Event review: FutureScot Cloud, Data & AI - Transforming Public Services

Digital transformation

This conference addressed some big topics: service delivery transformation, public sector data management, migration to cloud platform services, and use of AI and automation to support service delivery. 

The keynote speakers laid out the challenges to be addressed over the day: data-driven decision making is enabled through appropriate readiness levels in technology, policy, and society; data management and information generation is complex and context dependent; balancing the risk and benefits of data sharing and distribution is always difficult, and is particularly hard for public sector agencies; and building internal skills in Cloud and AI is a prerequisite for successful service transformation.

Change programmes present many organisational challenges, and one way of accelerating change and lowering risk when building new capability is through partnerships and collaborations. This was the one of the main themes for EPCC’s Director, Mark Parsons, who was one of the keynote speakers and experts on the opening panel session "Maximising the value of data and using cutting-edge data science techniques". EPCC has a long history of helping partners develop and adopt new technology through partnerships, with EPCC leading with expertise in compute density and data management scale that organisations may not have previously encountered. 

Cloud

The Cloud track session started with the Cloud migration journey of a major UK telecoms company. 

Establishing a Cloud technology centre of excellence to co-ordinate and promote best practice and to drive the IT change programme across the organisation was one of the key recommendations. Hybrid Cloud operations were presented, arising from the need to balance and match the organisation’s required technology service portfolio with the products and solutions available on the major public Cloud service providers. Whilst it is possible to put all IT functions on to one Cloud service platform, some existing functions may be more easily migrated and optimised on one platform than on another.

The follow-on session focused on the need to build a solid Cloud skills base internally in an organisation in order to maximise the effectiveness of staff and to build confidence even when most of the migration process may be outsourced to a third party. 

The training syllabus of one major Cloud service provider was presented with a detailed discussion of the specialist tracks available and how these might align with existing organisation competencies and functions. Cloud skills are particularly important post-migration when groups associated with DevOps, DevSecOps, and FinOps become engaged with optimising the form and maximising the value of the service delivery processes. 

AI

The conference culminated by bringing all the themes together in the session for AI in support of human decision making and enhancement of service delivery. 

This session was led by Public Health Scotland (PHS) project team members responsible for delivering performance improvements and forecasting demand for care services. The two project case studies presented were important for very different reasons. The first case study was a decision support system underpinned by a Bayesian statistical model rather than an AI model. This case study reminds us that it is not always appropriate to use AI models when alternative, established, and less complex methodologies can provide an adequate solution. 

The second case study was an example of PHS partnering with the Alan Turing Institute (ATI) and EPCC to use the National Safe Haven to develop a new foundational AI model. The development and adoption of new foundational models for any purpose is complex and expensive. It also requires AI skill sets that are more particular to research and data science institutes rather than public service agencies. When new models must be developed, or existing foundational models substantially optimised for specific data sets and use contexts, partnering with a specialist is one way of managing the research risk.

Perhaps of most importance to EPCC is that the service development model presented was one in which the risky and costly development work was performed in the NSH environment in partnership with ATI and EPCC. Once the foundational model had been built and tested it was deployed in a production environment, on a Cloud platform, where it was evaluated and improved with live data, before being rolled out as a live service.