Look around. AI is everywhere. From our daily use of Siri, Cortana, and/or Alexa to Netflix’s ability to recommend exactly what we like to watch on TV at any time. For someone like me who has been working in the field of AI for more than 20 years, the state of AI today continues to amaze me. In the earlier days, creating AI models was, in a word, difficult. Building an AI algorithm, testing it, and re-training the model. This was a hard process. Now, our ability to store and process large amounts of data; readily-available, easy-to-deploy AI; and cognitive and analytics services in the cloud have changed everything, giving AI techies and business leaders alike massive abilities we couldn’t have imagined just two decades ago.
As AI applications roll out every day, the question remains, “What’s the business ROI?”
This is where scaling up AI becomes difficult. How can we leap forward in our efforts to ensure that AI’s tipping point for digital transformation delivers real business value in a manageable, customer-centric, outcome-focused way?
A closer look at the “40/60 breakdown”
Within the last 5-10 years, the ability to create data insights is now thought of as the easy part. Many cite this piece of the bigger AI picture as 40 percent of the effort. The challenging part lies in the other 60 percent: consuming and operationalizing these insights in a way that is meaningful and relevant to customers of industrial businesses towards outcomes such as productivity, efficiency, profitability. Solving this part of the AI equation is critical for businesses that want to strategically prioritize and scale AI applications.
In my perspective, I see two foundational barriers and solutions to seizing AI’s business value:
Challenge #1: Maturity or ability of existing business processes to consume AI insights
Schneider’s Chief Digital Officer once said that building a digital twin without having a way to respond to the insights it reveals is like having a phone that never rings. Likewise, building AI models without knowing how to interpret, manage, and act on the insights leaves any of us with just a shiny object that has no real applicable value. The gap here is brought on by a lack of maturity or ability of business processes to consume these insights because they were never designed for this purpose, i.e., they were never designed to be AI-first.
Here are two recommendations to answer the call of AI:
- Adopt an AI-first mindset and revisit business processes, workflows with an AI lens by looking at every step along the value chain with an AI perspective, for example: Can my maintenance and operations teams respond to and act on predictive models that flag potential asset failure? Do the workflows they follow account for AI to create insights from data?A partner in our Schneider Electric Exchange open digital ecosystem, InUse is a tech company that provides a SaaS application that allows machines to talk to improve production performance of factories. Based on the digitalization of the manufacturer’s industrial know-how and machine data processing, the application delivers connected services that solves production issues within factories. InUse has been in the TOP 5 winners of our recently closed open Innovation Challenge for a way to optimize the “rinse over run” for Clean-In-Place process for Food & Beverage applications. As an example of a success story, InUse develops a portfolio of connected analytics and services to tackle Hellenic Dairies’ plant activities including connected cleaning-in-place cycles, in order to optimize their duration and reduce water consumption by 20%.
- Speed up change management: This step involves first ensuring buy-in at the executive level as part of a broader strategy. From there, communicating across the enterprise is critical, focusing on topics such as customer-centricity, data privacy, and security. Part of Schneider’s digital strategy includes working from our IoT-enabled EcoStruxure™ architecture to develop with customers and partners such as Accenture and Microsoft a portfolio of EcoStruxure Advisor digital service offers that can be iterated and scaled to customer needs.
Challenge #2: Too many small, unrepeatable, not scalable PoCs
Creating too many small, unrepeatable PoCs has no business value here. To overcome this common pitfall, Schneider works to:
- Define the “what” and “why” with scale in mind: As AI is about much more than just integrating data and showing it in a dashboard. For example, in the Oil & Gas segment, maintaining asset performance and efficiency of onshore oil pumps is expensive. We co-innovated with Microsoft a way to harness machine learning at the edge. Our Realift rod-pump with edge analytics improves operator efficiency by proactively identifying pump problems / abnormalities through machine learning algorithms. This application can rule out false positives with predictive analytics and, better, know the onshore pump jacks’ condition two miles below the surface. This proof of concept has broader-reaching appeal, as we have seen a 15% increase in productivity when onshore pump is fully optimal thanks to edge analytics.
- Prioritizes where: Determining which AI projectswill have the most business impact requires focusing on bigger problems to solve instead of scattering AI projects to address too many smaller problems. As I mentioned with InUse, resource efficiency, including water and energy, often is a driver behind the business rationale for top-line projects.
- Addresses the how by extending innovation to the ecosystem of experts, e.g., why the algorithm makes business sense, how to best improve the AI models, and, more important, identifying concrete ways to use the data insights to drive strategic business priorities. We have big problems to solve and nobody can solve them alone. Here you can for instance leverage the capabilities of the data engineers in the Schneider Electric Exchange ecosystem, as well as technology partners that can join forces to create unparalleled value towards the requirements of end users.
We’re ready for the “real work”
It’s true: AI is not all hype if done right and at scale. The business and customer value can be created. By overcoming the glaring obstacles, we can ensure a favorable and smooth tipping point when we know and implement the mechanisms to consume and act on insights from AI models. We see the future of AI. Let’s capture it today.