology Sales, IBM India & South Asia, outlines how orgnisations while embracing an AI-first mindset, position themselves to outperform competitors, drive sustainable growth, and create new business models to reshape their future in a rapidly evolving digital landscape. He further talks about how IBM watsonx is helping enterprises scale and accelerate the impact of AI with the data across their businesses.

In today’s data-driven world, how can technologies like AI & automation transform an organisation and expedite their digitalisation journey?

The transformative power of AI is helping businesses automate repetitive tasks, streamline operations, and enhance decision-making processes with data-driven insights. This not only optimizes workflows but also enables organisations modernise their IT infrastructure, fostering agility and innovation. Automation, when combined with AI, empowers teams to focus on higher-value activities, improving overall productivity and customer experiences. Moreover, AI-driven analytics can uncover patterns in vast data sets, allowing businesses to anticipate market trends and tailor offerings to meet customer needs. As companies embrace an AI-first mindset, they position themselves to outperform competitors, drive sustainable growth, and create new business models, ultimately reshaping their future in a rapidly evolving digital landscape.

How can enterprises measure and quantify the ROI of AI / Generative AI projects, especially considering factors like time to deployment and ongoing maintenance costs?

Enterprises should establish clear metrics aligned with business objectives, such as productivity gains, cost savings, and enhanced customer satisfaction. Time to deployment can be assessed by comparing project timelines against expected milestones, while ongoing maintenance costs should be tracked to determine total cost of ownership.

By analysing AI model performance post-implementation, considering both qualitative and quantitative outcomes, enterprises can better understand AI’s impact on operational efficiency. Ensuring high-quality, AI-ready enterprise data is crucial, as it directly influences the effectiveness of AI solutions and their ROI. For example, we just announced our Granite 3.0 models, which combined with tools like InstructLab allow cost improvements of anywhere between 3x and 23x as customers move workloads from large models to smaller, fine-tuned models.

What are some of the key challenges organisations face when transitioning from exploring Generative AI to achieving tangible ROI?

The major challenge is managing bias in AI models, as they are trained on vast datasets that may contain inherent biases, leading to skewed outcomes. Additionally, the opacity of some models complicates the ability to audit and validate their outputs, making it difficult for organisations to trust the results. Hallucination, where models generate factually incorrect information, poses another risk, necessitating robust oversight and governance mechanisms. Intellectual property concerns regarding content generated from potentially copyrighted material further complicate implementation.

How is IBM watsonx helping enterprises scale and accelerate the impact of AI with the data across their businesses?

IBM’s watsonx provides a comprehensive platform that addresses key challenges in data management, training and tuning models, and governance. The platform has three offerings. With watsonx.ai, businesses can simplify developing AI solutions and deploying them into their applications of choice. The next offering is watsonx.data, which helps effectively manage complex data environments through open lake house architecture, streamlining data access and reducing processing costs. Finally, watsonx.governance ensures ethical AI practices by automating model monitoring and compliance.

Further our new Granite 3.0 8B and 2B language models are designed for enterprise AI, delivering strong performance for tasks such as Retrieval Augmented Generation (RAG), classification, summarisation, entity extraction, and tool use. These compact, versatile models are designed to be fine-tuned with enterprise data and seamlessly integrated across diverse business environments or workflows.

What ethical considerations should businesses keep in mind when using AI and GenAI technologies, particularly in relation to customer data and privacy?

I believe that businesses who want to employ AI must adopt responsible AI, adhering to the principles of explainability, fairness, robustness, transparency and privacy to ensure that the technology benefits everyone, not just a few. We must always keep in mind that the purpose of AI is to augment human intelligence; data and insights generated belong to their creator; and AI must be transparent, explainable, and free of harmful and inappropriate bias.

Yet our IBV CEO study noted a contrast between understanding the need and actual adoption of AI governance policies – only 42% Indian CEO respondents said they have good Generative AI governance in place today. That’s why we launched watsonx.governance to help organizations in their effort to govern AI. This AI governance platform meets enterprise needs in three areas – Lifecycle governance; Manage risk and reputation; and Compliance with regulations, industry standards and internal policies.

How can organisations ensure that Generative AI models are trained on high-quality data and remain resilient to adversarial attacks?

Organisations must implement a robust data governance framework. They should focus on curating and validating data sources to ensure quality, relevance, and compliance with regulations. Employing data preprocessing techniques can help filter out noise and biases, enhancing model performance. Additionally, organisations should conduct regular audits of training datasets to identify vulnerabilities and ensure continuous improvement. Utilising automated tools can facilitate monitoring and management of data integrity throughout the AI lifecycle.