Over the past several years, imaging giant Xerox has added automation to its comprehensive lineup of products and managed services. The Norwalk, Conn.-based company has combined robotic process automation (RPA), intelligent document processing (IDP), AI and generative AI to capture, digitize, store, retrieve, understand and classify its clients’ documents across a wide range of industries to automate—and wring cost and time—from their processes.
To accomplish this, Xerox has implemented a multichannel capture strategy leveraging its imaging centers, third-party offsite facilities, and on-site scanning at client locations as well as their multifunction devices. By applying advanced intelligence, Xerox enables clients to understand, classify, extract, and efficiently process physical as well as digital sources such as electronic documents, emails, and web forms. They can then use that data to automate complex workflows. Shivani Agarwal, Vice President of AI and Intelligent Automation, and Miguel Nunes, Global offer lead for Intelligent Document Processing, have a shared vision that includes offering Xerox’s own end-to-end automation and document processing capabilitiesby leveraging market-ready tools from best-in-class providers in a Global Automation Platform that delivers secure, cloud-based, AI-enhanced document capture, workflow orchestration and process automation to its customers.
Existing Managed Print Services (MPS) clients will enjoy even more advantages. They can seamlessly connect their print ecosystem with this advanced workflow automation space through a distributed capture app on their Xerox® ConnectKey® multifunction devices. The app integrates with their business processes, serving as an additional inbound method among the multiple ways to inject documents or data into the workflow, enhancing efficiency and connectivity.
Integrating technologies for end-to-end document management and process automation
According to Nunes, Xerox has built a proprietary workflow orchestration management tool that enables the company to control services from end-to-end. The intelligent document processing piece, which Nunes has been heavily involved with, has added a wide variety of capabilities to the tool, including semantic analysis and language detection.
One of the early use cases for the capability has been important to banks and other financial services organizations such as Insurance companies.
“We’ve been working on building models for document classification,” he says. “Several of them are for fraud detection, which use what we called similarity and matching. The models use the functionality to detect image tampering by matching faces on ID cards to photos, check a proof-of-address against a database or detect particular objects in ID cards.”
One lender used the platform to ingest 50 million paper and electronic loan application documents per year. Using similarity and matching for fraud detection in addition to RPA, computer vision and machine learning, the client was able to process 138,000 applications each month, reduced paper applications from 80 percent of the total to 40 percent and yielded insights from the data that enabled the lender to streamline loan processing and help reduce credit risk.
On the RPA side, Agarwal says Xerox’s RPA technology partners perform a range of automation activities from simple tasks to complex AI-driven document understanding that enables classification, extraction and language recognition.
She cited several case studies that spotlighted how AI-enabled RPA supercharged process automation for a Xerox client. One company that provides support services for law firms wanted to automate a manual process that received requests to create deposition notices.
“By nature, deposition notices are extremely unstructured,” Agarwal explains. “So, we used document understanding solutions to find specific fields within that document. With the availability of LLMs, we can query the document for specific information that we need. Who is the client? Who is the deponent? Who is the lawyer? Does this notice have a signature? When does the meeting need to be scheduled? Do they need a translator? We can extract all the necessary information that we need from these unstructured documents and then use RPA bots to push this information into the clients’ system from where an Outlook invite gets sent to the required parties.”
The RPA solution reduced processing time from an average of six minutes per notice to less than 30 seconds and the number of people required to do the work from 27 to 2.
Staying ahead of the curve
AI has radically changed the nature and capabilities of data-intensive business process automation in a very short period of time. Agarwal and Nunes have been on the right side of the technology curve, anticipating and incorporating AI and RPA and developing other features that make the platform better. But technology will always evolve, and the pair understand that continuing education and trend spotting will be vital to their efforts to improve the platform.
One way they can do that is to use Xerox itself as a test bed.
“Most of our clients are thinking about adopting generative AI,” Agarwal points out. “There is a lot of talk, but there are not that many use cases. We have been able to test all these technologies and execute use cases for Xerox itself. In our call centers, for example, we use generative AI to read all incoming emails and classify them as a billing-related query, a service-related query, a supply-related query or break-fix issue. That classification happens automatically, and we can create tickets in our CRM systems faster and get those issues resolved faster. We did this at Xerox last year when the technology was very new and just being talked about, so this year we were able to market this to our clients and win some business to implement similar use cases for them.”
In addition to being able to generate their own proof-of-concept for new technologies, Nunes notes Xerox has a rigorous client feedback process that often leads them to new technologies that are in demand in the field. It’s important, he says, to always keep an eye on the future.
“The idea is that we always have our road map in mind,” he concludes. “We want to include all of the things our clients ask for, create use cases, discuss with clients and then try to build, evaluate, or buy the technology we need to make those changes happen.”
