Getting AI-Ready — CDO Insights from the AI & Analytics at Wharton and Informatica Symposium

The rise of Artificial Intelligence (AI) has ignited a firestorm of discussion around its impact on the future of work. A notable event addressing this issue was the AI & Analytics at Wharton and Informatica Symposium held on Wharton’s campus. This gathering, in collaboration with Informatica and the Wharton AI & Analytics Initiative, brought together chief data officers (CDOs) and thought leaders to explore the topic, “AI and Data Readiness for Digital Transformation.”

This article examines the main insights from the event, focusing on the relationship between humans and AI, the challenges of adopting large language models (LLMs), and the specific hurdles faced by CDOs.

Humans and AI: A partnership, not a replacement

One of the most prominent themes that emerged from the discussions centered on the misconception that AI will replace human workers. Leading professors from the Wharton Al & Analytics Initiative emphasized the importance of viewing AI as a tool for augmentation, rather than displacement. Stressing this point, one of the CDOs drew an analogy to the evolution of cell phones.

They went on to note that as these devices became more efficient and affordable, their widespread use among workers didn’t free up time; instead, it created new expectations for constant availability and increased output. 

Similarly, AI was highlighted as having the potential to amplify human capabilities, leading to greater productivity and the creation of entirely new opportunities.

There was agreement that the focus should be on leveraging AI's strengths in areas like data analysis and pattern recognition while utilizing human judgment and creativity for tasks that require empathy, critical thinking, and ethical decision-making. This human-AI partnership holds the key to unlocking the true potential of artificial intelligence.

The thorny issues of Large Language Models

Large language models (LLMs) are powerful tools with the ability to generate text, translate languages, write different kinds of creative content, and answer your questions in an informative way. However, their adoption comes with its own set of challenges. Professors attending the symposium highlighted several key concerns:

• Intellectual Property (IP) protection and improper usage: LLMs can be trained on vast amounts of data, raising questions about ownership and potential misuse of copyrighted material. It is critical to establish clear guidelines regarding responsible data usage and IP protection in the realm of AI.

• Data security: The effectiveness of LLMs hinges on the quality and security of the data they are trained on. Robust data security protocols are essential to ensure the integrity and ethical use of these models.

• Hallucination and bias: One of the biggest challenges of LLMs is their tendency to generate outputs that are factually incorrect or misleading. Developing techniques to mitigate "hallucination" and ensure the accuracy of information produced by AI models is crucial.

• Costs: Developing and maintaining advanced LLMs requires significant resources. Finding cost-effective ways to train and run these models will be essential for wider adoption.

• Explainability: Understanding how an LLM arrives at a particular answer remains a challenge. Enhancing the explainability of these models is necessary to build trust and enable their use in AI applications that require clear reasoning.

• Organizational factors: Implementing LLMs effectively requires addressing organizational hurdles. This includes establishing clear roles for humans and AI, capturing and distributing knowledge efficiently, and verifying and adjudicating outputs generated by LLMs.

CDOs: Navigating the AI labyrinth

While LLMs present unique challenges, CDOs face an additional layer of complexity in their quest to leverage AI effectively within their organizations. Some of the key challenges identified at the symposium include:

• Leadership challenges: Striking a balance between excitement about AI and cautious investment is crucial. Leaders must avoid getting swept up in the hype while recognizing the potential of AI to drive innovation.

• Demonstrating value and ROI: CDOs need to prove the business value of AI initiatives quickly, while also mitigating the risks associated with irresponsible AI use.

• Policy and governance: Developing robust policies and governance frameworks for data, models, and orchestration is essential. Ideally, these frameworks should leverage existing tools to minimize the need for custom coding.

• Data quality and accessibility: High-quality, organized, and unified data is a prerequisite for successful AI implementations. Additionally, making data management tools more user-friendly and accessible through natural language interfaces can empower individuals to interact with data more effectively.

• Data literacy gap: CDOs recognize a significant data literacy gap within their organizations. Bridging this gap requires innovative solutions, such as making data management tools more intuitive and fostering a culture of data-driven decision-making.

• No-code, governed AI models: The development of user-friendly, no-code tools for deploying AI models is essential for wider adoption. However, these tools must include robust governance mechanisms to ensure responsible use. One notable example discussed is the use of RAG (Retrieval-Augmented Generation) applications.

Incorporating LLMs within an organization’s own data can enhance the relevance and accuracy of responses. However, the widespread adoption of RAGs may be hindered by a lack of skilled personnel. A standardized framework for deploying GenAI is a valuable tool for transitioning pilot projects to production environments and demonstrating ROI quickly. Additionally, such a framework can also help overcome the skills gap that arises from building systems from scratch using multiple technologies.

The road ahead: A collaborative journey

According to our CDO Insights Survey 2024, CDOs expect to employ at least five tools in most cases to navigate their AI labyrinth. The great news is that Informatica customers are well-positioned to get ready for AI. With the Informatica Intelligent Data Management Cloud™ (IDMC) customers have a complete data management solution at their disposal and can address multiple aspects of today’s CDO concerns that were highlighted during the symposium.

IDMC includes tools from our AI-powered CLAIRE Copilot, which aids in building data literacy, to automated data governance for managing data pipelines and AI models. Additionally, it encompasses security and privacy controls to make data ready for AI use, demonstrating how Informatica can help accelerate AI readiness.

Stay tuned for a more detailed output from this in-depth discussion by CDOs and thought leaders. A whitepaper will be produced, exploring techniques to help organizations overcome all these challenges and quickly show returns while responsibly deploying AI applications.

About the Author:

Nick Dobbins, an accomplished technology leader with over twenty years of experience in driving innovation within data management, has held various positions ranging from software engineering to technical sales leadership. He is currently positioned as the Vice President, Worldwide Field CTO at Informatica — a global enterprise specializing in cloud data management.

In his capacity, Dobbins provides guidance to clients on addressing their most complex data management challenges, collaborating closely with CTOs, CDOs, and other leaders in data management to devise strategies that yield successful business results. Acting as a liaison between the field and product leadership, he steers innovation through the development of roadmaps and acquisitions.

Dobbins remains vigilant regarding emerging trends, ensuring that Informatica remains at the forefront of enabling our customers' success.