Introducing Insight’s New Virtual Help Assistant

Published 06/29/2018 02:48 PM   |    Updated 06/29/2018 02:54 PM
Throughout the past year, we’ve witnessed an explosion of next-generation experiences being created using chatbots. After all, working alongside bots makes our lives and jobs easier.
 
The rapid adoption of chatbots and the increase in trending content supporting them made some believe the paradigm shift would happen overnight — but we’re just getting started in this slow-burning bot revolution that will continue for decades.
 
It’s not incredibly difficult to build a chatbot. But creating one people actually want to use takes serious and thoughtful consideration. Not only are there massive technical challenges  —  such as understanding user intent  —  to overcome, but it’s a whole new canvas for user experience. For designers working on chat, the text itself becomes the only tool in the design kit.
 
To strike the right balance of agility, complexity, efficiency and scalability, we took a hybrid approach to create our chatbot and leveraged our partnership with Oracle to help make that happen. In this article, discover how we chose what to tackle first and why less is more.
 

The plan: Create a useful chatbot.

 
A chatbot that’s useful first manages the most common needs of today. A chatbot that’s useful always is intelligent, predicts interactions and is flexible enough to pivot in real time. This really aligns with our business imperative at Insight: to help businesses manage today and transform the future.
 
Getting a chatbot to solve today’s business challenges requires quick action, a lean design process and intentionally aggressive decision-making. Ultimately, we decided the more agile, the better for everyone — and that meant defining a minimum viable product to deliver now while continually iterating to stay relevant. With the plan in hand, it was time to choose the framework architecture.
 

The platform: A framework that empowers users

 
As experts in technology best practices, we demanded very specific requirements for our chatbot in order to serve our clients best. That’s why we chose to partner with Oracle for this endeavor and became early adopters of the new Oracle Virtual Assistant. This partnership gave us the freedom to customize an already intelligent solution, feed content to the bot from multiple channels without additional integrations, and give users the opportunity to switch from bot to human at their request.
 
The assistant offers a blended technology solution where Artificial Intelligence (AI) and machine learning work in harmony. The bot initially gains knowledge from content located in our help center, including frequently asked questions, product profiles and how-to articles relating to the procurement platform. Natural Language Processing (NLP) enables the bot to interpret and predict end-user intent, so each conversation feels more natural.
 
This type of technology structure allows the bot to collect content from multiple channels. Robust access to content gives the bot an edge to provide better answers to questions about products, order status types, general shipping times and more.
 
At any time, if a user wants to speak to a real human, he or she can choose to do so within the chat window. This was a must-have on our list. Oracle Virtual Assistant is one of the few products available that enables a seamless transition between bot and human.
 

Setting expectations about core competencies

 
Chatbots are still relatively new, and capabilities vary. It’s important to clarify upfront the purpose of the bot and address any limitations. That’s why we named our chatbot Virtual Assistant. Transparency gives users freedom to interact how they choose. Users are more likely to be patient with a learning bot when they know they’re interacting with one.
 
As users converse with the bot, it analyzes intent, pairs questions to answers and predicts conversation flow. If a user chooses to escalate to a human, the bot analyzes that conversation to gain understanding so that fewer conversations will need to be escalated in the future.
 
In the chatbot infancy stages, machine learning technology may require human intervention. Oracle Virtual Assistant provides a dashboard to continually monitor the bot and view interaction trends, history, issues and escalations. Analytics surrounding performance can be measured to compare bot and human interaction to define key performance indicators.
 
If the bot responds to questions incorrectly or recognizes an inquiry it can’t answer, the program will flag that conversation for review. Chat administrators can then use the platform to input additional standard messaging or refine existing messaging.
 
Setting expectations upfront provides users with an optimal chatbot experience that limits frustrations — while the machine learning framework ensures every future chat is better than the last. When the bot needs guidance on speech variety and natural conversation flow, human intervention on an as-needed basis enhances overall performance. This blended approach to technology is a win-win situation for all.
 

Developing a next-generation experience

 
Embracing emerging technology that provides real value creates nearly limitless future possibilities. As the chatbot evolves, so will the needs of our users. For now, we’ve created the right strategy, content and framework to answer the interactions clients are currently having with Insight.
 
The future of the chatbot will include harnessing the micro decisions we make on a daily basis and identifying new opportunities to help. Whether that’s turning our Virtual Assistant into a personal shopper who can create and send quotes, assist with orders or field communications for account executives, next steps include providing a solution where people need it most. Say hello to our new Virtual Assistant and help us create the road map for the future.
 

Interested in a custom chatbot?

 
In the digital age, if you’re not innovating, you’re falling behind. See how this technology can help your business soar.
 

Learn more

Is this answer helpful?