Why AI still needs humans in the loop

Artificial intelligence (AI) is a prediction engine. It can predict which movies we want to watch and voice assistants can predict what we're going to say. And, when we use AI in our personal lives, near-perfect answers are acceptable. Even if we only like a couple of Netflix's recommendations, or if Siri occasionally misunderstands a word, we'll still use those services.

In businesses, AI accuracy is critical

One can't say the same for AI in the enterprise. Near-perfect AI predictions lead to missed opportunities and money left on the table. Whether a business is deploying AI to manage risk in a lending portfolio or plan inventory for a retailer, AI applications must pass the accuracy test before they go live.

The machine learning paradox

It's no wonder, then, that many companies struggle to get results from AI. The challenge in enterprise AI applications is that 100% accuracy is often hard to achieve, making it difficult to replace traditional processes or methods with them.

But machine intelligence gets better over time. As more training data goes through the system, it learns and becomes more accurate. This means that AI applications must be used to become predictive, but predictions must be accurate for AI applications to be used. This is the machine learning paradox.

The AI imperative

Unfortunately, delaying AI initiatives while tackling the accuracy challenge isn't always an option. Businesses that don't invest in AI now will risk being disrupted by others that do. So, how does an enterprise take the first step?

A properly built AI application can get us to a reasonable starting point – predictive accuracy of 80% or more. If we use that 80% while working on the remaining 20%, we can get closer to the 100% mark. But how?

Humans in the loop

This is where humans come in. By combining an AI deployment with a managed service layer, an organization can manage the 20% inaccuracy and learn from exceptions. For instance, if an AI application for invoicing resolves an invoice with low accuracy, the system might route it to a human for quality control. Or, a computer vision application that predicts the cost of vehicle damage for an insurer may hand off low-accuracy photos to a human.

As a human processes these exceptions, the AI application learns and improves its algorithms to increase accuracy. This is why keeping humans in the loop is essential when fine-tuning AI applications.

Prediction, judgment and action

Although there has been much discussion about AI “taking people's jobs", we can also see that AI will augment other jobs and create entirely new ones.

Previous technological breakthroughs have freed up humans to focus on higher-value tasks – like moving from data entry to financial analysis – or entirely new lines of work. Even as AI gets better at predictions, the need to keep humans in the loop will grow because humans will still need to decide how to take action based on these predictions.

Augmentation instead of automation

To dive a bit deeper, AI is augmentative because it solves only one part of a three-part problem – prediction first, then judgment and, finally, action. Imagine a driver on a racetrack. If AI predicts that another driver is going to overtake them from the left, they must judge whether to stay the course, block, or accelerate. The driver then acts by executing the move safely and swiftly.

Similarly, magnetic resonance imaging (MRI) uses a variety of signals bouncing off internal body parts to predict what lies under the skin. But a doctor has to apply judgment to decide the best way to treat an injury revealed by an MRI scan.

The future of humans in the loop

Ultimately, throughout all of this change and innovation, humans and machines will continue to coexist. AI will augment human work while humans continue to give feedback to fine-tune the machines. No matter where the AI revolution takes us, one thing is for certain: there will always be humans in the loop.

The article was authored by Sanjay Srivastava, Chief Digital Officer at Genpact, and a version of this was originally published on Forbes Technology Council.