Speakers

Cal Al Dhubaib
Head of AI & Data Science at Further

For Business

Training 2 or more people?

Get your team access to the full DataCamp library, with centralized reporting, assignments, projects and more

Getting ROI from AI

November 2023

Summary

As AI increasingly finds its way into business practices, the focus is now on ensuring these technologies deliver substantial value. A significant statistic indicates that 90% of AI projects fail, and Kal Al-Dubeib stresses that this problem often stems from human, not technical, shortcomings. The mismatch between AI models' outputs and actionable insights for humans is a significant hurdle. Overcoming this involves promoting AI literacy, understanding data quality, and aligning AI solutions with real business needs. The discussion also delves into the complexities of AI implementation across sectors like healthcare and advanced analytics, emphasizing the necessity of human intervention and contextual understanding in effectively deploying AI solutions.

Key Takeaways:

AI projects often fail due to human, not technical, issues.
Successful AI implementation requires aligning AI outputs with actionable business insights.
Promoting AI literacy is important for understanding data quality and AI effectiveness.
Human intervention remains vital in AI deployment, especially in complex fields like healthcare.
AI solutions must be designed with context and trust to deliver substantial business value.

Deep Dives

AI Project Failures: A Human Problem

Kal Al-Dubeib presents a significant insight: the majority of AI project failures are not due to technological problems but human ones. "Often the failure is more of a human problem than it is a tech problem," he explains. A common pitfall is the disconnect between what AI models predict and how these predictions translate into actionable insights for ...
Read More

humans. The challenge lies in closing this gap by ensuring that AI outputs are relevant and actionable for business stakeholders. This requires a thorough understanding of both the technical capabilities of AI and the business context in which they operate. By focusing on human-centric design and promoting AI literacy, organizations can enhance the success rate of their AI initiatives.

AI Literacy: The New Digital Literacy

As AI becomes increasingly integrated into business operations, the need for AI literacy becomes vital. Kal suggests that the 2020s will be known as the era of AI literacy. This involves understanding probability-based workflows and the relationship between data quality and AI performance. As Kal states, "The quality of data and the relationship between data and the performance models that we use in the wild" are vital. AI literacy equips stakeholders with the necessary skills to critically evaluate AI solutions, identify potential biases, and make informed decisions. By investing in AI literacy, organizations can empower their workforce to effectively and responsibly use AI technologies.

Practical Applications: Healthcare and Advanced Analytics

In sectors such as healthcare, AI holds huge potential to transform operations and improve outcomes. Kal shares examples from healthcare projects, where AI models predict patient readmissions and identify high-cost procedures. However, he emphasizes that the success of these models hinges on human intervention and contextual understanding. For instance, a readmissions model was successful because it provided not only predictions but also explanations of the factors contributing to high-risk cases. This enabled healthcare providers to make informed decisions and adapt interventions accordingly. Similarly, in advanced analytics, AI solutions must be aligned with business objectives and designed to enhance human decision-making processes.

Designing AI for Trust and Action

Trust and actionability are key components of successful AI deployment. Kal advocates for designing AI solutions that inspire trust and facilitate decision-making. This involves introducing stress testing into risk management processes, as exemplified by OpenAI's research on stress testing models. By understanding the limitations and potential biases of AI models, organizations can mitigate risks and design solutions that are reliable and trustworthy. Kal concludes with a call to action: "Let's try to make AI a little bit more boring." By demystifying AI and focusing on practical applications, organizations can truly utilize the potential of AI to drive substantial business outcomes.