Amazon Web Services (AWS) announced the general availability of its Augmented Artificial Intelligence (A2I) service, which is designed to streamline human review of low-confidence artificial intelligence (AI) workloads.

The managed service is designed to make it easier for users of Amazon's various AI capabilities to implement human review and improve the overall accuracy of machine learning (ML)-based image analysis, optical character recognition, transcription, and spoken language. “This interplay between machine learning and human reviewers is critical to the success of machine learning systems,” Amazon claims.

Until now, Amazon says implementing human review, while necessary, has been a difficult and time-consuming process. “Developers sometimes spend more time managing the human review process than building the intended application,” the company said.

To remedy this, A2I provides developers with more than 60 human-review workflows for common ML tasks, allowing predictions from Amazon’s Rekognition and Textract services to be verified more easily.

Amazon explains that developers can choose to have AI/ML predictions below an acceptable threshold reviewed by Amazon’s massive workforce of independent contractors working for Mechanical Turk, or by third-party organizations.

While A2I aims to help developers improve ML code with low confidence scores — those where the model is essentially making an informed guess — Amazon says there are also cases, such as those in law enforcement, where human review is “strongly recommended” even when the confidence score is high.

Amazon’s customers say that “there are critical use cases where human judgment is required, like in law enforcement investigations or times when human review can be used to resolve the ambiguity in predictions when confidence levels fall below a given threshold for less sensitive use cases, and the current human review process involves a lot of custom effort and cost,” said Swami Sivasubramanian, VP of machine learning at AWS, in a statement.

One use for this technology would be to validate the accuracy of ML used to extract financial information from scanned mortgage documents or to identify counterfeit items online.

Over time, the cloud provider says this human touch can help to improve ML algorithms and improve their overall confidence rating.

A2I is available in multiple markets across the United States, Canada, Europe, and the Asia-Pacific region.

Alongside the announcement, Amazon also highlighted several enterprises and organizations, including the United Kingdom's National Health Service (NHS), T-Mobile, Deloitte, and Belle Fleur, which expressed interest in using the cloud provider’s AI technology and managed human review services.

“The NHS is investing in the promise of AI to improve the quality of public healthcare across the UK. Human judgment is critical and in fact, is often required for decisions involving medical payments,” said Chris Suter, head of cloud platforms and innovation at NHS BSA, in a statement.

Domestically, T-Mobile, which recently acquired Sprint, is using Amazon’s AI technologies to improve its customer service experience. In a statement, T-Mobile ML engineer Heather Nollis said the carrier is already using ML to provide “relevant information, such as account details and available discounts, in real time to our customer care agents while they are in live conversations with customers.”

Using A2I, Nollis said the carrier will be able to ensure that T-Mobile’s models deliver quality insights by having humans validate random samples.