Sometimes, inspiration strikes in obvious places. Other times, it comes from those offhand moments where you’re focused on a very different goal. In truth, it’s hard to tell what this moment was.

At the WorkFusion Ascend conference in NYC, a presenter was talking about the best way to ensure a new bot or automated workflow was working appropriately: “Create a fidelity loop.” The conversation focused on building in a process to review and audit anything that you automate. Whether it is a system that the bot checks itself or having a human double-check your elements, you need a way to check your work.

In the realm of RPA and greater AI adoption, especially as you move into cognitive intelligent systems, this makes perfect sense. You can’t trust a novel system to fully understand the way it adapts to new or different information or problems.

We’ve seen plenty of sci-fi stories show AI run amuck as it decides the best way to keep things clean or prevent suffering is to simply wipe out the human race. Now, no RPA is evolving a la Skynet right now. However, they are constantly looking for ways to streamline processes and reduce manual tasks. If left to their own devices, they may make decisions that prioritize this above business intelligence the valuable data you use elsewhere.

Build a fidelity loop where processes are reviewed, and you can understand how improvements operate within your organization as a whole. AI systems require it, but maybe your non-automation efforts should include it too.

Revisiting Workflows and Operations

One of the core debates at Ascend was whether companies should work to streamline processes before they automate or if adopting RPA first would force people to streamline a process so it could be repeated. The latter comes with a bigger time crunch that some said may not end up with the best possible process, but at least comes up with one that’s useful.

Of course, as with any dichotomy, some tried to present a third option where companies could do both at the same time. It is possible but unlikely, simply because it requires a large number of people in the same room and agreeing before, during, and as large sums of money are spent. The agreement is difficult to achieve at any single stage, and arguably harder to maintain throughout all three.

So, What’s A Company to Do?

I found myself agreeing with the push to review workflows before a process is automated, all thanks to the unrelated call for a fidelity loop.

There may be multiple ways to perform the same process and achieve the same goals and results, but it’s likely that one option is better in the immediate and aggregate. Fidelity to the outcome can point out which at least works best for you. Honest discussions and reviews of the metrics will point to a proper tactic.

It’s an intelligent first step for using systems that can mirror your activities and learn to adapt to hiccups — you can get a solid RPA toolkit through WorkFusion’s Automation Academy with some clear examples.

That critical eye is also relevant and useful for reviewing the processes that automation and RPA will impact. When you clear up time for your team, how do you best respond? Where do you reinvest when you get a 10x ROI? What do you no longer need?

It also should be seen as a call for the industry to be honest. In the recorded sales pitches, you’ll rarely hear anyone note that RPA can reduce workforce requirements, but we all know it’s there. Gartner estimates presented at the conference said that a team of dozens can manage thousands of bots.

Fidelity is important, and it’s something we owe the space and all the people who work in it.