Artificial Intelligence for Enterprise

Andrew Ng says, “AI is the new electricity”.

AI, Artificial Intelligence is a set of tools that mimic human cognitive capabilities associated with learning, reasoning, and decision-making. It either provides support to humans in Augmented Intelligence (e.g. breast cancer diagnosis) or replaces them in redundant tasks in Automation (e.g. Virtual Assistance).

With that being said, major companies like GAFAM (Google, Apple, Facebook, Amazon, and Microsoft) are going full speed with this new energy, AI. Nevertheless, other organizations should also benefit from AI, but how?

First and foremost, organizations should comprehend why they need AI, how it will benefit them, and can they cope up with it:

• What are our desired outcomes?
• How will cognitive technology help us achieve these outcomes?
• What is our long-term vision with this technology?
• Do we have strong executive support?
• Can we adapt existing processes and roles?
• Do we have the necessary skills within our organization?
• Do we have the IT environment we need to get started?

At that point, come the five steps of incorporating AI into the enterprise, as defined by IBM’s experts [1]:

(AI ladder steps[2])

Modernize: Build IA before AI.

The organization should initially build an Information Architecture to place data within the reach of its employees. Some companies still store data on physical material like papers, optical discs, or hard drives; which puts them at extremely grave risk of data loss and reduces their data’s availability. However, this concern might be addressed by modern Cloud solutions that provide both security and scalability. By using a reliable Cloud service, the organization can store and protect a huge amount of data. Besides, Cloud makes data accessible anywhere and anytime.

Collect: Make data simple and accessible.

After establishing the right Information Architecture, data should be collected to improve business quality. While data can be either internal (e.g. process’ KPI) or external (e.g. customers’ preferences on a shopping website) two solutions might be imagined. Internet of Things that transmits data of multiple connected objects to the internet or Big Data scraping programs that retrieve data from an external source. Moreover, it is recommended to collect more data, as it often results in better model performance.

(Models' performance display as training data size increases. [3])

Organize: Create a business-ready analytics foundation.

Data should become business-ready. Put differently, teams must process data and identify the most relevant to focus on. For this kind of work, data engineers happen to take the lead. They prepare documentation, that contains information about each data, to help better navigate the latter. Furthermore, they manage the different types of data at hand: structured (e.g. SQL databases), semi-structured (e.g. e-mails), and unstructured (e.g. images).

Analyze: Build and scale AI with trust and transparency.

The most crucial thing for the organization in the analysis step is transparency with stakeholders about how data will be managed. Machine learning engineers analyze and extract valuable insights from the data with various technologies of prediction and classification. These technologies are highly developed that they embody an incredible milestone for science.

Infuse: Operationalize AI throughout the organization.

At this point, the organization is completely aware of its knowledge. Therefore, it should work on incorporating the gained insights into the business. It could adopt a new marketing strategy as its analysis revealed a new segment of consumers to focus on. It could tune a chemical process that creates so much waste. And why not further deploy AI? Use virtual assistants to deal with clients, automate processes, and many more.

To summarize, enterprises should consider this modern science of artificial intelligence as it enables extensive growth. They must investigate the ground for AI. Additionally, they ought to integrate AI by scaling its ladder. They should initially establish an Information Architecture to permit better data management. Afterward, they must collect and organize their data. Next, they have to analyze this data to get business insights. At long last, they should infuse the results of these analyses into the business. Eventually, enterprises should adopt a continuous motion of artificial intelligence development.

References:

[1] The AI Ladder – IBM

[2] How to Scale the AI ladder – Daniel G. Hernandez, General Manager, IBM Data and AI

[3] Model Selection and Evaluation for Machine Learning: Deep Learning in Materials Processing – Adam Kopper, Rasika Karkare, Randy C. Paffenroth & Diran Apelian

Further readings:

Reshaping business with artificial intelligence – MIT Sloan Management Review

Scaling the AI Ladder – Rob Thomas, Senior Vice President, IBM Cloud and Data Platform

The impact of AI on business and society – Financial Times

AI, automation, and the future of work: Ten things to solve for – McKinsey Global Institute 

The AI Ladder: Accelerate Your Journey to AI – Paul Zikopoulos & Rob Thomas