Hollywood's portrayal of artificial intelligence has fascinated people for decades. Personal digital assistants,...
such as Apple's Siri, Microsoft's Cortana and Amazon Web Services' Echo blend basic speech inquiry and response with the cloud. But users often are surprised to learn a cloud intelligent agent has broader use -- and even a place in the enterprise.
Most end-user devices have limited onboard intelligence. In addition, cost and performance issues make it difficult to fulfill user requests. Having a cloud agent act as the information-processing point offers a solution. Users ask the agent a question and receive an answer instead of raw data. This saves access bandwidth and device resources.
To gain the benefits of an intelligent agent, IT teams can use basic IT or application design principles. To do so, enterprises must consider four steps.
Step one: Design thin-client, user-device applications with minimal local processing. It's risky to put too much intelligence in user devices, and creates version problems for mobile users. Linking a Web or API-based front end to a browser or another device element will force your application to offload processing and confine user data exchanges to questions and answers.
Step two: Think of application and information resources as tools that an agent can use on behalf of its user. Web-based applications create barriers to intelligent agent adoption when they combine multiple transactions into a single component. When creating an application, each step should be a separate component.
Many enterprises struggle to implement this step. However, RESTful design principles -- or treating all information and processing as an inquiry/response resource -- move others in the right direction. A job is a series of inquiries and responses, and this structure can be converted for an intelligent agent to use.
Step three: To evolve application front ends toward intelligent agent functionality, define front end tasks that respond to specific user requests. Normal transaction processing is workflow-oriented and involves search and updates. Intelligent agent processing must be need-driven and thought of as a simple Q&A. For example, a worker's question might involve looking up a customer name to get the transaction number, or reviewing open orders and account information. That single worker's request creates several transactions, but must generate a consolidated response for the worker. The agent process, which becomes the application front end for the worker's activity, is responsible for that conversion.
This step requires question interpretation and structured response development from composite information. Expand basic agent capabilities for text recognition and response without introducing additional speech recognition and generation problems. It's helpful to develop a guide for workers that specifies acceptable question forms. Enterprises can incorporate additional question formats until the dialog becomes rich.
Step four: Apply AI principles to these evolved front-end processes when necessary. Start with speech recognition and voice response first; this should be relatively easy if user dialogs are laid out properly.
How far should enterprises take free-form inquiry? True AI isn't voice input and response -- it's using machine analysis for question-and-answer generation. Each question is a series of inquiry and response interactions with application components or database elements. The purpose of AI is to create human-like interactions, but the goal is to enhance worker productivity. Poorly formatted question-and-answer interactions will quickly destroy any semblance of productivity, so start with basic recognition of general questions. Expand recognition as workers and enterprises understand use cases and productivity gains.
Each step toward personal agent consumerization in the enterprise speeds the pace at which workers accept and, ultimately, expect it. CIOs can benefit to get ahead of this trend.
About the author:
Tom Nolle is president of CIMI Corp., a strategic consulting firm specializing in telecommunications and data communications since 1982.