Data Action Layer April 22, 2022
Automation isn’t about the replacement of human talent. This is a conclusion that many businesses have reached as hyperautomation has been taking off across the globe. Instead, they’ve realized that the purpose of technologies such as robotic process automation (RPA) and intelligent document processing (IDP) is to reinvigorate operational efficiency by augmenting human employees. Although straight-through processing (STP) can work in some instances, there are many increasingly complex tasks businesses want to automate that require oversight, input, and guidance.
McKinsey found that businesses viewing automation as more than just a way of eliminating jobs are far more likely to meet their goals. A landmark 2018 Harvard Business Review School study came up with similar results: firms where humans and machines worked together saw the highest financial performance gains. No wonder many experts recommend implementing automations with humans in the loop.
What is human-in-the-loop automation? It’s when you automate a formerly manual business process but maintain human intervention in critical stages along the way to provide some missing data, do a validation, make a judgment call, correct an error, review or approve an action, or otherwise influence how the automation will proceed.
You’ll want to keep humans involved because automation technologies—even those considered intelligent—do best when the tasks are well defined. They don’t adapt well to change, nor can they consider unexpected events or facts. Even the best AI models, for example, need to be checked continuously because they “drift” as they encounter new data and situations. Therefore, keeping your human problem-solvers involved is a smart move.
Initially, continuing to keep humans involved in what should be a fully automated process may seem counterintuitive. But there are many situations where humans are needed to move the processes forward. Here are a few that are top of mind for most businesses.
The automated model (or script, or bot) often needs more data. This could be highly-variable data or information not available in digital form or data that only humans can infer—perhaps from a voice recording or video—to inform the context of a situation. Keith McCormick, the chief data science advisor at CloudFactory cautioned in his blog that in such cases, the human should not necessarily override what the automation is doing but just provide the information required. “The human in the loop should be there to fix, modify, provide, or clarify the inputs to the model,” McCormick said, “Not to override the model.”
This is the classic situation where you need a human in the loop. For example, say you’re leveraging automation in the accounts payable department to process hundreds or thousands of invoices a month, and some of the invoices are missing critical information like the purchasing number (PO). An intelligent document processing solution may have rules in place to flag these types of anomalies. Still, a human will have to decide whether to proceed with payment, manually add the PO number themselves, or reject the invoice until proper documentation is provided.
Some processes are so important or trigger such critical actions that you simply must put a pair of human eyes on them before allowing them to proceed. A “bad call” could result in hefty fines and legal issues in highly regulated sectors, like financial services. Having a human in the loop offers additional validation, extends accountability, and ensures compliance.
It makes sense to establish frequent check-ins on the results your automation technologies are producing. If, over time, more and more prospective clients are being rejected by your know-your-customer automated system, it’s probably time for a human to step in and fine-tune some of the rules.
Critical thinking is a human superpower, while bots live by the rules we set in place. When Covid hit, employees leveraged their problem-solving strengths to adapt their existing automation processes and implement newly automated workflows quickly and effectively. This scenario wasn’t something that could have been predicted and programmed ahead of time. Therefore, keeping humans in the loop is an excellent choice for companies seeking operational agility.
Automation technology is indispensable in today’s business world, but so are your employees. So stop considering them as separate elements and start thinking of them in complementary terms. Because when you combine the 24/7 operating power and efficiency of automation with the creativity, analytical thinking, and decision-making of your people, you’ll be able to reap more considerable benefits and extend your competitive advantage.
You already have the people, now add some workflow automation technology into the mix. Start your free trial today.
Fine-tuning is not the only way to get relevant, domain-specific responses out of an LLM. Alkymi’s team of expert data scientists explain an alternate route.
Find out which type of automated document processing solution is right for you: data extraction, an IDP, or a complete business system for unstructured data.
We’re partnering with Portfolio BI, a provider of portfolio analytics and reporting solutions, to bring structured and unstructured data sources together.