While issues around data silos, integration, quality and governance may be hampering artificial intelligence (AI) adoption in organisations around the world, AI itself could be deployed to support data integration and quality, and enable the interfacing of data governance, thus preparing the foundation for its own adoption.
AI has matured and found its place in all aspects of the data and process world.
As part of the AI arsenal, machine learning (ML) has enabled the speed of decision-making to be increased within processes (and hence output results) using the patterns, facts, findings or state of past experiences; ie, previously generated data and business or other parameters (weights and indicators) based on affinity and collusion. This has been and is traditionally done manually and subjectively.
Key AI drivers include the automation of both simple repetitive tasks and complex processes behind robotics application, cost reduction in the processing of data into information and knowledge (insights), hence less resources (people), and faster time to market, to present and to react.
Technically, the integration of AI and ML into all business and technical processes can uplift/enhance/leverage the latter investment. AI also uses its findings (fact base) to intelligently enable quick generation of insights with little or no manual intervention − and hence support for just-in-time/before-time decision-making, and tactical and strategic initiatives.
Data hurdles in the way of adoption
AI’s potential has made it a priority across industry sectors around the world. IBM’s latest Global AI Adoption Index finds that almost a third of IT professionals surveyed are using AI, and 43% of businesses said they had accelerated their AI rollouts due to the COVID-19 pandemic.
According to the report, the top drivers for AI adoption included advances that made AI more accessible and changing business needs due to the pandemic.
AI’s potential has made it a priority across industry sectors around the world.
However, the report found the top barriers to AI adoption were limited AI expertise, increasing data complexity, management of data silos and a lack of tools or platforms for developing AI models.
The report also found that over 90% of businesses using AI today say AI’s ability to explain how it arrived at a decision is critical.
AI has people-independent capabilities that can be trusted once tested; therefore it can be applied to data governance processes (data and process quality, security, usage, monitoring and oversight).
This means it has capabilities to address the very concerns that hamper adoption: data quality, data silos and the ability to explain how it arrived at a decision.
Much like the cream in a layer cake, AI can orchestrate the full data processing lineage, from data acquisition, through integration and quality, to dissemination, having (linking in) governance oversight rules over the entire process – this holds together the components of technical data management tools (“inner layer”) and governance tools (“outer/top layer”) to deliver an elegant whole.
AI can kick off the data sourcing and integration processes, do data quality checks, then take the full process and link it to a governance oversight business process or control/approval interface. This explains how the AI arrived at a decision using governance tools to surface or make transparent the data lineage, showing the processes followed from source to target, and across multiple data management and governance competencies.
Traditional manual approaches to the integration of tools or building up of rules is time-consuming, but by using AI to orchestrate these tasks, organisations can reduce the time and resources needed, and across different toolsets. AI can take over much of these workloads, generating prompts or calling in other functionality only if it is necessary to address a problem.
Some tools automate some processes in the data journey, but many don’t invoke automatic prompts − so bringing in AI to the process allows it to ask for a decision and feed the decision back into the process to continually improve it.
Silos become less of a problem when all of the data is aggregated in a data lake with AI/ML deployed to look for common values or search strings and based on that, find linkages across all the data, much in the same way an internet search engine finds relevant information.
When it comes to applying analysis, once data has gone through the mill of quality and governance processes, the organisation is then able to apply analytical models to a trusted set of data with the silos bridged.
Share