top of page
  • Madeline Weiss, Director

ADVANCED PRACTICES COUNCIL (APC) INSIGHTS - FEBRUARY 2023 MEETING HIGHLIGHTS


Advanced Practices Council members – senior technology and data executives across industries – met in Atlanta in February to continue learning from exemplary researchers and practitioners (including themselves) on topics they voted as high priority for their future success.


A continuing theme is digital transformation because it leads to many benefits (financial as well as preparation for future growth and competitiveness). Evolving technologies needed for digital transformation are remarkable and dazzling, but the technology is easy compared with adapting the company’s culture, structure, talent, and governance to reap the benefits.


Developing and Measuring Digital Maturity

Dr. Gerald Kane of the University of Georgia reported from his research (20,400 survey responses and 100 executive interviews) that companies further along in digital transformation were more nimble, were better able to scale, had greater stability, and were better able to integrate new capabilities (often from third party partners).


These companies’ cultures were characterized by greater risk appetite, more empowering leadership styles, more flexible work styles and experimentation, and greater agility.


These companies tend to have cross-functional project teams with considerable autonomy regarding how to accomplish team goals. Team members’ performance is evaluated based on meeting team goals. Executives who were interviewed explained that cross-functional teams reflect the nature of digital, which requires working across functions and allows for getting beyond organizational bureaucracies.


How are digitally transforming companies acquiring their talent? Not only do they attract people who want to work for digitally savvy businesses; they also develop current employees through training and on-the-job learning opportunities.


Governing Digital Transformation Across the Enterprise

According to Dr. Didier Bonnet of IMD, good digital governance is essential for reaping greater financial performance with digital transformation. Such governance entails matching transformation scope to the company’s culture as well as overcoming structural and process obstacles.


Governance models vary across companies and industries. For example, Starbucks has created new digital roles including dedicated digital executives with sufficient resources. GE Digital has created digital units focused on delivering digital services. Michelin has created digital units in some cases and has bought digital operations in others.


Although successful governance practices vary, Bonnet recommended that all organizations strive for more data-based decision-making and self-organizing practices, such as team autonomy, cross-collaboration, job fluidity, and self-forming teams.


Scaling AI

Dr. Amit Joshi at IMD noted that AI initiatives generally originate in business units as islands of experimentation that address local problems and are funded locally.


Over time, these islands of experimentation morph into centers of excellence that are not as local. In some companies, centers of excellence become incorporated into federations of expertise with a library of tools and practices that can be customized for local needs.


Making the leap from islands of experimentation to centers of excellence helps to increase the impact of the AI initiatives and allows for sharing of best practices. It requires prioritizing efficiency over agility and placing the new centers of excellence in divisions that facilitate execution.


Some organizations will make the leap from centers of excellence to a federation of expertise, where the centers of excellence become shared resources to both IT and business units. This allows for standardization of data, practices, tools, and governance. It also allows for incorporating domain knowledge, which might not have existed in the centers of excellence. It requires integrating agility with efficiency through data science sandboxes and joint funding between the centers and business units.


Edge AI

After defining edge computing as any computing and networking resource along the path between data sources and cloud data centers, Dr. Mudhakar Srivatsa of IBM described the key benefits of edge computing by sharing examples. Edge AI can visually inspect objects (e.g., fire extinguishers), optimize assets and production (e.g., inspection of an electrical room to capture the heat signatures of assets), and improve security and prevent losses at points of sales (e.g., McDonalds).


While many new edge AI systems help solve real-world problems, creating and deploying each new system often requires a considerable amount of time and resources. Each new system requires a large, well-labelled dataset for the specific task you want to tackle. The AI model must learn to recognize everything in the dataset before it can apply the model to a specific case.


The next wave in AI replaces the task-specific models that have dominated the AI landscape to date. New models, called foundation models, are trained on a broad set of unlabeled data that can be used for different tasks, with minimal fine-tuning. We’ve seen foundation models in imagery and language (GPT-3). Input a short prompt, and the system generates an entire essay, or a complex image, based on your parameters, even if it wasn’t specifically trained on how to execute that exact argument or generate an image in that way. These new models, as the name suggests, can be the foundation for many applications. The model can apply information it has learned about one situation to another.


Foundation models have been used at the edge with significant benefits. They were 47% more effective in modelling liquified natural gas reactors and capturing system behavior than previous models. They were 9% better at predicting when oil well submersible pumps would fail than previous models. They were 12% better at predicting impurities in a chemical process in real-time than previous models. At the same time, these foundation models required much less effort for data aggregation to deliver higher accuracy rates in a shorter time, thereby increasing revenue potential.



bottom of page