With all of the excitement around artificial intelligence (AI) and its rapidly expanding capabilities comes the fear that it will replace people’s jobs. What this concern fails to take into consideration is the distinction between the decisions AI now makes on its own and the judgment calls that belong to people.
Start with AI’s core strengths
Let’s start by focusing on what AI is best at: completing high-volume, repeatable work. Computers have been automating tasks for a very long time, but AI takes it to another level entirely. Where legacy computers needed very specific inputs and very specific coding to arrive at very specific outcomes, AI is more flexible. It can take an input, compare it to previous inputs, learn from them, and produce an output based on the patterns it identifies. This is similar to how people arrive at decisions, hence the term “artificial intelligence.”
Removing the need to tell a computer program exactly what to do with any input that it might receive is a major paradigm shift in how we approach automation. With technology like RPA, it is prohibitively difficult to identify every possibility, and then write rules for each one of them. It also introduces the risk that a rule was coded incorrectly, producing an incorrect output and leading to costly errors
Unlike traditional task automation, AI learns
AI also learns very effectively from its experiences of successful past transactions and any feedback it receives. Not only does this make the technology easier to implement, it also introduces the concept of “confidence.” Confidence is an AI’s unique ability to check its own work. Because it learns from history, it can evaluate the quality of the patterns it builds and uses to arrive at a decision. Put simply, if it doesn’t have enough history to produce a highly confident output, it can bring that output to a user for validation.
When you, the user, provide feedback, you are also providing another data point for that pattern, allowing the AI to make better decisions. This feedback is often referred to as “training.” Think of it like a teammate who isn’t sure what to do next, so they ask a colleague or supervisor for help and learn from their feedback.
But judgment calls require people
However, it’s important to remember that while the AI behaves similarly to a human, there is no “human-like” mind behind that calculation. People will always be necessary to provide AI systems feedback and to make the judgment calls that shape the business.
Not every decision is a judgment call. Routine decisions that follow a written policy, like approving a compliant invoice or expense, are decisions AI agents now make on their own, with every action logged and auditable. Judgment calls are different. They lie in a gray area where there are multiple defensible alternatives and outside factors that may not be fully captured by policy. A person’s role is to select the best of these, incorporating subtleties an AI might not understand.
How to partner with AI for maximum benefit
Here’s an example: An AI receives an invoice via email and begins to process it. It ingests the data, identifies the associated purchase order, and performs a three-way match. These are tasks, with a correct or incorrect result for each step. However, on this invoice, there is a price that is outside of the company’s tolerance. Now what?
First, the AI saves the finance team hours of work on the dozens of invoices that never need to be touched. It pinpoints the one line where there is a challenge on a single invoice, and routes it to the right person with the full context attached. Next, someone on the team needs to be involved to decide whether to update the purchase order or push back on the supplier to change the billing price. Dozens of outside factors could influence the best decision in this scenario, and people are better suited to picking the best alternative.
Balancing AI tasks and human judgment
This is why people will always be needed for judgment calls when working with AI. And that’s a good thing. AI can do routine work more quickly and accurately than people can, and people bring the context and accountability that judgment calls demand. It’s all about balance.
When you use AI to automate a process in your business, focus it on the work your standard operating procedures already describe. Ideally, that same automation resolves what your policies cover, recognizes when a judgment call is needed, identifies the person who needs to make that decision, and provides them with all the information available to help them make it. You get the best of both worlds.
Businesses need people. AI can free them.
Fears of any new technological development are understandable. They’ve been a part of every human leap forward. What we’re finding today, however, is that people are increasingly using AI, whether they’ve been directed to or not, in order to be more productive.
The key lies in recognizing the difference between a routine decision and a judgment call. Finance organizations that understand this distinction and implement a comprehensive AI strategy will thrive over their competitors by freeing people from routine work.
Providing a partnership between human and artificial intelligence means everyone wins. Businesses become more efficient, employees engage in more meaningful work, and customers receive better service. The future of work isn’t about AI versus humans. It’s about AI and humans together, each playing to their strengths.
