Transforming Customer Support Triage with Simple Automation + AI

  • 29 February 2024
  • 1 reply
Transforming Customer Support Triage with Simple Automation + AI
Userlevel 1
  • Most Valuable Pathfinder
  • 0 replies

Many people may assume Intelligent Automation can only execute rule-based tasks and processes. But the recent onslaught of AI technology is challenging—nay, enabling—us to shift what we ever thought possible with Intelligent Automation. The unruly, unstructured tasks can now be streamlined with automation.

I joined TriNet, an organization that provides businesses with full-service HR solutions, in 2018 when it had just completed an initial POC use case with Intelligent Automation and received funding to launch a new automation program. Since then, I've helped expand and mature the program with the help of Automation Anywhere's innovative tools. Until now, our toolset mainly included AARI and API Tasks. However, like many of you, since AI came onto the scene as a cutting-edge tool to bolster automation efforts, I have been eager to find a use case to POC its viability within our organization. Today, I'm sharing the use case in which we were finally about to debut the power of AI and how it unexpectedly caught fire across the organization.

The Use Case


The TriNet Customer Experience team has a central intake for support cases in a CRM tool. From this central intake, two different teams manually triage cases out to other specialized teams within the department, depending on the criteria they identify within each case. A lot of manual review often meant slow turnaround times, and manual hand-offs opened up the potential for error in translation. With this use case, we recognized a relatively simple solution with automation, as well as an opportunity to leverage AI to optimize the process even further.

The Solution


First, we connected the CRM tool through API for an automation to pull the support case files from the central intake. Then, we added an AI layer on top of the automation to parse each support case and determine to whom it should be routed. To accomplish this, we worked with our internal Analytics team to train a predictive model using a plethora of data and keywords and set parameters of a confidence threshold. Upon analyzing a file, if the predictive model's confidence level meets our threshold, the case file is updated in the CRM tool and sent to the appropriate specialized team.

Security Considerations


The biggest hesitation for most organizations to integrate AI technology is security, and that was no different for us. Firstly, we spent a great deal of time training the predictive model with substantial data. It was tested and tested and tested quite a bit before we began using it. Also, we felt this particular use case was a safe initial POC because the model and data are only used internally and, in the slight chance a case be routed to the wrong specialized team, the data is not highly sensitive. The data is not shared over the internet, and the only people accessing the predictive model are within the TriNet network. Finally, we restricted access to only the automation ID calling the model. With all of these considerations thoughtfully covered, we have confidence in the process and have had no security issues arise.

The Impact


Besides the pride of executing our first AI use case (which is a reward in itself), we’ve seen this process bolster the ROI our program is generating significantly. And what we didn’t expect with this specific use case was company-wide impact. This seemingly simple solution designed for the Customer Experience team is actually highly reusable, and 40 different teams across the TriNet organization are now relying on this automation.

Make AI Work for You


AI is a sophisticated, sometimes complex technology. But putting it to work for you doesn’t have to be complex, especially with the tools available from Automation Anywhere. We took a simple automated solution, beefed it up with just a hint of AI, and created a massive impact. I encourage you to step out of the rule-based mindset and reconsider how you could optimize even the simplest solution with AI. I bet you’ll be pleasantly surprised with the results.

1 reply

Userlevel 5
Badge +9

Love this. Great breakdown @Eric K! Not only did you clearly explain the situation your team was facing and how you approached it, but also the guardrails in place to make sure that incorrect decisions made by the automation wouldn’t have a negative impact on the business or the organization’s reputation. Thanks for sharing your lessons learned and the impact of your AI-powered automation!