This Week in ML & AI avatar

Dexa/This Week in ML & AI

Learn more
Published Apr 21, 2017

Selling AI to the Enterprise with Kathryn Hume - #20

Kathryn Hume delves into the challenges and strategies of integrating AI in enterprises, highlighting the need for education, cultural adaptation, and strategic use of AI for competitive advantage, while advocating for a nuanced approach that prioritizes augmentation over full automation.
Episode Highlights
The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence) logo

Topics covered

Popular Clips

Episode Highlights

  • Augmentation

    Kathryn Hume highlights the strategic decision enterprises face between augmentation and automation in AI adoption. She notes that many non-technical stakeholders mistakenly assume a leap from manual processes to full automation is feasible, but practical challenges like data scarcity and system mistrust hinder this transition 1. Instead, companies often start with simpler algorithms, gradually building complexity as they gain confidence in AI's capabilities. Hume shares an example from the tax advisory sector, where AI augments human expertise by providing dynamic alerts on legal updates, rather than replacing human judgment 2.

    How people work will change when they have machines as a companion and component, but most of the time, just from a pure technical perspective, it's so hard to have the amount of data that's required to train a system that could really perform with the levels of accuracy that are required to do risk oriented tasks.

    ---

    This approach allows enterprises to enhance productivity while maintaining human oversight.

       

    Competitive Edge

    AI offers a competitive edge by enabling innovation and product differentiation, but it requires a strategic approach. Hume explains that successful companies often have a clear problem to solve and seek external expertise to evaluate their AI initiatives against industry leaders like Google and Facebook 3. Many enterprises mistakenly believe they have sufficient data for AI projects, but often lack the structured, labeled datasets necessary for advanced techniques like deep learning. Hume suggests that AI and data science are intertwined, with AI being an extension of data science that handles more complex data types like images and speech.

    We have this false impression that just because it's a big enterprise, they're going to have lots of data. They do, but they haven't been considering data over the last hundred years with an eye towards building machine learning products.

    ---

    This perspective underscores the importance of data readiness and strategic planning in leveraging AI for competitive advantage.

Related Episodes