3 Things to Know Before Starting Your AI Journey
AI-Powered Search Engines—referred to as “Insight Engines” by Gartner and “Cognitive Search” by Forrester—can deliver significant value to organizations these days, provided certain risks are avoided.
As the name suggests, these engines automatically analyze enterprise content and data to surface relevant information and insights. They automatically surface relevant information for users within the context of their work and, depending on the specific use case, result in better decision making, superior customer service, more effective management, and significantly improved performance overall.
In recent years, technology to simulate human intelligence processes, or artificial intelligence (AI), has become an increasingly useful component in these systems for automating and in some cases scaling certain types of analysis. The science of getting a computer to perform specific tasks without using explicit instructions, better known as machine learning (ML), has produced so-called “supervised” learning algorithms that take labeled data sets (training sets) to automatically detect patterns and generate models that can be used to label new unlabeled data sets. Deep learning, a subfield of machine learning, involves the training and application of advanced mathematical structures referred to as artificial neural networks. Deep learning enables advanced processing of free text content and opens the door to newer and broader types of artificial intelligence such as open question answering and user intent recognition.
That all seems pretty compelling but maybe not so straightforward. In fact, according to a recent survey conducted by IDC, 25% of organizations worldwide that are already using AI solutions report up to a 50% failure rate. A lack of skilled staff and unrealistic expectations were identified as the top reasons for failure. [IDC global survey of 2,473 organizations, May 2019.]
As an independent software vendor providing an AI-Powered Insight Engine, Sinequa has direct experience with many AI initiatives. Below are the best practices we recommend following to ensure success.
Identify and Focus on the Right Use Case
First and foremost, “doing AI” is not the end game, addressing a valuable use case is. If a use case requires AI, validating that it is compelling to the business is key (ROI, usability, etc.). At Sinequa we have created a dedicated AI department to work with customers on their AI-related projects. Experts within this department analyze, jointly with customers, the highest-value use cases for AI usage, the data they have (while checking its adequacy), define metrics that are used to measure results, and use the results to prioritize and design customer road maps.
It is critical that the use cases be grounded in reality, meaning they should be well within the capabilities of today’s technology, and the use cases should merit an AI-powered approach. Many use cases can be accomplished in a simple and straightforward manner without the need for artificial intelligence technology. However, there are certain use cases that can only be realistically and cost-effectively accomplished with AI technology, like auto-classifying large volumes of confidential documents or auto-routing large volumes of support tickets based on the case details, for example.