Previously, we described our process for enabling disciplined innovation at Centric. Now, we’ll share how that process led to a new product, the Centric Virtual Assistant (CVA) bot.
Innovation is a cornerstone at Centric. Because it is so important, we created a disciplined innovation process at the company. Once we established this process, our Innovation Advisory Team was eager to identify a proposal that solved a real business problem while providing opportunities for innovators to work with new technologies.
After many meetings and engaging discussions, the Centric employees in the Innovation Lab narrowed the final contenders to four. They then formed teams to create proposals for each idea before voting for the winner on a weighted scale. The results were close, but in the end, one innovation project emerged as the victor that would move on to development.
The winning proposal presented a novel way to solve a common business problem: easily finding that one document you need among the thousands stored in multiple repositories, including Microsoft Teams, One Drive, SharePoint and more.
The winning solution would become known as the Centric Virtual Assistant (CVA). At its core would be a chatbot – or bot – employees could interact with to find their documents. Instead of having to know exact search terms, they could use more natural language to tell the bot what to find, and the bot would do the hard work of finding it.
While that may sound straightforward enough, creating the CVA required several developers and Scrum Masters to coordinate the work of Centric data architects, data scientists and programmers. Our teams clocked time as volunteer hours as they navigated the many steps needed to get the job done. Amazingly, before the CVA project, no one had worked with the tools used!
The innovation team completed their work in about three months, and while they had hundreds of tasks, we can break the CVA creation process into five major steps.
Step 1: Determining a Minimum Viable Product, What the Bot Would Do, and Its Future State
A common practice for innovators is first to create a minimum viable product (MVP). MVPs are products built with only the most essential features needed to get them functioning. Innovators can put the product in front of users to test their response, using that feedback to refine the new tool and create additional features.
To create our MVP for the CVA project, Centric innovators decided to limit its scope to one repository: Microsoft SharePoint. Their inspiration came from The Lounge, Centric’s SharePoint-based intranet site that is an entryway to the internal and external sites, applications, tools and documents we use for our daily work. The Lounge also connects employees to other internal SharePoint sites, including one that houses a centralized collection of business development assets, proposals, case studies and more. Recently, we used Viva Connections to integrate The Lounge into Microsoft Teams, allowing employees to begin using Teams as a platform.
The problem? While SharePoint is great at enabling collaboration, its search feature requires a specific information architecture, content definition and taxonomy. Centric employees often need to find documents based on document type or a specific phrase within a document, which can make SharePoint search a challenge. For the MVP, we planned to customize SharePoint search to enhance how our employees use it.
With the project’s scope limited, our innovation team created user stories to determine what it would do. Once the bot greets users and they enter their natural language search term, it would then communicate with Microsoft Azure Cognitive Search using a custom application programming interface (API). The custom API would make searching for metadata and full-text content easy, allowing Cognitive Search to crawl SharePoint, find matching documents and display them to the user.
For the bot’s future state, we could adapt it to use artificial intelligence to learn more about its users and store that data. The next time that user entered a search term, the bot would remember their previous searches and return even better results more quickly. In addition, AI could allow Centric to collect useful KPIs to report on bot usage and improve future searches.
Step 2: Cloning “The Lounge”
Before we could start creating the MVP, it needed a sandbox to play in. That made creating a clone of The Lounge’s SharePoint repository on our Microsoft Azure portal tenant our team’s first step. The clone site allowed innovators to work on duplicates of legitimate, real-life documents without affecting employees’ experience on the actual Lounge site.
Step 3: Designing the User Experience (UX) and User Interface (UI)
With The Lounge’s actual data secured, our team could start designing the bot’s user experience. Because of the Innovation Lab’s democratic process, they put four potential designs to a vote.
The winning design was a simple webpage, rather than a more complex approach that would allow the bot to “float” over The Lounge’s homepage. Instead, users would access the bot by clicking an icon on The Lounge’s homepage. Once on the bot’s page, they would begin talking to it through a Centric-branded interface, complete with the company’s colors and logo.
Each of the branding elements was a separate project to address. The team had to build and integrate everything – from Centric’s logo and colors to the avatar that personalizes the user experience.
Step 4: Building the Bot
The team now needed to build the “brains” behind the API. Fortunately, Centric India had already done work on that piece. The Centric India team’s first task was to select a platform for building the bot. After evaluating several options, they chose to build it with Microsoft Power Virtual Agents, a local platform that can be extended with the Microsoft Bot Framework.
Power Virtual Agents’ graphic user interface (GUI) made designing the bot go faster, but the challenge was to determine how it would interact with users, how it would retrieve documents using Azure Cognitive Search and present the results.
As the bot’s search engine, Cognitive Search was its most important component. Members of the innovation team back in the States built the custom API, which interprets Cognitive Search’s unique requirements and our own data storage approach. The API makes it easy for the user to simply provide a few search terms and some text content to get results, no matter how we store the data.
The stateside team then worked again with Centric India to create stories for various search scenarios and fed those into Microsoft Power Automate Flows to create virtual flow charts, which interact with Cognitive Search to determine which path the query takes to retrieve the right document and return the result.
Step 5: Putting It All Together
With the bot and the web page built, the only task remaining was to drop a link into the page’s chatbox, embedding the bot into the page. Our innovation team then began integrating the bot and troubleshooting, testing, fixing and retesting to prepare it to go live.
With CVA in place, guessing file names and scrolling through irrelevant results with SharePoint search is set to become a thing of the past at Centric. Instead, users can engage in a more natural, intuitive conversation with the bot to lead them to exactly what they want.
Of course, as the original proposal described, we expect CVA will continue to evolve as Centric employees experience it and we add more data sources. Though the project’s goal was to develop the MVP for Centric, not as a marketable tool, employee feedback and additional capabilities could one day make it a product that could benefit other organizations struggling to search multiple repositories. The Centric CVA could become an off-the-shelf solution Centric could configure for each client and provide additional value to our engagements.
Conclusion
Innovation is an evolving journey, not a destination, and the same is true for CVA. We will continue tweaking it, learning from our experiences, and finding more ways to make it a valuable tool for anyone searching for information across different databases.
But as exciting as CVA is, its true value for Centric is the innovation process that allows employees to bridge the gap between needs and solutions while learning more about new and emerging technologies. It helped us live up to the Centric Innovation Lab’s goal to help employees think differently about their work and look for opportunities to improve.
Our Centric Innovation Team
Beth Beasley, Bryan Butler, Nick Cocco, Eric Galluzzo, Chantelle Heaton, Sumit Jain, Donnie Kerr, Mukesh Joshi, Rahul Kumar, Faisal Malik, Uriah McCanna, Ryan McComb, Ted McKenna, Paul Newkirk, Chris Pommerenke, John Quinn, Carmelita Resh, Adam Sebetich, Kevin Shaffer, Joe Setiabuti, Keith Shaffer, Abhinav Singh, Raysa Suarez, Zach Tesler, Ben White and Shawn Wallace