Three keys to AI success in the enterprise
Many CIOs have artificial intelligence (AI) on their radar. A recent survey by Genpact and Fortune Knowledge Group of senior executives found that 82 percent of respondents plan to implement AI-related technologies in their company over the next three years. Yet a few organizations are ready to realize the full potential of these technologies.
The reason is clear, as AI presents a new approach to computerized models. Using pattern recognition, natural language processing (NLP), and computer vision, these machines are able to significantly increase predictive decision making and automate complex work. Through machine learning and neural networks, they can quickly process large amounts of data, which humans cannot, to support decision making and gradually become “smarter.”
While the number of applications is growing, few companies are successfully using AI technology to drive improvements in their enterprises. The problem often lies in a lack of focus on what problems are these systems actually trying to solve. In reality, success with cognitive technologies requires a defined goal, strategy to reach it and the right tools.
Define the Goals
Technology decision makers must first identify the best areas to apply AI and define the ultimate goals of the project. For instance, these tools may be used to create new products or offerings, improve a current product, optimize internal operations, improve customer experience, or lighten workloads for staff members.
One global financial services firm needed to extract and normalize financial information coming from customers in different formats and languages. Its previous process was manual, making it hard to keep up with its global operations. The firm recognized that it needed to automate to speed things up, reduce customer cycle time, increase customer satisfaction, and cut costs.
Technology decision makers must first identify the best areas to apply AI and define the ultimate goals of the project
With these goals in mind, the firm introduced extraction, machine learning, and computational linguistics. Now, it can pull financial information from any format, including spreadsheets and images. It can normalize the numbers automatically and understand footnotes for analysis. By automating, the organization has cut costs by 70 percent and improved customer satisfaction, compliance, and on-time credit decisions.
Establish the Strategy
An effective strategy should address content, people, and change management. Companies need to determine what content would best support their project and goals. For instance, a wealth management firm has to compile reports on its clients’ financial health using information from hundreds of outside custodians in a variety of formats. The firm determined it needed to be able to leverage the content from these different sources in a standardized way. Using a combination of NLP, machine learning, and computational linguistics, it can now pull and organize financial data from any source. Financial advisors can monitor assets in real time and create on-demand reports. Most importantly now the firm has enriched content to derive patterns for proactive advising.
When it comes to people, companies need ones with domain knowledge and understanding of AI technologies. Individuals should know how these systems can be applied and would integrate with other processes. They should be able to communicate with executives and relate them to key issues of the business. Companies that do not have staff with this domain expertise can look towards outside consultants.
Another aspect of people is managing change. AI may not sit well with current employees, out of fear of job displacement. Organizations need to communicate that these systems are not meant to displace jobs, but augment them. In fact, most projects see minimal layoffs.
Identify the Right Tools
Companies need to decide if they will build or buy their AI tools, use proprietary or open-source software, and implement standalone applications or a broad platform. The answers depend on each problem set. If you have dynamic, structured data, machine learning will likely deliver the insight you need. If you have unstructured data, computational linguistics will work better. Companies without experience may want to use an AI platform that brings together a variety of different tools, including data engineering and robotic process automation, that automation that can complete the process.
For example, a large consumer packaged goods company uses retailers to execute a large segment of its go-to market strategy. Retailers claim a portion of promotion costs from the manufacturer. Since everything was manual, the company would often audit only 30 percent of the invoice value, resulting in overpayments. The company wanted to reduce manual errors using AI.
Without an off-the-shelf solution, the company turned to a broad, AI-based platform to carry out the automation of the trade promotion process. It employs computational linguistics, pattern recognition, and machine learning to extract contract terms and validates them against invoices and other data. If all is clear, it triggers the payment. Using the platform, the company has seen 60 percent productivity gain across the end-to-end process and an additional $21 million per year in revenue from reduced overpayments.
While cognitive computing is at the top of mind for many CIOs it should not be seen as a passing phase. It will dramatically shift the way companies do business long into the future. Companies need to begin establishing their AI goals, strategies, and arsenal of tools, or risk being left behind more innovative competition.