There’s no doubt that artificial intelligence is changing the world. AI’s potential to automate many tasks that are being done by humans is unparalleled, and it is expected to contribute $15.7 trillion to the global economy by 2030. With such a wide-reaching impact, some roles will inevitably be made redundant by businesses adopting AI into their operations. Some experts believe that data entry is one of these jobs, due to being composed mostly of simple, repetitive tasks.
Many businesses are turning to AI because of increased speed and cost savings, and it allows the business to allocate its human resources to other tasks. However, AI is not a silver bullet that can perform all needed tasks without the need for human intervention. Alan Bandell, Founder and CEO of Axion Data Services, an industry leader in providing data entry outsourcing services to companies all over the US, believes that the future of data entry will be intertwined with and majorly transformed by AI, but it won’t make human expertise obsolete.
Since opening its doors in 1996, Axion Data Services has seen how technological developments have led to huge changes in the data-entry industry. According to Bandell, the company has lost a number of clients over time, due to the clients’ business decisions in response to changes in technology and business environment.
Despite these challenges to the business, Axion is continually adapting with the times, drawing from the Japanese concept of kaizen, or continuous improvement, which Bandell learned when he was stationed in Japan during his military service. It is incorporating cloud and AI technology into its operations, combining the efficiency of automation and the ingenuity and expertise of humans.
Axion, whose technology infrastructure is housed on the Google Cloud Platform, has experimented with both intelligent character recognition (ICR) and optical character recognition (OCR) technology. However, the results were less than positive, as the majority of Axion’s projects have historically included handwriting and the technology did not produce positive enough results to warrant the investment.
While AI technology can automate data entry to a large degree, Bandell says that it is unable to perfectly decipher handwriting, especially in comparison to the team of experts he has in-house. In early AI technology tests, Axion has found that the software can recognize around 70% of handwriting, which means that human operators still need to do verification, especially for projects where a high degree of accuracy is needed, such as financial and personally identifiable information.
“During AI processing, the AI engine signs a confidence level score to each field. In their early stages of evaluating AI, the confidence levels achieved for handwritten text were in the 70% to 73% range. For numeric and date information, the results were in the high 90% range. On the other hand, human experts on their first pass can expect a confidence level in the range of 96% to 98%, without verification,” said Alan Bandell.
Axion currently employs a three-pass process, where a document is processed twice by different individuals, and then a computer program flags any discrepancies between the two operators’ work. These discrepancies would be reviewed by a senior operator or supervisor, ensuring that all data entered is as close to perfect as possible.
Axion incorporates AI to handle the first pass, followed by human verification of questionable data. The second pass to review questionable data will be done by a live operator. According to Bandell, while this process will be slower than a fully automated one, it is more accurate and still offers a speed advantage over the fully manual process, as each operator is validating the AI’s work rather than another operator’s.
“Applying AI technologies will both be to our clients’ benefit and ours,” Bandell says. “Typically when we are doing research for a project, we are entering data directly into our clients’ in-house systems. Furthermore, the research we do tends to require judgment and decision-making requiring human involvement,” said Alan Bandell.
Thus, Bandell believes that both automation and manual data entry will be able to coexist in the foreseeable future, as both have unique strengths and weaknesses that the other can complement.
