After a number of years of growth in the venture capital and startup communities, AI is now ready for business.
Artificial intelligence (AI) has come a long way since its inception in 1948. Lately it seems not a day goes by without hearing a story about AI. In a recent BBC World Service interview, AI expert and pioneer Max Tegmark of MIT, declared “AI can do anything”! That sounds promising.
One of the more tangible AI applications for business is gaining better customer insights. In its report, The Future of Automotive Retail, EY predicts the “huge data generated from multiple customer touch points will result in various CRM [customer relationship management] considerations.” among other specific changes to the automotive retail landscape.
An example of a real-world business problem that can benefit from AI and machine learning (ML) is predicting customer churn.
Using Volkswagen’s customer automotive churn dataset, the big data analytics firm MapD developed and implemented a workflow on GPUs to respond to that business problem.
CBInsights says 2018 is the year “machine learning gets normalized”
The normalization of AI in business is underscored by Arden Manning of Yseop in his article Top 5 AI Business Trends Transforming the Workplace in 2018 in which he describes the trend towards sensible and workable business solutions.
Manning’s practical approach includes:
1. Specific Solutions
2. Use Cases
3. Functional AI
4. Machine Learning and Expert systems, and
5. The Death of the black box
1. Specific Solutions
Manning is direct on this point: “People [are] no longer looking for cool tech, but for specific solutions.”
The short video below describes how Clinc and their AI scientists built a personal assistant app to “connect individuals more quickly and easily to their financial information.”
Clinc didn’t want users to be constrained by the traditional approach of interfacing with the application using touchscreen menus. With the goal of making finance easier, they created a conversational interface. To enable their AI, they used Intel® Xeon® processors.
The app is cool tech. More importantly, it also solved a specific problem.
2. Use Cases
Specific solutions require use cases and Manning’s advice to solution providers is they need “to define clear use cases so that businesses can quickly identify and understand what each solution can and cannot do.”
An interesting use case is a chatbot that “will listen in on agents' calls suggesting best practice answers to improve customer satisfaction and standardize customer experience”. AppliedAI has compiled a searchable list of over 100 AI for enterprise use cases with chatbots among the 11 categories.
- 3. Functional AI
According to Manning, AI needs to be functional, meaning it is “software that has a clear application in business.”
You probably already use a voicemail to email service, where the voice recording is turned into a text transcript. This feature used to be only available to enterprise customers but is now available for the consumer market.
Some cities use car location and speed data to understand traffic flow and adjust the timing of traffic lights and reducing gridlock.
Physical security systems now use facial recognition to supplement their existing proximity card-based access control systems to enhance security. Similarly, license plate number recognition has existed for some time for toll road or congestion fee billing and locating stolen vehicles.
- 4. Machine Learning and Expert Systems
Expert systems are also called “rule-based systems” and they perform tasks normally handled by a human being with a vast knowledge in a certain area. Typically, expert systems are used to automate time-consuming tasks.
Siemen’s Mindsphere is a development platform with an open API. One example of its application is in breweries. By automatically collecting production line data, Mindsphere analyzes and visualizes the data to quickly identify anomalous conditions detrimental to output. In the past, data collection would have been collected manually and entered into a spreadsheet before finally being examined and analyzed. By the time the analysis was completed, production output would have been negatively affected resulting in waste.
- 5. Death of the Black Box
Unless you’re a data scientist, understanding the output of a deep learning artificial neural network (ANN) is not immediately obvious.
Once again, Manning is pragmatic: “Businesses will want more details on how algorithms come up with those conclusions” in order to translate the conclusions into actions. You’re going to want to understand the logic of the explicit and hidden layers before taking any real decisions for your business.
AI is not perfect but it’s getting there and there are more and more practical, implementable business solutions available now than ever before. Talk to a CIARA specialist for help with your AI hardware needs.