New study: Government GenAI optimism may be outpacing ability to deploy

Despite high hopes and dedicated budgets for GenAI, government departments need to increase guardrails and training, while synthetic data could be a missed opportunity.

A new global study, Your Journey to a GenAI Future: A Strategic Path to Success for Government from SAS and Coleman Parkes Research, reveals that government departments lag well behind other sectors in the adoption of generative AI. However, 60% of government respondents believe GenAI will drive innovation, and those that have begun using it are already seeing improvements in employee satisfaction, compliance, and operational costs and time savings.

Despite trailing other sectors by 10% (44% vs. 54%) in the current use of generative AI, the success of these early adopter departments suggests enormous potential for the technology. Those benefits could arrive soon, with 84% of government decision makers saying their organisations are planning to invest in GenAI in the next financial year, and 91% of those respondents already having a dedicated GenAI budget.

All sectors surveyed, including the banking, healthcare and life sciences industries, shared top concerns about adoption of GenAI which included data privacy, data security and AI governance. However, government respondents had larger concerns (52%) about cultural resistance to change compared to other concerns (46%) and believe compatibility with legacy systems could be a challenge.

Additionally, the promise of GenAI in government may be imperilled by inadequate regulatory preparedness and lack of understanding of GenAI, relative to other industries. While many organisations have rushed to put GenAI guidance in place, only 52% of government organisations have a policy stating how employees are and are not allowed to use GenAI at work, compared to 61% across all sectors.

The study found that government departments set aside less of their budgets for governance and monitoring than other sectors; 64% have allotted one-tenth or less of their GenAI budgets to governance and monitoring. Additionally, 50% of public sector respondents said they either don’t have a framework or that it’s ad hoc or informal, in comparison to 39% across the board.

GenAI regulation is moving quickly, so keeping up with it while unlocking the technology’s value is a universal challenge. However, government may be less prepared than other sectors, as 51% of government leaders say they’re fully or moderately prepared to comply with current and upcoming GenAI regulations, compared to an average of 58% across all sectors.

Awareness is also a concern, as only 35% of public sector employees are familiar with their organisations’ adoption of GenAI, far less than the 46% average. These lagging indicators could be the result of a problem at the leadership level, as only 38% of senior government decision makers say they understand GenAI and its impacts on business processes well or completely, compared to 48% across all sectors.

Commenting on the research, Nicky Furlong, Director, Public Sector at SAS UK, said:

"While government departments in the UK and globally may not have been the first to adopt GenAI, they are now well-positioned to enhance productivity and transform public services with this technology. Early adopters are already proving that the more it's used the more confidence will grow, leading to accelerated innovation. 

“With the right processes and policies in place, tools like large language models, digital twins, and synthetic data hold immense potential for transforming government services for employees and citizens."

Low interest in synthetic data, artificial data that accurately mimics real data, could be inhibiting innovation in government. For example, synthetic data of simulated traffic flows could help transportation departments test a road improvement with what-if scenarios even if they only have a few months of traffic data. Since it can mimic sensitive data, it can be created to train and test a system that processes health records, student records or tax information.

However, the study found that 32% of government decision makers would not consider using synthetic data. This exceeds the mere 23% of respondents across industries who are averse to its use.

“Synthetic data is particularly relevant for government departments that must follow strict data privacy regulations,” said Furlong. “Governments can use synthetic data for various purposes, including research, testing and analysis, while mitigating risks of violating privacy regulations or exposing sensitive information.”