Generative analytics and business intelligence is in reach but requires key capabilities to be in place, says Carlos Canales, Carlos Canales, senior manager IT excellence at Sopra Steria Norway
While technology is speeding up dataflows and improving data-cleansing techniques for analytical purposes, enterprise architects should consider decentralising data architectures to unleash untapped benefits for performance management and decision making.
Exponential advancements in generative data visualisation have enabled major analytics and business intelligence (ABI) vendors to develop features that autogenerate key performance indicators and cross-visually interactive dashboards – that is machines creating insights without human intervention. This may soon reduce the need for business intelligence dashboard designers.
Knowing what to visualise requires domain knowledge and experience – but AI will soon outperform humans in such tasks too. However, when applied in ABI, generative data visualisation falls in the same trap traditional ABI does: data quality and misaligned context.
For AI to flawlessly take over the front-end of ABI, a key business capability needs to be in place: data products.
Data products is a concept drawn from one of the four principles under data-mesh architecture: data as a product. The most concise definition I have come across in the industry states that a data product is a ‘managed, reusable dataset created for the purpose of creating value’.
Why is this business capability key for generative ABI?
Although there have been standard frameworks partly addressing the data quality problem, such as Total Data Quality Management or Data Management Body of Knowledge, they have not been able to safeguard and keep the governance of data quality and metadata in such a tight and collaborative alignment with the business as a data product architecture does.
Traditionally, most organisations have designed data-quality solutions -- and related-governance in their data architecture domain – i.e., strictly within their information systems architecture. This has led to centralised custodianship of data and material gaps between business and IT.
Treating data as products, however, breaks the paradigm of centralised data quality monitoring and the use of static data structuring. Instead, when AI discovers’ and uses properly catalogued data products (in a decentralised data architecture), the generated insights get contextualised to end-users with domain viewpoints.
Therefore, with its domain-oriented approach, data products can solve data quality and context-misalignment issues from a governance and architectural perspective.
Well-governed and contextualised data: a clearer path for generative insights
The starting point of data products is that they serve analytical purposes and, as defined by data mesh’s inventor Zhamak Dehghani, have eight characteristics:
- Discoverable
- Addressable
- Understandable
- Trustworthy
- Natively accessible
- Interoperable
- Valuable on its own
- Secure
When a data product capability is created in an organisation, the physical and logical layers of analytical datasets become governed and organised by the properties listed above. As a result, large language models can not only find the nuances of the data quality in those data products, but also programmatically contextualise the insights produced based on the metadata of the datasets used in graphical representations (e.g. KPIs and dashboards).
With natural language inputs, users are then able to interact with data that is more closely- and regularly validated by domain experts.
Rapid adoption of data products in the ABI industry
Unlike traditional datasets used for analytical purposes, which normally lack cross-functional governance, data products are backed by a logical architecture embedded in both the business domain and the enterprise IT artifacts. Such cross-functional governance and architecture ensures proper usability (corroborated by the business) and consistent interoperability.
These architectural capabilities may now be possible in your existing IT stack. Major ABI vendors are rapidly deploying in their platforms a feature known as data catalogue – i.e., in the context of data-mesh, an enterprise-wide “à la carte” of data products.
As data cataloguing features in ABI platforms mature, organisations can reap the benefits of generative technology with well-governed and trusted data products – provided that such business capability is in place.
Technology has now done its part in bringing us as close as we have ever been to trusted generative insights. Now it is the hands of enterprise architects to establish decentralised data architectures in their organisations and thus unleash these technological benefits that add exponential stakeholder value.