It's time to make the Machine Learn your business

There is a lot of talk about AI (Artificial Intelligence) and ML(Machine Learning). Is it just a hype? Or can AI add substantial business value to your business? Let's take a look at some of the basics and ground truth about AI, the opportunities and challenges it offers to business and some real use cases which are providing differentiated advantage to the early adopters. Also discussed is an approach for business to derive value from this disruptive technology.

What is ML? Is it same as AI?

In a broader sense AI can refer to any of these technology systems – Robotics & autonomous vehicle, Computer vision, NLP (natural language processing/understanding), Virtual assistant (/chatbots) & Machine Learning.

Machine Learning can be a common ingredient to some of the above systems. ML as such refers to any computer program which can self learn from its experience (/dataset) without the need for explicitly programming for each experience.

How ready is AI for business?

AI has been around for decades. But it hasn’t come to the main stream commercial space until now. Now it may be changing. Many tasks which were considered to be requiring human cognitive capability like identifying patterns, making meaningful insights, taking decision and forecasting can now be done by AI powered applications.

What has changed? Abundance of data is one reason. Big data and IoT sensor technologies have largely contributed towards this. Second factor could be the reduced cost of high computing power (e.g., GPU for deep learning). The third reason is the advancement in the AI technologies itself. All these are enabling the AI to start solving business problems which requires cognitive capabilities.

Some successful & impactful use cases

AI is already being used by some of the businesses to drive value. Adoption is varying from domain to domain - High tech, Automotive and Financial services are the leaders in AI adoption. Education and Healthcare domain seems to be the slowest in the AI adoption. Increased profitability is realized by many of these early adopters. Some of the impactful use cases from various domains are given below.

Accurate forecasting in retail

German e-commerce company Otto through data analytics found that the number of items returned can be reduced significantly if shipments are made within 2 days of ordering. Also they found that the customers prefer single shipment for all the items in their orders. Otto found the solution by using AI. A deep learning algorithm analysis 3 billion historic transaction using 200 variables to predict what customers will buy. The AI system is able to predict with 90% accuracy what will be sold within 30 days. By using the system, surplus inventory to be kept is reduced by 20%. The returns per year is reduced by 2 million items.

AI to make flights more efficient

GE conducted Kaggle challenges (Flight Quest 1&2) which identified AI usage for increased efficiency in flight. The winning algorithms uses national airspace data and determines the most efficient routes, speed and altitude in real time. It considers various parameters like weather, wind and airspace restraints. The solution provides 12% efficiency improvement. GE has estimated that the proposed solution can reduce the annual fuel consumption by 360 million gallons and save $3 billion per year for the airline industry.

Technology Assisted Review (TAR)

TAR is having a significant impact in the legal domain. First recognized by US court in 2012, TAR is used in the discovery phase of a legal proceeding for identifying the documents (including emails, pdfs, presentations, databases, voice messages etc) which are relevant for a law suit or investigation. TAR involves multiple technology and process and may be implemented in different ways. But typically core is a supervised learning algorithm where expert tags documents as relevant or non-relevant. This is used by the algorithm to identify similar documents for further analysis by expert. This way the system can sift through huge data sets in reduced time frame. Studies have shown that it reduces the total hours spend in the eDiscovery phase by 20-30%, at the same time increasing the accuracy.

Predictive Maintenance in energy plants

In most of the coal power plants, aging equipment is a major problem requiring constant maintenance and resulting in down time. One coal power plant was able to use AI based predictive analytics to predict timing of failures, 6 months in advance with 74% accuracy. This kind of predictive maintenance can help in 15 to 20% increase in their profit margin.

Personalized cancer treatment using AI

In the healthcare space many companies uses AI for providing accurate disease diagnosis to personalized treatment. For example Oncora medical, a startup company, uses AI with integrated digital medical database providing Precision Radiation Oncology platform. This enables personalized cancer treatment for patients providing better results. IBM Watson also offer solutions enabling personalized treatment for patients. 

AI in adaptive learning

In education field AI is used for bringing value in different areas. Some of the AI applications are - for identifying skill gap as per future industry needs, for identifying behavioral changes in students warning school drop outs, for enabling personalized learning, and even for enabling virtual teaching assistant. Quantum, a startup, offers AI based adaptive learning and assessment solutions. Quantum Adaptive learning is playing a key role in the Artificial Intelligence market in the US education sector which is expected to grow at a CAGR of 47% during 2018-21.

Challenges in AI

Unavailability of AI based solutions for enterprise level business opportunities is a concern. There are generic AI based platforms from various vendors. These include IBM’s Watson, Amazon ML/SageMaker, Microsoft Azure ML studio, and Google ML Engine / TensorFlow. These platforms offers specific ML functionalities like speech recognition, classification, regression, anomaly detection etc. But there are not many mainstream enterprise application products helping businesses to easily take advantage of AI. This can be addressed by developing custom solution for the business by partnering with startups or other third party. 

Second challenge is the acceptance of the AI solution by the existing staff. For an AI project to be successful, it is important to have upfront buy-in from all stakeholders including the experts who are required to train the ML system. This requires openness and transparency of the project being executed. It should not remain as a high-tech solution being implemented within the organization by a small group of experts (internal or external). If you are partnering with a startup or other third party, ensure that the partner also has a culture of openness and willingness to share the approach details with the relevant teams in your organization.

Scarcity of capable ML experts is another issue the industry is facing. To overcome this, companies adopt various strategies including acqui-hiring (acquiring startups for talent), internal reskilling, tying up with external partners (typically startups) etc.

How Businesses can realize the benefits of AI driven growth ?

As per a McKinsey study in 2017, AI has provided higher margins for early adopters and expect the performance gap with other firms to widen in future. The evidence suggests that AI can deliver real value to serious adopters and can be a powerful force of disruption. What is the recommended approach for a business to adopt AI? The six steps recommended are

• Define a strong AI business case and connect it to the organization's strategy
• Build internal capability or partner with an AI startup
• Have a discovery phase to analyze the core opportunities/challenges to see whether a differentiated solution can be found through AI. The effort need to be co-led by business and technical leaders
• Have a long term plan but start with small goals
• Take an agile approach - focus on working prototype and expand incrementally
• Re-design processes workflows to incorporate AI insight into the workflow deriving value

References

IDC FutureScape: Worldwide IT Industry 2018 Predictions

HBR: The business of Artificial Intelligence

McKinsey Global Institute: Artificial Intelligence the next digital frontier

10 Top takeaways - Forrester AI readiness study

Fortune: 50 companies leading in AI

The economist: How Germany’s Otto uses artificial intelligence

Top Artificial Intelligence Companies in Healthcare