Artificial Intelligence Overview
Artificial Intelligence (AI) is a rapidly growing field for business organizations. Findings in the latest Vantage Partners’ annual executive survey, first published in 2012, indicates that an overwhelming 97.2% of executives report that their companies are investing in building or launching big data and AI initiatives.
Current AI fields can be separated into two different categories: General AI and Applied AI. The former involves broad applications of AI, such as computers having the ability to conduct problem solving, similar to a human. Whereas, the latter is an application that is meant to perform a specific task, like suggesting new music or finding the shortest route for your drive.
At Intersys, our highly skilled consultants can design and deploy real world AI applications to solve a variety of problems. Some of the AI projects we have recently completed for clients include developing AI chatbots for business, usability testing powered by cognitive models, and natural language processing platforms.
Artificial Intelligence Use Case #1 – Chatbots for Business
Outline of Concept
Implementing an automated conversational bot (chatbot) into your business logic can increase user acquisition, retention, and sales. Mayor social media platforms are developing APIs for businesses to connect their bots. Those capable of taking advantage of these APIs and platforms are able to expose their business to millions of users. A good bot design can have an enormous impact on user experience and satisfaction due to their natural way of communicating with the end user.
Chatbot technology offers an alternative solution to mobile application development. Nowadays, phone storage space is a highly competitive area (e.g. for most users, downloading an app requires uninstalling another). Chatbots can leverage social media apps as a channel to expose a business. Nonetheless, the challenges of building a chatbot are considerable – from system architecture to machine learning.
Businesses always strive to enhance user experience and customer service. Nowadays, technology empowers us to address this problem in creative and unique ways. New advancements in machine learning and data engineering enable us to create highly reliable intelligent systems capable of engaging with users at a personal level and providing high-quality customer assistance.
We are now seeing increasing interest in the development of this technology in areas related to virtual assistants, customer service automation, and sales chatbots. Nonetheless, the development of such systems usually requires a significant investment: A huge team with domain experts and developers, maintenance of cloud services, and considerable development time.
Intersys Solution and Application
At Intersys, we identify the challenges in developing an intelligent chatbot that meets your business needs. We’ll team with your the engineering workforce to build, customize, scale and deploy chatbot projects. This allows you to focus on providing the specifications for building your bot according to your business needs, instead of maintaining a highly complex engineering project. Our team is capable of integrating natural conversational capabilities to your bot and customizing the question-response flow (rule-based chatbot or intelligent bots). Furthermore, chatbot projects developed at Intersys can be seamlessly connected to the most popular social media platforms such as Facebook Messenger.
The technology stack used in this project was selected to guarantee performance and scalability. We use the Scala programming language with the Akka Toolkit to follow the reactive manifesto (i.e., elastic, responsive, resilient, message-driven). Our system architecture integrates Kafka for publish/subscribe messaging patterns and Cassandra for message persistence. Our machine learning algorithms run using Spark, Akka, and Tensorflow.
Artificial Intelligence Use Case #2 – Usability Testing Powered by Cognitive Models
Outline of the Concept
When interacting with a User Interface, users can execute three kinds of actions: cognitive, perceptual, and motor. A combination of these actions produces a specific outcome, for instance, logging in to an app. A User Interface that provides a seamless journey for the visitor increases conversion rates.
Customer/ Business Need
Increase conversion rates – there can be macro-conversion (create account, download demo, purchase, etc.) and micro-conversion events (spend more than 1 min in a particular page, navigate more than three pages, etc.). Increasing conversion rates for specific actions on a website or app can lead to higher user retention rates and increased revenue for businesses of all types.
Intersys Solution and Application
A cognitive model is able to produce a sequence of cognitive, perceptual and motor actions with three characteristics: (a) the latency of actions is plausible according to what we know about how the human mind works; (b) the computational limits of the model are set based on the known limits of our cognitive system and (c) a comprehensive record of every action is produced.
It takes around 8 seconds to perform a login to a page (inserting your login and your password and press enter). A cognitive model produced 343 actions during this time (255 cognitive, 40 visual, and 29 motor).
Such a detailed account of behavior represents an opportunity for UX: By analyzing the sequence of actions, it is possible to detect opportunities to improve the interface. Some generic examples are these:
- If several milliseconds are spent looking for a particular element in the interface, a change is needed to make this feature more salient
- If several cognitive actions are required to produce a decision, a key piece of knowledge may be missing; hence particular training may be needed or a hint can be provided to the user
- If a considerable amount time is spent on manual actions, such as mouse pointer movements, the layout can be changed.
ACT-R (cognitive architecture), Cassandra, R, Python, Java/Scala
Artificial Intelligence Use Case #3 – Natural Language Processing Platform
Outline of Concept
Develop a Natural Language Processing platform capable of ingesting, storing and analyzing large amounts of text and audio. Results are made available through a collection of microservices, which combined represent the servicing layer of the platform. The platform is capable of ingesting batch and streaming data. The servicing layer support publishing feeds in batch and streams, as well as on demand restful services. Supported services include sentiment analysis, thematic analysis, voice recognition, intent classification, entity extraction, and auto-summarization. The platform also provides full text search on the unstructured text data.
Businesses have a huge amount of Natural Language data but have long struggled to extract insight and derive value from this data. Using Natural Language Processing, businesses are able to better know, understand, and react to their customers, their employees, and to public opinion. Examples of data sources include call center recordings, chat messages, emails, social media, customer reviews and feedback, and customer survey responses.
One example use case is to convert call center voice conversations to text, then preform sentiment analysis, thematic analysis, and auto-summarization to each interaction.
This enables use cases such as:
- Following up with a customer to proactively upsell/cross-sell additional products.
- Flag poor customer interactions for additional follow-up with the customer by a supervisor. These flags can also be used to provide additional quality assurance and coaching of customer service agents.
- Summarization of calls can help with workforce planning and training. It also helps deliver insight into why people are calling the customer service center. For example, are people calling about billing, technical support, to ask general questions? This in turn can also be used to help fine tune IVR systems.
Intersys Solution and Application
Intersys Consulting will engage with you to deliver the overall platform as an end to end solution. We have developed IP in the domain of NLP which accelerates the time from investment to value. We will partner with you to ensure that all components of the application pipeline are integrated and implemented. We support on premise, cloud and hybrid deployments, and have connectors to integrate with a wide range of technologies.
Java/Scala, Python, Kafka, TensorFlow, Spring, GPU computing, ElasticSearch