Johnson Controls is an example of a company that has its automation act together. The company has been around the block with robotic process automation (RPA) but started encountering use cases involving unstructured data that RPA couldn’t handle. After taking stock of its arsenal, it realized it was missing a tool with artificial intelligence capabilities that could take on use cases requiring intelligent document processing.
The use case that pressed the issue was invoice processing, according to Chris Engel, Senior Manager, Robotic Process Automation and AI, for Johnson Controls (JCI). Engel spoke to Brandi Corbello, VP of business development at Indico, for a webinar on automation technologies. Corbello was a natural choice as she is an automation expert herself, having until recently served as VP of Transformation for the commercial real estate giant Cushman & Wakefield.
Their discussion covered the possibilities inherent in automation use cases involving unstructured content, from those that are pressing today, like invoices, to others on the horizon, including business modeling and predictive analytics.
Related content: Achieving unstructured data automation with RPA
Taking on an “impossible” invoice task
JCI is a multinational focused on infrastructure and technology for buildings, from industrial HVAC systems to fire suppression, building security and digital solutions and services supporting smart building technology.
As such, it deals with “hundreds of thousands” of invoices per day across the organization, worldwide, Engel said. While some may think of an invoice as a structured document, and hence a candidate for RPA or template-based solutions, that is simply not the case at JCI’s kind of scale.
“Imagine trying to keep tens of thousands of templates that are all geared toward Grainger invoices and Fastenal invoices and so on. It’s just impossible,” he said. The Indico Unstructured Data Platform enables you to create models based on the “concept” of an invoice; not one geared to any particular vendor. “So, the next invoice that comes along from a vendor we’ve never seen can still run through the system.”
JCI’s pilot involved an RPA-based front end to take in PDF invoices and enter them into the company’s accounts payable (AP) workflow system. From there, the PDFs were manually opened, with relevant data extracted and cut/pasted into the company’s ERP system. The pilot focused on replacing that manual step with an IDP platform, using it to pull out relevant fields and values from each invoice, enter them into the ERP system, and find the matching purchase orders and the goods-received notices.
“If they all match up, you wind up having a touchless AP system,” Engel said.
In practice, exceptions in the business process and other “internal workings” sometimes thwart the automation routines. But the pilot proved the technology works, he said.
“Our goal from an IT standpoint is to get up in the 90s in terms of a percentage being handled hands-free,” Engel said. “I’m confident we’ll make it to 90%, probably by the end of the year.”
Applying analytics to unstructured data
Corbello noted automation projects do tend to bubble up issues with business processes, but that can be a good thing. “Automation in general, especially in larger organizations, does expose some of these business exceptions and really forces us to think about our business processes,” she said.
JCI is also thinking about how to apply IDP technology to unstructured data for future use cases, many of them involving analytics. An IDP platform can take unstructured data and turn it into a structured format, enabling companies to then apply powerful analytics technology to it and gain new insights.
“We’re also looking at a couple of use cases around supply chain modeling and being able to do predictive analytics on some of the information and the supplier metrics we get,” Engel said. The idea is to understand where things are trending and potentially take action to prevent issues.
Another is being able to use an IDP platform to normalize contracts, whether from vendors or clients, and better understand their terms and conditions to identify risks.
Similarly, around contract renewal time JCI could examine contract terms and marry it to data around customer churn to identify clients that may be at risk of changing suppliers, along with upsell opportunities.
“Those cases are there, certainly,” he said. And he expects use cases will only grow as the tools continue to mature in their AI capabilities.
Those are just some of the highlights of Corbello’s interview with Engel. To learn more about why he’s so bullish on intelligent document processing in general, and Indico Data in particular, check out the full interview below.