Unleashing productivity in government

A new era of productivity growth

After some stops and starts during and after the dotcom era, governments seem to have found their digital footing.⁸ Leaders worldwide have digitally transformed operations and service delivery through the mainstream adoption of cloud computing, improved commercial off-the-shelf software, and a surge in open-source development.⁹ Software as a service has enabled agencies to access first-rate programs without needing to build expensive custom systems from scratch. Agile development has helped leaders break up large technology projects into smaller, manageable modules, which has helped cut costs and increased efficiency and productivity.¹⁰

With the advent of artificial intelligence, governments have reached another jumping-off point in their journey toward productivity and efficiency. Digital transformation has enabled agencies to set up the foundational digital infrastructure required to benefit from advances in AI technologies. The power of AI technologies has been demonstrated in detecting fraud, reducing costs, optimizing resources, improving customer experience, and streamlining back-end processes.¹¹ Trelleborg, a Swedish town, led the way in 2015 by digitizing its social benefits program, using robotic process automation to slash processing times dramatically.¹² Many other European nations and offices have followed suit, tapping into expanding AI-based capabilities.¹³

The US Internal Revenue Service (IRS) provides a peek into how digital technologies, AI, and process transformation can help improve government productivity. The COVID-19 pandemic strained IRS operations, which was already struggling with a lack of resources and outdated technology.¹⁴ Staff shortages added to the burden of managing tens of millions of stimulus payments and reconciling these payments during the tax season. As a result, customer service suffered. Only 9% of phone calls got answered, the unprocessed paper tax returns backlog surged to 21 million, and the agency kept taxpayers waiting for an average of 251 days for responses to proposed return adjustments.¹⁵

The IRS turned this around by using technology upgrades, achieving process improvements, and hiring new service representatives. By automating the paper returns–scanning process, the agency could scan 80 times more returns in the first quarter of 2023 than in the whole of 2022. Nine of the most common notices, such as earned income and health insurance tax credits, were moved online. Previously, taxpayers had to respond to these notices via email. The agency hired an additional 5,000 new customer service representatives, which slashed phone wait times from 27 minutes in 2022 to just four minutes by April 2023. The agency also opened 335 new taxpayer assistance centers nationwide, quadrupling the number of taxpayers served in person.¹⁶

An increase in staffing certainly drove up at least some of the agency’s 10x productivity improvements. However, the core change that should persist beyond the hiring surge was the implementation of AI technology. These outcomes align with Deloitte’s 2016 analysis that suggested AI’s potential to dramatically reduce one of regular government employees’ biggest paperwork and operational activities: documenting and recording information.¹⁷

Generative AI, the technology’s latest iteration, has the potential to automate and accelerate countless repetitive tasks such as retrieving relevant information, communicating with others, processing information, analyzing data, and offering informed advice.¹⁸ With this technology, government employees can have decades of policy knowledge at their fingertips, policymakers can compare policies across time and jurisdictions, and case workers can input and retrieve client information remotely.¹⁹

This trend focuses on governments navigating the transition between digital transformation and the AI leap. Advances in AI provide an opportunity to kickstart a new era of productivity acceleration in the public sector. Just like digital transformation ushered in improved efficiencies, AI promises something similar, if not much more. The next era of deep productivity improvement in government has just begun.

Breaking trade-offs

The public sector’s mission of focusing on equity and accessibility can complicate the pursuit of efficiency, cost savings, and productivity. Postal services need to deliver packages to all addresses, not only to the nearest shipping centers. Public schools must educate all students, not only the highest-achieving. Similarly, agencies must provide high-quality services to all population cohorts, not just digital natives.

Unlike private sector companies, government agencies do not have the luxury of quickly redirecting budgets and staff from underperforming departments to promising new initiatives. Service mandates cannot be ignored, even if they’re expensive, and traditional cost-value trade-offs often do not apply. Therefore, achieving 10x productivity improvement lies not in cost-cutting but in foundational digital infrastructure coupled with the power of AI.

Convergence: The key to 10x change in productivity

It’s happening now. Agencies are using a mix of tools to drive 10x improvements in efficiency and productivity. They are strengthening digital infrastructure by leveraging cloud computing, developing enterprise data lakes, and enabling analytics at scale. They are breaking down jurisdictional silos and barriers to improve data-sharing and integration across government levels. They are embedding AI capabilities into existing government systems and processes. They are driving public workforce skill development to enable employees to leverage new technologies such as generative AI. Above all, they are prioritizing human-centered design to apply technology solutions to real-world problems.

The sweet spot is at the intersection of these technology, policy, and process tools—a synchronous dance between the different elements to generate value that is greater than the sum of its parts. Consider how combining different tools can have tangible impacts on constituents and operations:

• Mobility data + Internet of Things data + analytics = Less congestion
• Robotic process automation + eligibility data + human judgment = Delightful customer experience
• Historical data + machine learning + digital twins = Predictive planning
• Cloud computing + data-sharing + process transformation = Life event–based services

Trend in action

Productivity is about achieving more with the same inputs. As technology has increased productivity, it has also raised our expectations of what government can and should provide. As a result, agencies are under constant pressure to find more ways to do more.

Government agencies have various processes and systems that can be made more efficient. The challenge is to identify and implement the right technological solutions, transform business processes, and update talent policies to fully realize their potential.

Transform business processes for the digital era

Governments have made great strides in transforming and improving digital services, but more needs to be done to reduce duplication, streamline processes, and accelerate digital transformation.

Consider life event–based service delivery. Individuals experiencing a major life event, such as a birth or death, have traditionally had to engage with different government agencies. However, these agencies can collaborate proactively to make the process easier for citizens. This effort requires a fundamental transformation of traditional processes, improved data-sharing, and digital technologies to anticipate citizen needs and guide them through their likely next steps.²⁰

Austria is a pioneer in delivering life event–based family services. In 2014, the federal government launched the Antragslose Familienbeihilfe program (child benefit without application) to automatically trigger child benefits for new parents. Before the program, parents needed to submit an application via mail, online forms, or an office visit. A caseworker would further process the application by manually collecting data from different systems and assessing the application for eligibility. This process could take weeks, even if the application had no queries or errors.²¹

The program automated the data transfer process within federal, state, and local agencies, eliminating the need for a parental application. A birth record triggers data transfers from the hospital to the central civil registry, the Ministry of Finance, and local tax offices, which disburse allowances.²² Within months of implementation, the Antragslose Familienbeihilfe program created massive value and efficiency improvements, with average processing time decreased to just two days, collectively saving 39,000 hours for citizens. Time savings internally in government was equal to 15 full-time employees.²³

Similar life event–based approaches have blossomed worldwide, including in Singapore, the United Kingdom, Finland, India, Estonia, and the United States.²⁴ However, these approaches wouldn’t work without advances in data integration and governance practices. Governments are moving beyond the days of unstructured, siloed, inaccessible data. Digital systems are increasingly freeing data from its traditional jurisdictional confines, making it available not only across government systems but also in machine-readable formats that can be integrated and ingested in large data analysis platforms.

In recent years, integrated data has had a profound impact on city operations. Real-time data from traffic systems, ongoing constructions, weather sensors, and other sources has been integrated to improve city management.²⁵ Cascais, a mid-sized city in Portugal, developed a cloud-based digital command center to manage city operations effectively. The city integrated data from various domains such as mobility, public infrastructure management, civic protection, emergency management, waste management, and Internet of Things devices embedded across the city. Its smart waste management system optimized waste collection routes and times to reduce operational costs by 40% and energy costs by 20%.²⁶

AI as a productivity accelerant

In recent years, agencies have embedded AI technologies into thousands of government systems and processes, often resulting in massive improvements in efficiency and productivity:

• In 2017, public hospitals in Queensland, Australia, introduced a tool to predict patient admissions, injury type, impact on bed availability, and staff vacations, helping forecast availability hours, weeks, or even years in advance. As a result, Queensland hospitals saved US$2.5 million a year, with improved patient outcomes benefiting the state as much as US$80 million annually.²⁷
• The US Food and Drug Administration’s Center for Drug Evaluation and Research used robotic process automation in its application intake process to slash application processing time by 93%, eliminating 5,200 hours of manual labor.²⁸
• The Police Service of Northern Ireland has implemented more than two dozen AI-based automation solutions across several operational areas, including intelligence and public contact. The solutions have helped the service cut turnaround time from days to hours by fully automating low-complexity tasks while flagging complex tasks for staff review.²⁹

The emergence of generative AI capabilities promises to accelerate productivity further. At a broader economic level, it is anticipated that large language models (LLMs) could affect at least 10% of tasks for four-fifths of American workers.³⁰ Globally, some experts predict that generative AI could boost the global gross domestic product by up to 7% and productivity growth by 1.5% over the course of the next decade.³¹

Generative AI–based tools have the potential to revolutionize how the public sector operates. By automating tedious, time-consuming, as well as many knowledge-based tasks, health care workers, child services caseworkers, defense and security analysts, embassy personnel, and other government employees will have more time to focus on high-value, intuitive, and creative work, including more time to interact with citizens.

However, a challenge for government technology leaders will be developing, testing, implementing, and quickly scaling potential generative AI applications. There is already a wide range of potential generative AI use cases that government leaders can prioritize. Government leaders are expected to initially prioritize applications that improve operational and planning areas (figure 2).

A southern state in the United States is developing one of the earliest generative AI application pilots at their state’s managed care Medicaid program.³² The Medicaid program is building upon its foundational digital infrastructure to explore generative AI applications. One such use case is a policy bot to help dozens of long-term services and supports staff to seek nuanced answers by sifting through hundreds of extensive guidelines and policy documents.³³

The application aims to allow staff to quickly access these documents—including rules, waivers, and guidelines. For instance, if a staffer asks about a level of care criteria, the bot will, in seconds, scan the whole set of provided documents to provide contextual answers in plain English. Ideally, new employees would no longer need to spend extensive periods of time memorizing or researching these guidelines to find answers to such questions. The policy bot also helps capture some of the tacit knowledge in the organization that could be lost when a senior professional retires.³⁴

The state Medicaid program leaders are already thinking about how to extend these capabilities to other areas. A digital assistant for workers could scan a client’s history and eligibility to help summarize outcomes for more timely analysis. A natural language processing tool, coupled with an LLM, could help contact center workers with foreign language translations and summarization for more contextual responses. And a generative AIOps tool could help information technology system administrators scan through thousands of system performance alerts to proactively identify and escalate problems.³⁵

Getting ready for a new era of human-machine collaboration

Although technology can drive significant change, people continue to be the most valuable asset for most agencies. However, this creates two potential human capital problems: ineffective human-machine collaboration and the ongoing AI talent shortage in government.

The past decade’s explosion of AI-powered tools has empowered public sector workers by automating routine work, allowing civil servants to focus on more complex, nuanced, and higher-order tasks. But with advances in AI technologies and the emergence of generative AI, these human-machine collaborations are set to become more intimate and valuable. There are many types of daily interactions workers can have with AI (figure 3), ranging from directing AI applications to perform work (machines as subordinates) to working with AI in open-ended, iterative, and interactive ways over time in true partnership (machines as teammates) to allowing AI applications to help guide and direct work (machines as supervisors).³⁶

Singapore was one of the first governments to launch a generative AI chatbot focused on serving 4,000 civil servants, with ambitions to extend it to the entire 150,000-strong public sector workforce. Singapore’s Government Technology Agency developed the Pair chatbot to assist civil servants in writing, research, and coding.³⁷

To better serve citizens, Singapore plans to replace its current “Ask Jamie” chatbot with a chat assistant informed by LLM engines.³⁸ The government has also updated its Data & AI Literacy ePrimer learning program for public officials to include content on generative AI, LLMs, prompt engineering, and limitations of LLMs.³⁹

The increasing use of AI technologies in government will require a corresponding increase in tech talent. However, this presents a particular challenge, as many companies are also vying for the same skilled labor pool. Bridging the gap through training programs can be a solution. While agencies will need AI specialists, adopting AI at scale will require improving the data literacy skills of the workers who will be responsible for purchasing AI tools and services or utilizing AI applications to deliver services to citizens.⁴⁰

The US government’s recent executive order on AI⁴¹ includes national AI talent recruiting in the federal government.⁴² To boost the supply of digital talent, the US government has previously relied heavily on “tour of duty” programs,⁴³ including 18F at the General Services Administration, the US Digital Services, and the Presidential Innovation Fellows program. Such programs can close governments’ talent gaps by creating opportunities for highly skilled professionals to contribute to public service.⁴⁴

What the 10x future holds

• Learning time for new workers plummets, as they can retrieve policy and operational knowledge through generative AI, quickly checking historical data to understand what worked or did not in the past.
• Information-empowered citizens are able to check the performance of government programs, compare with other programs across jurisdictions, and provide more informed feedback to policymakers.
• Field workers are able to exercise greater informed discretion, as AI technologies enable quick access to information needed to make judgments; supervisors can review and evaluate these discretionary actions more holistically.
• Next-generation contact center experience with generative AI capabilities helps provide real-time AI translation, contextual understanding through shared data, and summaries of user options.

Steps governments can take now

• Develop strong AI leadership in government, similar to the digital and data leadership governments developed in the past two decades. This will be key to developing AI standards, a shared governance model, and making a stronger case for using AI in government.
• Redefine AI strategy for the generative AI era. Government AI and technology leaders should experiment and identify the most impactful use cases for the evolving tools and adjust workforce strategy to incorporate new skills into specific job roles and tasks. Agencies can develop policies and guidelines for how people should—and shouldn’t—use the technology for specific applications and build trust by fostering a collaborative culture.
• Focus on workforce experience, as technology transformation, especially at this scale, can be challenging and could generate a culture backlash. Agencies should take a measured approach to skills training with programs designed around learners and smooth the transition by involving managers and frontline workers in designing or rolling out new tools. Also, they should analyze the positive impact of higher productivity on public sector wages and talent attraction and retention.
• Create ethical guardrails for new technologies to address privacy, security, hallucinations, ethics, and fair decision-making challenges.
• Take a lead on generative AI development by going beyond the role of user of the technology. Governments as buyers, regulators, and infrastructure providers can help catalyze the innovation ecosystem and accelerate AI entrepreneurship and economic growth.⁴⁶