Your 5-Step Journey from Analytics to AI

Most organizations have come to understand the importance of being data-driven. To compete in a digital economy, it’s essential to base decisions and actions on accurate data, both real-time and historical. Data about customers, supply chains, the economy, market trends, and competitors must be aggregated and cross-correlated from myriad sources. 

But the sheer volume of the world’s data is expected to nearly triple between 2020 and 2025 to a whopping 180 zettabytes. That makes it unfeasible for mere mortals to harness it strategically without some automated assistance. This is where artificial intelligence (AI) comes in.

How do you introduce AI into your data and analytics infrastructure? To companies entrenched in decades-old business and IT processes, data fiefdoms, and legacy systems, the task may seem insurmountable. It’s not—but it might require some top-down rethinking of workflows and mindsets.

 1. Develop a strategy to liberate data 

If your company currently walls off certain data to limit its accessibility to certain departments, liberating it across the organization is an essential first step. The goal is for all decision-makers—from the CEO to the front-line employee—to be working with a current, holistic version of the truth. Seeing and analyzing partial data, data that’s out of context, or stale data can mislead decisions and have unfortunate consequences. 

To enable greater access to data, initiate a conversation between business and IT leaders to review how data currently flows in your organization. Determine who “owns” the data and what controls are in place for accessing it. This investigation will help you identify the organizational and infrastructure changes needed to open up data access across the company. 

2. Consolidate data 

Consolidation creates a single source of truth on which to base decisions, actions, and reports. Which type(s) of storage consolidation you use depends on the data you generate and collect. 

One option is a data lake—on-premises or in the cloud—that stores unprocessed data in any type of format, structured or unstructured, and can be queried in aggregate. Another option is a data warehouse, which stores processed and refined data. If you opt for a data warehouse, define master data to enable easy search queries. This is important, for example, if different departments organize and classify the same information in different ways. You want a single record that’s consistently searchable and accessible. 

You might have reasons to run a data lakehouse and possibly other, purpose-built data stores. It’s important to connect them all in a secure, well-governed system. Then virtualize your data to allow business users to conduct aggregated searches and analyses using the business intelligence or data analytics tools of their choice. 

3. Set up unified data governance rules and processes

With data integration comes a requirement for centralized, unified data governance and security. This task involves setting consistent policies across data, services, and applications that strike the right balance between data security, compliance mandates, and worker productivity. 

Refer to your Step 1 inventory of data resource ownership and accessibility. Create a map of which resources should be made accessible to whom. Adopting a zero-trust security approach limits access to only those users and applications that require it and has quickly become a best practice. It’s also hugely beneficial to deploy a data catalog or other centralized management mechanism that automatically discovers, tags, and catalogs data so you can manage and audit your policies all in one place. 

4. Consider deploying analytics-as-a-service 

To allow business units to access and use the data in a cost-effective, secure manner, you can create an analytics-as-a-service model. Each department or business unit is assigned a “workplace” within the unified data infrastructure. Business units can bring in their own data, and access the superset of data aggregated from all the other different sources. 

5. Start small with AI

Becoming data-driven and automating with AI and machine learning (ML) algorithms can seem overwhelming. It’s best to break it into manageable pieces. For example, consider identifying a single, measurable use case for AI that gets you started with a proof of concept (PoC) that could be a stepping stone to other initiatives.

Focus on a specific business problem to be solved. Think it through, end to end, from implementation feasibility to identifying the key performance indicators (KPIs) you’ll use to measure return on investment (ROI) and project success.  An example might be a manufacturing company looking to detect faulty parts as they arrive at its factory in order to reduce the number of defective products coming off the assembly line while lowering associated warranty repair costs. 

Outcomes you can expect

Data analytics powered by AI has created a wealth of business opportunity. Insights gathered from real-time and historical data allow leadership teams to quickly adjust to unexpected change, improve customer experiences, become more predictive, and use automation to streamline processes for operational cost and time savings. Dell Technologies’ research indicates that organizations that have integrated AI into their operations have seen a 92% reduction in downtime, a 39% increase in IT productivity, and a 19% drop in the cost of business operations, on average.