How Atom Ai and RFP Assistant have Evolved and What it Means for the Future

In a prior WWT Research Note, How WWT is Harnessing Generative AI to Drive Internal Business Value, we gave readers a glimpse into our company's internal approach to AI adoption. We outlined our methodology for assessing and validating generative AI (GenAI) use cases and shared how we were beginning to apply that framework to several internal use cases, including the ATC Assistant and the RFP Assistant.

This Research Note shares some updates from our AI journey thus far, with a focus on how these two intelligent digital assistant applications have evolved since they were first conceived. Both applications leverage WWT's internal data sources to create custom solutions for different business needs.

As a quick refresher, we designed the RFP Assistant to streamline the RFP intake process via embedded large language models (LLMs) that can quickly summarize, qualify and RFP answer questions and requirements based on an existing library of curated information. The business goal of this application is to reduce the time and effort required to respond to proposals. Similarly, the ATC Assistant was initially conceived as a way to extract insights from the valuable data collected every day in our Advanced Technology Center (ATC) — a unique lab environment that our clients, partners and internal employees leverage to validate, integrate and deploy next technologies.

Let's explore how these two GenAI-powered applications have evolved in the last six months and see what impact they've had on their respective lines of business.

The evolution of two GenAI applications

The ATC Assistant becomes Atom Ai

Of the two AI-powered applications in question, our ATC Assistant has undergone the greatest evolution. The drive behind this evolution was to create an even more robust chatbot that can help clients, partners and employees get the information they need about WWT more efficiently. We're currently calling this new version of the ATC Assistant "Atom Ai" (formerly WWT GPT), though the application will likely be officially branded prior to external release.

So how did the ATC Assistant evolve into Atom Ai?

Data source evolution

Rather than limiting the ATC Assistant's training dataset to two primary sources (proof of concept data from ServiceNow and our ATC Repository), our experts saw value in widening the dataset aperture to include all wwt.com. Atom now taps three internal data sources to educate WWT employees, clients and partners about our capabilities.

In addition to adding a new dataset in wwt.com, we also added the ability to convert natural languages to SQL queries and leverage SQL functionalities to extract ATC project insight from our ServiceNow database.

Target audience evolution

The target audience of the ATC Assistant was originally defined as WWT ATC engineers and managers. Atom Ai, on the other hand, has been designed to cater to more than 4,000 WWT employees across sales, services, operations and other teams. We're already seeing our external sales teams use Atom to:

Quickly locate information about WWT's current capabilities and services
Research and reference relevant case studies by technology and industry
Cross-reference WWT-related or niche technical terms
Draft emails leveraging internal data that would normally require a huge manual lift to find

Business goal evolution

The original goal of the ATC Assistant was an intelligent application that could provide detailed and specific insights into WWT's wide range of proof of concept (POC) results. With its expanded access to wwt.com data, the underlying goal of Atom is to increase employee productivity and provide insights across five main use cases:

Research WWT's services and capabilities
Research past projects and case studies
Draft external-facing materials
Identify the point of contact and subject-matter experts for a certain domain
Research technical domain expertise with WWT's perspective embedded

The RFP Assistant learns to generate proposal responses

The RFP Assistant has undergone its own evolution in the last few months.

Functionality

The previous version of the RFP Assistant focused on providing a summary based on uploaded RFP document(s), plus an initial qualification score based on Salesforce data from similar past RFPs with regards to industry, customer and technology.

The RFP Assistant has since evolved its capabilities to be able to generate a first draft of a multipage proposal response with great user control. For example, our Proposal Managers can now use the RFP Assistant to:

Generate an outline based on uploaded RFP documents
Generate customizable prompts based on each section
Draft a final response leveraging internal WWT IP

Data sources

The only data source of the previous version of the RFP Assistant consisted of manually uploaded RFP documents. The latest version leverages five data sources to generate RFP responses that strive to incorporate the key aspects of WWT's messaging, knowledge and expertise. These data sources include:

Past proposals: The RFP Assistant leverages historical information, verbiage and formatting from hundreds of past proposal responses to cut down on the proposal response time.
Q&A pairs for questions: Our Proposal Team maintains a rich library of proposal questions and responses. These cover some of the most asked-about topics, often having to do with company information, achievements and capabilities. This rich data source is maintained by the Proposal Team as a library of information to inform proposal responses.
Long-form common question responses: The Proposal Team's library also contains a collection of longer-form answers to questions based on technologies, industries and services.
Customer reporting (e.g., BVAs/BVRs): WWT often reports directly to clients in detail about the specific ways our solutions and services have helped drive business value. This often takes the form of a business value assessment (BVA) or business value report (BVR). Including these data sources helps the RFP Assistant application inject specific business value into its responses.
wwt.com: Data from wwt.com has been used to train the RFP Assistant to leverage WWT's thought leadership messaging on various topics. It gives the AI tool the ability to reference previously published case studies as well as information about our marketed services and capabilities. This data source gives the tool access to our web-based sales materials, like sales enablement pages and solution area information. And finally, wwt.com is home to an array of company-related news through press releases that touch on everything from our partnerships to new capabilities to achievements to environmental commitments and more.

Impact and what's next

Atom Ai is still currently in beta testing mode with around 4,000 employees having access to the tool.

In our initial A/B productivity testing, we're seeing productivity gains north of 30 percent across designated use cases. This represents a real and significant time saving for WWT employees. We plan to give Atom access to even more data sources that cater to both internal and external use cases.

Our Proposal Team is currently incorporating the RFP Assistant into its daily workflow. As they continue to test and refine the application, they expect that it will significantly reduce the time it takes to qualify and respond to RFPs throughout the year. As the next steps, we will keep improving the current version's ability to summarize, qualify and generate accurate proposal responses and provide more granular control to the proposal team.