Where are we at and where are we going? Cloudera leaders on integrating AI into successful government transformation

Where is generative AI already being utilised effectively in government - and where can it be harnessed to improve future government delivery? 

We spoke to Dr Chris Royles, Field Chief Technology Officer (EMEA) and James Underhill, Senior Account Director - Government at Cloudera about what Cloudera provides to government and what the future of AI holds for public bodies.

Cloudera provides a platform for organisations to manage and interpret data of all types, on any public or private cloud. It integrates multiple platforms (data meshing), standardises the data for different use cases (data fabric) and allows data to be moved seamlessly from different locations. Its unique Data Lakehouse leverages the data using analytics tools to provide trusted insights and power applications and systems.  

What Generative AI systems does Cloudera provide and which systems are optimised for public service usage?

“They all want AI, some of them don’t necessarily know what they want to do with it, but they know that they have to, or they should,” says Underhill of government departments. The key is to focus on solutions, outcomes and use cases which will actually deliver better services, before deploying the technology. 

Therefore Royles highlights that “getting data AI-ready is the first step in the journey” which Cloudera takes its customers on. In government, data silos present a significant challenge to sharing and delivery so the key is to create a “fabric that can federate across those data silos.” 

Ensuring the data is trustworthy by overcoming fragmentation is key to building effective AI models and tools for the future. However, it is important not to lose “that rich vein of context specific information” in the process, says Underhill, as this will inform the most responsive services.   

Underhill emphasises the particular importance of data privacy and security within the public sector to upholding citizen trust. With departments sitting on legacy systems and data spread across many domains, Underhill highlights how Cloudera can help build a “central view of who I am as a citizen” which holds across all government interactions. 

Cloudera’s recent acquisition of Octopai B.I. Ltd., now Cloudera Octopai Data Lineage, will help organisations with big picture data management and governance to help facilitate linked up services. 

What are the benefits of Cloudera’s joined up data lakehouse system? 

Cloudera’s unique offering is its data lakehouse, a joined up system which standardises data. It is essentially “one system for doing different types of workload,” says Royles, with use cases including business analytics, building models and integrating a model with enterprise. Its design improves ease of use and is lower cost for the consumer.    

This system is particularly valuable for building and scaling AI models as it facilitates model reproducibility and regulation to verify exactly what data was used to train the model. Royles explains that the user “can effectively version control their data” allowing them to audit and evaluate data from this to create future models. 

This iterative approach is something Underhill highlights. “Knowing that this is a process, it’s a transformation” is key within government, he says. It is important to recognise that models will need time to adapt and improve, then they can deliver excellent outcomes. 

What kinds of things are Cloudera doing to support successful AI adoption in government?

Underhill explains that Cloudera firstly recognises that departments possess varied data infrastructure and therefore there is different value to be gained from AI adoption. The principle is connecting data from different silos to create value and high insight quality, recognising that change is gradual and “there are different stages on that GenAI pathway,” he says.

Typical cases of government AI usage include looking at how we “check for error, check for fraud, check for consistency, match, merge, de-duplicate,” say Royles. Where AI is used in fraud and error detection, it is already saving billions of pounds of public money.  

AI has also been adopted within applications to provide data stories to citizen-users, helping organisations share data more effectively. Additionally, they have facilitated the use of AI to verify information provided in government questionnaires and check it against previous records for reliability.  

What advice do you have to public servants for integrating AI into their systems? Both in terms of their own workflow and public delivery

Royles says it is key to understand how data is used as part of a wider process across government. He emphasises the value of creating those data maps previously mentioned from the backend to the interface with the citizen. 

This “helps you decide what you have today, what you might need to change as the next step and also what processes you might need to put in place,” he says. Cloudera Octopai Data Lineage  will help to deliver this service.   

In addition to ensuring data is AI ready, Royles says the priority should be getting processes AI-ready as well. Focusing on “lower use, higher value use cases” which help public servants learn more about future use cases is important.  

Despite encouraging AI adoption, Royles highlights that “there’s always going to be people in the process.” The key is to unlock the productivity of the people, the government officials, by reducing human error and allowing the processing of vast data sets. As Underhill explains, embracing this cultural change is needed to create effective service transformation. 

Current uses of AI to streamline customer service operations so civil servants can focus on complex cases, for lesson planning to enable teachers to focus more energy on pupil interactions and for red lining financial contracts to free up lawyers to focus on higher level tasks are some examples they highlight of successful AI integration.  

As Royles observes, “that time wasn’t saved, it was moved to solving harder problems” where the expertise of civil servants can be best put to use. 

Where does the future of AI use in government lie?  

Underhill highlights how he thinks creating customer 360 is key. In future citizens should interact with government services and find they already possess relevant information on them, regardless of the touchpoint. This joined up service already exists with the Tell Us Once initiative, for example.

Similarly, Royles speculates that AI could be used to help improve citizen’s journey through government by recognising people not only as individuals, but also as members of a household.    

In terms of policy, Royles observes there is a new appetite for data sovereignty and creating a sovereign data footprint by fencing in an organisation or country’s data. He suspects this trend will continue in the future leading to private AI with unique systems controlled by one organisation protecting data, IP and leading to full control of what AI is trained on.  

In terms of technology, Royles is excited by the potential to overcome the weaknesses of Large Language Models (LLMs). LLMs are restricted to delivering a certain set of tasks due to being trained at a specific time with limited access to the outside world preventing them from adapting to, for example, legislative changes. 

Two new architectures are being built around the models to address this issue: 

1. Retrieval-augmented Generation (RAG): letting the model interact with other systems like databases to enhance context and accuracy 
   
   
2. Agent Networks: providing an intelligent team of AI agents with different specialisms, like searching the web or mathematical calculations, who can collaborate to solve tasks with the most up to date information. For example, these agents could integrate citizen records with most up to date legislation. Cloudera's Agent Studio is facilitating the development of these teams    

Overall, Cloudera presents an optimistic picture of how AI can be successfully adopted by government. Through adapting public sector culture, focusing on use cases and taking an iterative approach, civil servants can embrace the opportunities AI provides.