Artificial intelligence (AI) is not just using statistics and mathematical modelling to draw correlations and conclusions from large sets of data. AI promises to offer enterprises “magical” attributes to solve business problems in much shorter time frames than humans can achieve. For example, retailers are using AI to comb through databases with the goal of balancing inventory levels and making recommendations for write-offs, expediting expiring inventory, or reducing inventory buffers. Large manufacturing companies are using AI to identify issues in production lines or to scan for quality defects – which improves product availability, and fewer returned goods. And in IT, CIOs are using AI to classify trouble tickets to drive the right prioritization, identify bugs in computer programs faster, and correct technology issues before they happen.
With all the benefits of AI, one might think that enterprises are adopting it en masse. However, in 2019, one analyst firm found that only 14% of global CIOs have adopted these advanced technologies! Clearly, there is a lot of business value sitting on the sidelines waiting to be achieved. These include innovative products for customers, employees and other stakeholders, lifesaving medicines, and better ways of working that can help businesses weather the storm and thrive in today’s turbulent economic environment. AI is here, now, and ready to start driving value in your enterprise.
AI terminology can be confusing. To demystify AI, it can be broken down into three pillars to help organizations automate, continually improve, identify new trends and patterns, and improve the customer experience. The pillars include supervised learning, unsupervised learning and reinforcement learning.
- Supervised learning – These types of algorithms use existing data to make predictions. An example would be combing through 36 months of accounts receivable data to predict which customer invoices might be paid late.
- Unsupervised learning – Enterprises use this type of AI to find correlations in data that are near impossible for a human to spot. For example, in search of vendor fraud, an unsupervised learning algorithm could scan tens of thousands of invoices, breaking them into clusters in order to find commonalities or disparities. Issues could then be flagged for data science teams to review in depth.
- Reinforcement learning – A technique gaining steam over the past 5-7 years, this type of AI incorporates rewards and penalties within the model as the algorithm learns to get the right results. These algorithms are often used to improve automation techniques or to make equipment more efficient and productive.
When considering which type of AI to implement, most organizations start with supervised learning because these types of applications mostly take advantage of structured data (e.g., names, dates, credit card information, invoices, etc.) which are more plentiful and accessible within an organization. Then, as companies expand their AI capabilities, unsupervised and reinforcement learning may provide even greater dividends and even larger benefits.
AI in action
While most of today’s IT resources are dedicated to running the day-to-day business, there is an opportunity for the transformative CIO to carve out a budget for innovative AI use cases. Here are just a few examples where enterprises are deploying AI to drive value practically, and in short timeframes:
- Improving application development – For a large US company, EY developed an application that helps development teams discover coding process flows, reduce risks and decrease onboarding developer time.
- Boosting code quality – EY teams are using AI to boost code quality and help enable a more efficient process to apply corrections. For example, one company is utilizing an EY software bot for program reviews in a current version of SAP and determines what code might not work for an upgrade. Beyond finding solutions, the bot can also fix the code automatically.
- Data cleansing – A US oil and gas company engaged EY teams to improve data cleansing initiatives. Mismatched data from various business units was causing inventory issues and directly affected their customer base. Using AI, and in just 10 weeks, EY teams improved data collection and quality efforts for 28,000 data fields and identified pricing and lead times for 1,684 materials. This saved hundreds of hours in data cleansing efforts, and now all business units within the company are using the same master data to help to harmonize data collection efforts.
How to increase AI adoption
To move from the “magical” promise of AI to the “practical” reality of driving business results, enterprises need an infusion of talent, data, and new applications.
- Talent – Taking full advantage of AI means going beyond traditional IT and business experts. A strong AI program will combine various skills, including data engineering, data management, data analysts, and data science, along with business acumen and subject-matter expertise. All of these capabilities must come together to effectively use advanced technologies to create competitive advantage for the enterprise.
- Data – As more data are available for consumption, there is a real danger of information overload. To counter the deluge, IT organizations are setting up data-refining capabilities to harness internal and external data sources into corporate data stores. Enterprises are creating strong data management programs to understand what data are stored in which repositories and mine them so that they are accessible to business users.
- AI integration – There are always new vendors in the market offering advanced AI capabilities across a wide span of business cases. However, the most advanced capabilities are of little value if the IT organization cannot implement them. Systems integrators can help implement these advanced technologies, engineering them to work with legacy systems to help drive better business decisions.
Practical AI starts with IT
As a proof point for the value that AI can bring, CIOs have an opportunity to start piloting AI within their own IT organizations. First, start by examining current IT processes and inefficiencies. Second, identify areas where AI can help improve day-to-day activities. Then create a small task force to tackle areas for improvement in short sprint cycles where value can be quickly articulated.
Where can AI fit within the IT organization? Some examples include fixing code, automating processes, monitoring for outages, and predicting upgrade issues. With cost savings, improved productivity or innovation as proof points, it becomes much easier to talk to other business leaders about how AI can drive similar value for their organizations.
Paul Barsch, Assistant Director, Ernst & Young LLP also contributed to developing this article.