From hype to real-world hero: experts on the deployment of ‘pragmatic AI’ in government
Much of the noise around AI focuses on anticipated future uses and disruption. But what about its practical application in the here and now? In a Global Government Forum webinar, public and private sector experts shared how AI is being used to streamline existing processes in government – with an eye on preparing for what comes next
Developments in artificial intelligence will have a revolutionary effect on the work of governments. It offers a whole host of benefits, from time and cost savings to the amplification of public servants’ efforts and better services for citizens. At a time when many governments are at the start of their AI journey, governments are also looking for AI applications that can make a real impact now.
During a Global Government Forum webinar, public servants from the US and Canada and an AI expert from SAS shared insights on how governments are moving beyond the hype to so-called ‘pragmatic AI’ and the practical application of tools and systems to streamline back-office processes.
AI a ‘valuable assistant’ in the US health department
Robert Owens, chief financial officer at the US Department of Health and Human Services’ (HHS) Office of the Inspector General (OIG), spoke about how his office is “at the beginning of its AI journey” – but that the vision for its use of the technology is clear and that it already plays a role in its day-to-day operations.
He explained how AI is being deployed among its specialists.
The OIG’s auditors and evaluators use AI for data analysis, risk modelling and compliance checks, and its attorneys apply AI to the process of reviewing contracts and legislation, and for summarising case law, research and other documents.
Its special agents leverage AI for investigative purposes, triage, anomaly detection and for trawling large data sets, and its managers and supervisors use AI to assist with “drafting, performance evaluation and summarising team performance metrics while retaining full accountability for decisions”.
He emphasised that “AI should serve as a valuable assistant, not a replacement for human judgement”, and that OIG employees are “fully accountable for the final product, even when AI helps accelerate the process”.
He also stressed that the verification of AI tools’ performance and accuracy are non-negotiables, having learned firsthand about the follies of over-trusting the technology.
Looking at the bigger picture, the OIG’s use of artificial intelligence aligns with the Trump administration’s AI Action Plan, he said.
He noted that the OIG returns US$12.70 for every dollar invested in it and receives “less than two cents for every US$100 spent by HHS”. This gap often forces “tough decisions”, he said, but also creates an opportunity to leverage technologies like AI smartly.
Read more: US transport department ‘to use AI to draft regulations’
Canada’s plans for enterprise-wide deployment of AI
Next to speak was Canada’s Jonathan Macdonald, director of responsible data and AI at the Office of the Chief Information Officer, Treasury Board of Canada Secretariat. He began by stressing that the public sector is one designed to change “drastically and continuously”, and that though the Canadian federal government consists of various layers, there are “remarkably similar threads” that run through each, and which AI is apt to simplify.
The federal government’s ‘GCtranslate’ tool is one example. Announced in September 2025, the tool is designed for use by departments and agencies to translate from English to French and French to English. Describing GCtranslate as the government’s “lighthouse project”, Macdonald said that though the project is in its early stages, results from the initial pilot are hopeful.
There is now an “ambitious timeline for a broad deployment”, he said, with the first post-pilot wave of deployment due to take place over the next few months. If all goes well, the government plans to commence a full rollout of GCtranslate by the end of 2027.
“We are estimated to process around or over 450 million words annually, and early estimates on savings are in the tens of millions. This is one fantastic example of a pan-government AI solution,” Macdonald said.
Innovation 2026
Innovation is a unique exhibition and conference that brings together government leaders from across the globe responsible for the transformation and acceleration of their public sector organisations and services. Held on 24 and 25 March 2026 at Excel in London, the event is co-hosted by the UK Government, UK Civil Service and the Cabinet Office, and covers innovation across a range of topics, including data, digital transformation, workforce, culture, sustainability, and much more.
Find out more about Innovation 2026 and register to attend here
The government has also published a register of AI use cases across the full breadth of its operations, which shows “400 instances” of AI-use from “over 40 different departments”.
“What we’re doing at the centre [of government] is build[ing] on the successes of enterprise-wide deployments and look[ing] to other potential use cases that have the opportunity and promise to scale up,” he explained.
The promise of agentic AI action taking
Antti Heino is principal AI advisor at SAS. He told the live webinar audience about the potential of agentic AI, as well as what he considers the “underutilised” role of synthetic data in AI model testing.
He said that though the latter is a relatively new development, synthetic data – which is broadly defined as artificially-generated information that imitates the statistical properties and patterns of real-world data – is useful for trialling AI without risking data privacy.
“We don’t have to use any real data for many types of phases in different projects, and that protects this sensitive information,” he said.
Turning to the topic of agentic AI, Heino explained that there has been an evolution from traditional machine learning for many types of predictive problems, to deep learning, to models that have made it possible to generate content.
“We got ChatGPT [and other LLMs], that were able to generate things. But then organisations understood that just generating media like text, audio and video is not enough. We need agents that can perform actions.”
When it comes to the public sector, Heino said that agentic AI is still in a relatively limited “research phase”, and recommended that the public sector approach agentic AI from a “decision automation angle”.
He elaborated: “If it’s more of an internal use case, you might be able to bring in these more sophisticated models, or larger language models, to increase the level of automation and increase the level of capability to create automated actions in your environments.”
Read more: New GGF research: Canadian public servants report digital progress but ‘persistent’ barriers remain
Though the use of agentic AI in government is relatively immature, it is being considered by the governments of Canada and the US.
Macdonald said the Canadian public service is experimenting with agentic AI and is “in the throes of developing guidance on the responsible use of the technology”, but added that he doesn’t currently see ambition for a broad deployment of agentic AI systems or AI agents.
Recalling some of the AI-related anxieties that followed ChatGPT’s debut, he said that agentic models represent “a force multiplier to that”, making it essential for governments to “meet the moment” with rules around responsible use.
Owens said that the OIG is not currently using agentic AI but that he has been “watching it as a US federal employee” and mentioned the Office of Personnel management’s proposal of a “unified HR system across the federal government”.
“[The Office of Personnel Management] doesn’t explicitly talk about agentic AI, but I am sure it’s within the goals of this procurement,” Owens said. “It has a very aggressive time scale to roll this out, and some very lofty goals to attain. So we are watching that closely, because we will have to move to that system when it comes online. It is promising great efficiencies, which will be awesome. So we’ll see how that goes.”
The deskilling effects of AI – and its offer of freedom for creativity
The conversation moved on to the impact of AI on public service jobs.
Heino said there is a danger that AI is applied in a way that does not give the people using the technology “any type of freedom to be more creative”, and emphasised that AI should only be used to automate “the parts you hate about your job” and not “critical thinking” or anything that involves initiative, benevolence or spontaneous action.
He added that “job roles should evolve” to reflect the efficiencies captured by AI.
Macdonald warned that “the risk of skill atrophy is real” with greater adoption of AI.
He gave an example: “I often use generative AI tools for recipes. I have never been so helpless in the kitchen as I am now, because if I don’t have my phone or a computer close to me, I almost forget how to fry an egg.
“Understanding the limitations of this technology is key. If you were a first drafter of things [before AI], now you’re going to be the first reviewer of things. Your job has changed, so you need to apply a different skillset to it.”
This will require leaders and managers to manage people differently too, he highlighted.
The ‘Pragmatic AI: Cutting through the noise to make artificial intelligence work in government’ webinar was held by Global Government Forum with the support of knowledge partner SAS on 12 February 2026. Watch the webinar on demand here and hear the panellists discuss a range of related topics including:
– More on how agentic AI can be used in practice
– The importance of accountability and transparency in the responsible use of AI
– How governments can set guardrails and allay civil and public servants’ fears about being replaced by AI
– Increasing AI literacy in civil and public services
– The implications of the use of AI in national security
– Mitigating the environmental impacts of AI
– Data privacy
About Jack Aldane
Jack is a British journalist, cartoonist and podcaster. He graduated from Heythrop College London in 2009 with a BA in philosophy, before living and working in China for three years as a freelance reporter. After training in financial journalism at City University from 2013 to 2014, Jack worked at Bloomberg and Thomson Reuters before moving into editing magazines on global trade and development finance. Shortly after editing opinion writing for UnHerd, he joined the independent think tank ResPublica, where he led a media campaign to change the health and safety requirements around asbestos in UK public buildings. As host and producer of The Booking Club podcast – a conversation series featuring prominent authors and commentators at their favourite restaurants – Jack continues to engage today’s most distinguished thinkers on the biggest problems pertaining to ideology and power in the 21st century. He joined Global Government Forum as its Senior Staff Writer and Community Co-ordinator in 2021.
Related Posts
