A shift in perspective
We are currently witnessing a shift in attitudes toward data. Data governance used to be about safeguarding, avoiding data loss and regulatory non-compliance. But with organisations generating more data than ever, they increasingly need to improve accessibility and derive more value from it.
This is easier said than done. Data protection regulations are becoming more stringent. In Europe, the EU’s General Data Protection Regulation (GDPR) – enacted in the UK as the Data Protection Act (DPA) – introduced strict new rules, with significant penalties for failing to meet obligations. Meanwhile, increased digitisation has increased the complexity of the IT estate.
The internet of things (IoT) has seen analogue infrastructure evolve into networks of data-generating assets. The Covid-19 crisis has accelerated migration to the cloud and normalised remote working. Organisations are now increasingly disparate and dependent on digital technologies for even the simplest of tasks.
Split responsibilities
Beyond technology and regulations, the biggest challenge businesses face is operational. Across every enterprise, responsibility for governance has split. At leadership level, this was once the sole domain of the chief data officer (CDO). But now it is shared with the CSO, the CRO, the CGO, even individual business unit leads.
Unsurprisingly, there have been negative impacts. Co-ordinating this group is difficult, so data initiatives often happen in silos. This leads to fragmented data estates, prevents holistic oversight governance, and weakens the compliance and security posture. Even more concerning, this fragmentation undermines the overarching business strategy by removing the ability to drive competitive advantage from data.
The ability to collect, analyse and derive actionable insights from data efficiently has become central to identifying inefficiency and opportunities in business. But data science tools like artificial intelligence (AI) depends on reliable data. Without it, we cannot trust the outputs they generate.
Forecast to add $15 trillion to the global economy by 2030, according to the PwC global artificial intelligence study, optimising information architecture for AI should be a strategic priority. Just as every business today needs a website, it will soon become impossible to conceive of organisations not using data science in some capacity.
The AI ladder
According to IBM’s recent study on global AI adoption, more than one in three businesses cite difficulties along their journey to AI. Optimising data architecture is a four-stage process, which we refer to as the AI ladder:
- Collect data: Locate data, store it, ensure accessibility;
- Organise data: Catalogue and clean data to ensure accuracy and compliance; set policies for use;
- Analyse data: Build and deploy AI models; monitor performance over time, update as required;
- Infuse AI throughout the business: Employ AI to automate rote tasks, optimise processes, generate insights.
While many organisations collect data, many don’t know what data they have, where it is, who is using it or for what purpose. Adopting AI too early and without intent can be costly. Similar problems tend to occur – data governance teams solely focused on compliance and security, business units hiring data scientists to develop applications based on siloed data, minimal collaboration between stakeholders.
What happens? Unsurprisingly, results are poor. Too much time is spent hunting down data. Data models contravene governance. Or data can’t be scaled to provide useful insights. AI bias can result in highly problematic outputs. Ultimately, little value is delivered.
Optimising IT for insights
Often, efforts to navigate the AI ladder are hampered by IT that impedes the movement of data. As organisations migrate to the cloud, it is important they use platforms built on a foundation of open-source and hybrid cloud. This ensures data is not trapped within proprietary eco-systems, allowing it to be accessed across any environment, whether on-premise, public or private cloud, a data centre, or at the edge.
The choice of technology partner is also important. Modernising business processes, information architecture and technology strategy will inevitably involve some degree of disruption. For many, it requires a cultural shift, too. A partner with enterprise expertise and an understanding of specific industry challenges is essential to smooth the transition. This is particularly relevant in highly-regulated industries like finance or healthcare.
Driving data value with Denmark’s biggest bank
IBM’s work with Danske Bank, Denmark’s largest financial institution, provides an example of the organisational transformation required. Servicing over five million customers across Europe, the company had big questions it needed to answer – how could it ensure the right people have access to and are using the right data at the right time?
Previously considered an IT problem, the first challenge was to alter perceptions and bring in stakeholders from across the business. Driving cross-divisional collaboration from the start of the project was key to building a new data ownership model that was fluid and responsive to changes across the bank.
The bank also modernised the foundation of is metadata management practices, using AI to discover, curate, clean and secure data in an automated, agile way. Now, Danske Bank has the trusted platform it needs to drive real business value from its data.
Regardless of whether an organisation is ready to begin climb the AI ladder, this approach will increasingly be the basis on which competitive business models are built. We need to stop looking at data governance as something companies do to protect assets, and start thinking about how it can proactively impact the bottom line.
Jay Limburn, director and distinguished engineer, IBM Data and AI
