Smart thinking: Why data is key to successful AI projects

Artificial intelligence (AI) is predicted to transform business and become commonplace in our daily lives – in fact, it has already begun to do so. AI can automate mundane tasks previously performed by humans, saving countless hours, enabling more accurate predictions and better decisions.

Data is the lifeblood of AI. However, many organisations struggle to produce accurate and up-to-date quality data for their machine learning and AI technology to learn from. It is the weak link in many AI projects.

According to 2020 Forrester Consulting study commissioned by IBM¹, organisations that scale ...AI are seven times more likely to be the fastest-growing businesses in their industry. However, many are struggling to adopt AI at scale because of gaps and errors in their data.

Improve data quality for smarter AI

Data repositories can help businesses organise their data and improve its quality.

Standard Bank of South Africa raised the quality of its data from six per cent to 98 per cent using IBM DataOps software. It now has a data catalogue to help it meet regulatory and compliance requirements.

Before working with IBM, the bank − which operates in 20 countries in Africa and has reported assets of approximately US$157 billion in 2019 − was investing tens of millions of dollars on data fixes in disparate places, says Simphiwe Phakathi, Executive Head: Relationship Banking PBB Africa Regions at Standard Bank Group & Dumisani Mthimkhulu, Head of Data Asset Management Platforms at Standard Bank Group. “We needed a disciplined data lifecycle approach that was sustainable.”

For example, IBM’s framework, called the AI Ladder, based on thousands of AI engagements, helps an organisation determine its individual AI journeys, and provides structured milestones to turn its AI aspirations into business outcomes.

Data is the foundation for any successful use of AI

Business customers face three main challenges when approaching data in their AI projects – collecting the data, analysing it and infusing it into their operations. Often, data is poor quality, full of errors or out-of-date, and scattered across dozens or even hundreds of IT systems, in a large, global company. Collecting this data, then analysing it and infusing it when there is no central system is a daunting task.

Creating a single repository to store, verify, organise, and analyse data is a crucial early stage in an AI project. After the data foundation is in place, an organisation can infuse the data into AI systems, using open source technology and private and public online cloud systems.

As an example of successful implementation of the AI Ladder, IBM worked with a large European Bank to streamline its data collection and data governance using IBM’s cloud-based data catalogue and platform, Watson Knowledge Catalog. The data platform is helping the bank to use AI to predict IT system breakdowns and fix issues in advance, so that customer service is not affected.

Energinet, Denmark’s transmission grid operator is harnessing the power of AI to help it better manage the electricity grid and reduce IT downtime. It worked with IBM to pilot a multi-cloud solution based on IBM Cloud Pak for Data and Watson Studio. The technologies help Energinet glean operational predictions from big data. In a pilot project, Energinet and IBM tested the AI capability by simulating outages with known causes and remedies and then compared the outcomes with experienced operators’ conclusions.

Cloud technology, analytics and data platforms can help facilitate AI projects, but no matter how advanced your AI is, it will only be as good as the data that is fed into the system. Creating a strong data foundation is always a smart move to put AI to work.