Lifting the limitations: a deep dive into enabling data-driven decision-making in government
Governments are placing greater emphasis on extracting and acting on the insights derived from data. At a roundtable, senior civil servants from a range of countries discussed their approaches to laying the groundwork for greater productivity
Data and AI hold significant promise in helping governments make the right decisions as quickly as possible. As one senior public servant put it: “We are trying hard to let our people know that without [data], decision-making is very limited. I mean, you can’t make an informed decision if you don’t have the information to make it.”
Yet around the world there is some way to go before internal processes enable evidence-based decision-making to be embedded as a standard practice.
This was the message of an online roundtable hosted by Global Government Forum with the support of knowledge partner SAS, in which 13 senior civil and public servants from 10 countries – including the UK, Estonia, India, Ghana, Mexico, and the European Commission – discussed the enablers and barriers to using data and AI for decision-making in government.
Examining issues such as the importance of using data not only to make decisions but to subsequently evaluate their efficacy, the roundtable offered nuanced views, insights and ideas.
Kicking off by asking how confident they are in the decisions being made in their organisations, there were mixed responses from participants at the event, which was held under the Chatham House Rule to allow them to speak freely.
While most were positive about decisions taken, there was recognition of the challenges with issues around data quality, standardisation and sharing across departments and policy areas from all countries represented. These issues were having a knock-on effect on efficiency, with duplication of effort “happening all over the place”.
Even in organisations that have advanced front-end digital services for citizens, “behind the scenes, the data is very segregated,” one leader admitted.
“At the policy level, I do believe that the decisions are as good as they can be and based on data… but everything in between – the managerial, everyday decisions that we have to make in a hurry – there’s a lot of improvement [needed] there, and I don’t know how to solve it.”
The roundtable built on the findings of Empowering public servants to make confident, informed decisions with data and AI – a report by Global Government Forum, supported by SAS and AWS – which is based on a survey, and insights from senior civil servants and private sector experts.
Different approaches to data sharing
It was clear from the discussion that in such big machines as governments, with their vertical structures and remits traditionally siloed to individual departments, embedding a data-sharing culture across the ‘whole’ is far from easy. But the ‘wicked problems’ facing governments worldwide require a big shift towards coordination and collaboration, with data sharing key to unlocking the insights needed for robust decisions and effective outcomes.
Participants described how their countries are working to apply ‘safe’ and consistent rules around data sharing, avoid overcollection, and utilise external organisations to sense check their datasets.
One participant described his country’s approach, including its data management system, which incorporates data, technology and interoperability standards and validations; ensures accurate categorisation of personal and non-personal data; and has mechanisms built-in to ensure data is kept up to date.
There is a “bottom-up approach to data collection” to encourage a focus on usefulness, and a ‘maker-checker-approver’ data lifecycle principle has been implemented – requiring three people to input, review and approve data before it is uploaded to a system – to support integrity.
Some countries are also connecting databases using APIs and making them available to academia and the private sector to help build AI models, enabling what one participant called “better refinement and a better decision support system”.
Similarly, one government department has entered into memorandums of understanding (MOUs) with universities to grant them access to data held by the department for analysis purposes. “This has allowed us to influence Masters and PhD-level research as well as influential reports – it gives us broader insight into our own datasets. This is my best piece of best practice,” the leader said.
More broadly, one participant described efforts to instil effective data-sharing – across government and other sectors – being driven by a central coordinating body which is responsible for overseeing the collection of data, standardising it and inputting it into the country’s national information system.
And there are other approaches too. Adding qualitative data into the mix is something one participant has been working on, as they described: “We are thinking about how to make qualitative, as well as quantitative, data available and useful to inform the development of interventions.
“We’ve had to persuade leadership to allow us to create a digital platform where actors across silos can review the data, contribute meaningfully to it, and use it how they see fit. This is less about centralising control but enabling action at multiple levels to happen, so it’s important to think about who is using the data and why.”
Delving deeper, there are other considerations around data access and sharing which may be out of governments’ direct control. A participant said one of their main challenges is that they don’t own all of their key datasets and are reliant on the private sector to share appropriately. This means that “sometimes our data exploitation potential is only as good as our commercial considerations”, they said, explaining that they are working to ensure their commercial teams engage digital teams when considering data-related aspects of contracts.
Phantom barriers, fundamentals, and skills
As well as actual barriers to data access and sharing, in one participant’s experience, public servants can be reluctant to share data for fear of falling foul of rules like the EU’s General Data Protection Regulation (GDPR), even if the action would actually be compliant.
“There’s often a perceived barrier that it’s not possible to share data, whereas actually you can do it with the right sort of permissions, and often you can build those permissions in really early,” he said.
This is where the implementation of a data governance framework can help, and as Jennifer Robinson, global strategic advisor for SAS’ public sector practice, noted: “[In terms of] being nimble and using data in light of a situation that requires fast response, I would think those of you that have data governance in place may be better equipped to be able to turn on a dime and use that data for fast decisions.”
Though, as voiced by one participant, even if an institution has a data governance framework in place, its people may not know how to apply the policy to a specific action, raising questions around training and awareness.
Some of the leaders were clear that confidence in decisions comes down to transparency and in knowing how a decision – made by a person or team or supported by AI – was made and based on what data, which also requires a focus on improving skills and capacity within civil services.
“Even if we overcome all these previously discussed barriers, there’s one that remains, and maybe even increases in a data-rich world: understanding data. Not data literacy in numbers, but understanding why and how different datasets were collected,” the participant said.
They shared an example that illustrates this neatly: different official databases – such as population, medical, and genomic registries – may each hold distinct information about the same individual. “All three might be different and yet not be incorrect data. So if you access all three databases and try to teach your AI, you really need to understand which database contains which data and why,” they explained. “Data literacy, in the deepest sense, remains the biggest challenge.”
The countries represented at the roundtable are working to bridge data and AI skills gaps in different ways.
Some are making it mandatory for all civil servants to be trained in new technologies, including AI.
Others offer technical internships to students in education so that when they join the public service, they are “able to provide the support we need” – an intervention that came about after a disconnect was found between the theory students learn and what the public service needs from them in practice.
Some have also introduced training programmes for senior civil servants that focus on making data, AI, analytical and technical skills a core capability.
This fed into a conversation about how to drive change from the top, with some participants noting that leaders aren’t always setting an example or creating a culture where people feel empowered to use data and AI to aid decision making – and about the importance of bottom-up exploration.
“I think it’s impossible to drive this from the top,” one participant said. “You need the support and understanding that is only driven, I feel, from the bottom up – by going into the institutions, creating the communities, and really driving this through the people who are actually on the floor by allowing them to experiment and to really try things out.”
Help to harness AI
Building capacity and improving processes so that civil and public servants have access to the data they need, when they need it, and in the format they need it in, isn’t solely about data use – it’s also a crucial enabler to harnessing AI to make autonomous or support human-made decisions, streamline workflows, and ultimately enhance efficiency.
There was agreement among participants that the problems governments are experiencing with siloed and poor-quality data was stymieing use of artificial intelligence and related productivity gains.
One suggested that “the better we get at going in the same direction and doing some kind of federated [data] system… then we can really start to utilise AI to its full extent”.
With greater adoption of AI comes greater access to evidence – but using this appropriately would require new sets of skills, one participant emphasised.
He said policymakers are now dealing with lots of content – from within government or via stakeholders – that is generated by AI, and that such content should be viewed “with the same healthy scepticism that any other engagement with evidence should be”.
“Understanding how to interpret new forms of data, often synthesised, is a part of good practice. I am hopeful the public sector will be able to adapt to this,” he added.
Evaluating the efficacy of decisions
Improving public servants’ capability to make confident decisions using data and AI is one thing but as two of the leaders voiced, reflecting on decisions made and evaluating their impact is also important but often overlooked.
“I’m confident at the point of the decision, but then there’s the accuracy, consistency, timeliness and provenance of the decision that we don’t necessarily always go back to and reflect on,” one said.
The other cited a statistic that only around 9% of public policy programmes in their country are properly evaluated.
“That goes to show that the focus on evaluation can be limited, and I think often where the long-term evidence base can be most useful is in those wicked problems where you’ve got cross-cutting issues [that touch multiple] departments – that’s where data sharing becomes really, really important.”
They acknowledged that “it’s very hard to measure some of the most valuable aspects. ‘Hit the target, miss the point’ comes to mind”.
Evaluating whether interventions are working can be key to keeping funded government projects going and make senior leaders vividly aware of the importance of data in their decision-making, as another leader described: “A lot of our work is dependent on how we are able to anticipate risk and design policy, dialogue, support and training for stakeholders within the region. Data has been critical. So what we’ve done over the years, especially because a lot of our programmes are funded, is to provide critical data to leadership for them to understand that without a push for self-sustainability, the centre might not be able to support itself.
“I think that push and acknowledgement of the dwindling funds was a wake-up call for our leaders to take data seriously. Now they’re actually pushing for data at every executive meeting so they can see how well we’re doing, where the anticipated risks are coming from, and whether the interventions are working.”
The roundtable demonstrated that even the most advanced governments have a way to go before data and AI is routinely used in decision-making. But as the need for better cross-government collaboration – and collaboration with external partners – intensifies, countries can learn a lot from each other about how to set the stage for effective data use and the productivity gains that follow.
The ‘Improving productivity in civil and public services through data and AI-driven decision-making’ roundtable took place on 24 September 2025. It was hosted by Global Government Forum with support from knowledge partner SAS.
Read the Empowering public servants to make confident, informed decisions with data and AI report.
About Partner Content
This content is brought to you by a Global Government Forum, Knowledge Partner.
Related Posts
