How AI Will Impact Organizational Structures

Artificial Intelligence is necessitating changing organizational structures and the creation of new roles, just as the internet did post-Netscape.

The first International Conference on the World Wide Web at CERN outside Geneva Switzerland in May 1994 is commonly recognized as the birthplace of the commercial internet. How business is conducted and who conducts business was forever changed. The internet spawned new businesses, new business models, new ways of doing business and so much more. All of which, over time, required companies to reorganize existing departments, create new internal organizations, invent new roles with people with new skill sets. The sum total of all the change is remarkable. The way companies were organized before the commercial internet and the way they are organized today are radically different.

In the ImageNet Large-Scale Visual Recognition Challenge in 2012, a team from the University of Toronto entered an algorithm called SuperVision and swept the competition. SuperVision’s domination is now being recognized as a turning point for the field of artificial intelligence. For the first time, machines rivaled humans in the difficult task of image identification. Many refer to this as the end of the second ‘AI winter’ and when interest and breakthroughs in the field exploded. About three years later, some of the first deep learning models become widely commercialized, democratizing artificial intelligence for business. We believe that this time period, 2012–2016, will be viewed as the birth of commercial AI.

Artificial Intelligence is no longer just the subject for research papers nor the experiments of university research laboratories and the benefits of AI are well documented.

Like the internet, Artificial Intelligence, once again, will change how business is conducted, and who conducts business.

Initially, we will see the integration (or destruction) of traditional business and technology silos to allow the cooperation and data aggregation that is critical to running AI programs.

AI models are, in some ways, like house plants. You need to care for them to keep them growing and thriving. They require continuous data collection, training, engineering, and monitoring to ensure their performance increases in accuracy and does not decay. New AI projects need close alignment to the challenges of the business and means to measure success. This all becomes a big challenge when you run multiple AI models, which forces data teams, technical teams, and business teams to all work together to ensure success.

 

 

We believe that the future of AI-driven organizations includes the birth of the AI Center of Excellence (CoE), a new department that will be business-led and follow the growth levers defined by the corporate strategy. The CoE will own, consolidate and govern all corporate data assets for ready access to any project or organization that needs to train models.

From a business standpoint, artificial intelligence programs at scale require tight alignment with core objectives. Machine learning models often provide the missing link between data, analytics, and measurable outcomes. Roles like Chief AI Officer, AI Strategist, and AI Product Manager will emerge to ensure data collection and AI use case identification all tie back to the corporate strategy. These new roles make sure success is well-defined and prioritize projects that require the least amount of resources with the maximum return on investment.

Technology, data, and DevOps are a few of the most important components of AI-driven organizations. New technologies will be acquired to capture, store, access, and process data. Models require a tremendous amount of data to perform accurately. And since ‘data is the new oil’, savvy organizations are hoarding data produced by everyone — and everything — from consumers to sensor to leverage for machine learning.

AI requires tremendous compute power to train and run models. Most businesses won’t keep expensive compute or storage hardware on-site and will favor a service-based model in the Cloud and pay based on consumption. Data collection is not a one shot effort, but instead an ongoing, strategic one. The AI strategist will help define what data exists that can produce key intelligence, and how to acquire it. Organizations need to continuously assess and collect new data that are fed into models. And, of course, all of this data and technology requires governance, privacy, and security.

Key components of operations in AI-driven organizations

 

 

One completely new component of AI-driven organizations is the concept of human-in-the-loop systems. Teams of people monitoring and providing feedback to models when they make mistakes. As stated earlier, models are like organisms and thrive when fed the appropriate data. They can improve with more data and training while in production. Conversely, they decay without supervision and with inappropriate data.

Teams of humans will be dedicated to manage models and give input when the model is unsure or not qualified to make decisions. Their input provides additional data to train models so that errors are not repeated. The human-in-the-loop system may place humans as a second opinion or act as a manual override. For example, in call centers, chatbots may handle most customer needs, but escalate exceptions to human experts.

AI will also drive operational changes, requiring teams, customers and suppliers to work alongside AI systems that will augment their decisions. It is important that organizations start planning now for a centralized AI organization which connects business, data, and engineering components to ensure outcomes far outweigh investment.

Is your organization ready for these changes, including wrestling with new types of ethical issues, bias in data and making decisions based on probabilities? How can you and your team get ready to confidently lead in this new era? Our mission at KUNGFU.AI is to help democratize these technologies and help our clients build AI organizations brick-by-brick.

Our recommendation is to start small with point solutions on narrow use cases. Once a few use cases are driving value, we can help your organization install technology, processes and reskill staff to develop and run AI programs at scale. This strategy is a winning one because it eases into the investment and de-risks the technology. Once value is proven, then full transformation can start to be embraced.

So one question remains. Is your organization ready to start this transition?

Want to learn more? KUNGFU.AI’s CTO, Ron Green, recently spoke on the topic of How AI Will Impact Organizational Structures at SXSW.