Catalytic: ‘RPA is the gateway drug for AI’

The immediate benefit of RPA is that it can eliminate a lot of repetitive manual labor and free up humans for what they are better at. But there’s also a side effect. RPA helps enterprises create a standardize framework for capturing data about how they execute processes as well as data about how processes can get delayed or stalled.

“If you set up RPA the right way by instrumenting the process, it’s possible to gather data to use as the training set for machine learning,” said Ted Shelton, Chief Revenue Officer at Catalytic, in an interview at Transform 2019. “RPA is the gateway drug for AI.”

An RPA implementation not only captures the steps involved in a process into a bot script, it can also set up the framework for understanding how a process is affected by different variables.

For example, a capital expenditure approval process might have a very specific flow. Different individuals might be involved depending on the price of equipment being requested. By automating this process and tracking it every step of the way and the responses to different kinds of requests, it’s possible to capture data about the steps of the processes and what factors into requests being approved, delayed, or disapproved.

Once the company has enough data points, AI models can be created to make recommendations. For example, an AI model might suggest that a particular purchase is likely to be delayed by a request for a better explanation. This can help save time for everyone involved and improve the process.

“This is why RPA is the gateway drug, it allows you to instrument the process,” Shelton said. “It is not just about process discovery, it can also help make sense of the results that happen in the process.”

Expanding the idea of process

Historically, processes have been thought about as a pattern of tasks repeated by multiple people. Human process experts would map out these steps to create a logical map of the activities necessary to complete them. These have typically been limited to things like managing customer data or facilitating purchases.

But other things have not traditionally been mapped. For example, if Sally is retiring, a process engineer may not map out the steps to throwing a great party. This might be considered a process in the strict sense of the word because it involves specific steps such as sending out invites, recording RSVPs, reserving a space, and requisitioning supplies.

“Automation technology will allow us to treat a much broader array of activities as processes that can be automated,” Shelton said. It will also make it easier to recognize commonalities between activities. For example, a birthday party and end of year holiday party might have similar tasks.

And throwing parties might not just be good for morale, it could also boost the bottom line by making it easier to organize better sales events. For example, Chad Rich, a senior director at E. & J. Gallo Winery, said some of the more creative salespeople tend be better at organizing parties for wholesale wine buyers. Consequently, Rich’s team is looking at how they can create a party process that helps organize the details for larger sales events for the whole sales team. This involves managing details like ensuring they have enough wine lined up for the event, ordering decorations, sending out invitations, creating themed music playlists, and ensuring the event aligns with new wine product releases.

Better instrumentation coming

RPA still requires a lot of human expertise to make sense of how a process works. Automated process discovery can help make sense of activities in a larger process, but they are still limited to understanding individual interactions. “Today I can take a particular task out of the process and automate it, but I cannot map it across the enterprise,” Shelton explained.

Eventually, process discovery tools could instrument different aspects of the workplace using AI agents to acquiring data from meetings and phone calls using automated transcription and natural language understanding tools. “It is not an intractable engineering problem, but one where the costs would far outweigh the benefits,” Shelton said.

In the short run, he expects tools to capture high value aspects of these for things like improving coordination of meetings or reducing the overhead for salespeople in the field. “With the right technology we can eliminate the coordination overhead,” Shelton said. These could end up being the building blocks for capturing the data required to infuse AI into more pockets of the enterprise.