The public sector has long struggled with paper backlogs and worker shortages, impacting citizen response times and public perception. In March of this year, the IRS recognized the need to hire 10,000 workers to support nearly 20 million unprocessed returns as they navigated 20 times the usual amount of tax filing backlogs. These backlogs tend to ignite interest in technology innovation projects emerging across the public sector as a solution to help alleviate matters like passport or work permit delays.
As modernization initiatives become prioritized, the trailing and historically slow-to-change industry often gets stuck on identifying the best and most effective path forward. But thanks to recent AI and Machine Learning breakthroughs, new solutions like Natural Language Processing (NLP) have gained momentum, transforming back-office processes and making it easier for machines and humans to communicate with greater speed, efficiency, and accuracy. But before we add fuel to the “robots are taking over the world” fire, let’s explore what NLP is, what it is not, and how it’s helping humankind in so many valuable ways.
The Myth: “Sentient AI”
Let’s be clear, despite the recent flurry of news stories surrounding the AI industry after a Google engineer claimed that the company’s language model had a soul, AI systems are far from sentient beings. While many machine learning systems exhibit humanoid characteristics, most experts agree that living machines are more science-fiction than fact.
The advancements in NLP are rooted in deep learning systems, which are designed to build and train neural networks that are inspired by the human brain. The abilities of Google’s Language Model for Dialogue Applications, or LaMDA, showcase how NLP and AI can have a real impact. Although these technologies are far from replicating the human touch, the models are learning and improving with the support of human knowledge workers and real-world data samples—which can be game-changing when considered through the lens of government and public sector applications.
The Reality: Intelligent Document Processing
As a society, we’ve grown accustomed to using conversational IVR (Interactive Voice Response) and online chatbots when communicating with business and public sectors. But as NLP advances, the applications to assist knowledge workers will be endless.
Different operations utilize deep learning systems and NLP to automate processes, including one that’s gained popularity over the last few years—Intelligent Document Processing (IDP). Leveraging NLP to process human-readable documents, the deep learning capabilities of IDP can handle more complex processes and content types than its legacy counterparts and extract information more seamlessly for downstream decisioning systems. For example, machine learning could classify, and index photos included in passport applications, and deep learning-powered applications could consist of caption generation for images.
Imagine, amid the search for thousands of knowledge workers to support the IRS, if an NLP-backed IDP could process and approve tax returns in a matter of seconds instead of weeks. And what if it could communicate directly with an individual in the case of an error? This technological advancement would provide more efficient service to the citizen in question and allow IRS agents to focus on larger cases and pressing tasks rather than getting bogged down with manual approvals or minor discrepancies (e.g., a missing signature) requiring citizen clarification.
Beyond NLP-powered data extraction and processing, other applications focused on the downstream stages of business processes, such as machine learning-assisted decision-making, could soon become a reality, providing insights in a readable format for human workers.
Transforming the Public Sector with IDP
No vertical is more primed for technology innovation than the public sector. Processing millions of forms, applications, documents, and more every day to meet the needs of their (growing) constituents, the public sector faces tighter turn time expectations, limited resources, and higher demand. Unfortunately, these arduous hurdles make it challenging to meet those expectations. When ranking customer satisfaction by industry, McKinsey found the public sector averaged 5.5 to 6 points out of 10. In contrast, other sectors such as e-commerce, credit-card providers, and pharmacies scored an average of 8 out of 10.
Automating the slowest, most data-heavy processes is crucial to managing growing document backlogs, decreasing response time, improving data quality, making data access faster, and freeing employees to focus on more critical consumer-facing interactions. But these challenges will only continue to mount, especially if agencies continue patching together legacy systems to index and transcribe documents, despite additional investments.
Natural Language Processing (NLP) and more advanced AI can impact the public sector, hastening response times, improving accuracy, providing a worthwhile customer experience, and allowing employees to focus on customer-focused activities.
The most difficult part of any transformation is taking the first step, but once it’s taken, the places we can go from there can be limitless.
The author, CF Su, is the VP of Machine Learning, Hyperscience
