Data strategy: The key to unlocking AI potential

By Ryan Levman, Director Data Engineering, Bell Canada  

In the fast-paced world of artificial intelligence (AI), data is critical. I’ve been working with data to power AI models since I joined Bell over 11 years ago, and it’s been a game changer. Even before ChatGPT became a household name, I had the opportunity to analyze unstructured text data to unlock the potential of large language models (LLMs) and see firsthand how instrumental data is in mobilizing generative AI use cases at scale. It’s made our business more agile, efficient and productive. 

However, the evolution from gathering data to developing a cohesive data strategy can be a challenge – particularly for companies with complex IT landscapes. Bell is a perfect example – serving Canadians across multiple brands, each with its own unique products, ordering systems and billing systems. It's an evolving landscape, with mergers, acquisitions and system consolidation projects happening all the time.  

Maintaining a single source of truth for all our customer metrics and attributes is definitely a balancing act. While developing our AI strategy here at Bell, our data modernization work has given us valuable insights into what makes a data strategy successful. Here's what we've learned. 

Align your business goals 

Too often, organizations get caught up in the excitement of AI possibilities without grounding their efforts in practical business needs. To ensure your AI initiatives deliver real value, start by defining what success looks like for your organization. Whether it's improving customer experiences, streamlining operations, or unlocking new growth opportunities, your objectives should serve as the north star for every decision in your data strategy. 

Remember: your data strategy is the foundation for your work in AI. By setting clear and measurable data goals, you create a path for effective AI implementation and foster a data-driven culture that benefits both your employees and customers. 

Start with quick wins 

Your goal may be a complete data transformation, but your business – and stakeholders – will benefit from activities that provide measurable progress in the short term. Here at Bell, for example, we identified an opportunity to improve service by streamlining customer data within a single interface.  To break down silos and optimize support operations, we focused on unifying transactional customer data and automating workflows.   

Early wins like these showcase the value of data and set the stage for more complex and impactful projects. More importantly, testing different improvements will give you invaluable experience and the understanding you need to guide your overarching strategy.  
  

Deepen the understanding of your data  

Before you can leverage AI, you need a holistic view of your existing data. This means not only knowing where your data is stored, but also understanding its quality, structure, and relevance to your business goals. Conducting a thorough assessment of your data landscape is essential for aligning AI initiatives with strategic objectives. This process involves evaluating your data sources and how information flows through different technology systems.  

This knowledge is vital, since AI systems are only as good as the data they are built on. This process will uncover potential gaps and key analytics use cases as well as data inconsistencies and areas of improvement, ensuring that when AI is applied, it delivers actionable and reliable business outcomes.

Prioritize security and privacy

It is important to ensure that security, privacy, and AI ethics are not afterthoughts. By making them fundamental aspects of your data and AI architecture, you reduce risk and also minimize the need to retroactively solve key concerns. Adopt a 'security-first' approach, implementing robust security measures, clear data privacy policies, and unwavering ethical AI principles aligned with your business goals. This proactive approach not only safeguards critical data but also fosters trust, minimizes compliance risks and accelerates AI project deployments.  

At Bell, we have the expertise and solutions in place to support a secure transition of your workloads to the cloud, allowing a seamless and protected migration. Our network is built with inherent security features, providing an extra layer of protection for your data in transit and at rest. 

Design an agile ecosystem 

Your data and AI architecture will become a core part of your business. Every organization has unique data needs and a distinct path to modernization. Accordingly, there isn’t a one-size-fits-all solution for a modern data architecture. Consider your current data requirements and choose an approach that can evolve with your business over time.

A modern data architecture should include the following elements:

 A solid foundation:

• Networking and cloud fundamentals: the backbone of your data flow.
• Data integration and storage: how you bring your data together and keep it safe.
• Data transformation: cleaning and preparing your data for analysis.
• Data governance: ensuring data quality, accessibility, security and privacy.
• Analytics and reporting: extracting insights from your data.
• AI/machine learning (ML): putting your data to work with AI models.
• Production and machine learning operations (ML Ops): deploying and managing your AI models.
• Security/audit: keeping your data safe and secure.

 The right tools: Consider factors like seamless integration, performance, scalability, costs and required customizations. 

 Sound operational practices: It's not just about the tools; it's about how you use them. Develop strong operational processes to make your data strategy work.  

Implement strong data governance 

Cloud scalability enables enterprise-wide data colocation, removing the physical silos that existed in the past. To achieve maximum business value from this collocation investment, it’s important to pay close attention to how you organize and govern your data. This includes how data is accessed, documented, recovered and protected. To do this successfully, you need a solid governance approach that accounts for the following: 

• Centralized vs. decentralized: Decide how you will manage your data governance.
• Metadata requirements: Document your data to make it understandable and usable. 
• Data cataloguing: Create a system to find and access your data easily. 
• Metric and attribute glossaries: Define terms and calculations that link to business KPIs. 
• Access management: Control who can access what data. 
• Data discovery: Make it easy for people to find the data they need. 
• Interoperability: Connect your data whenever possible. 
• Data products: Define ownership, value and quality for your data. 
• Virtualization strategy: Know when to use virtualization and when not to. 

Investing in these key areas – especially when a substantial number of data producers are collaborating in a single cloud environment – will help you keep your data organized. It will also help create a consistent experience for data consumers. 

Build your data strategy with Bell  

Ready to elevate your enterprise data strategy? Our Bell Data Strategy Accelerator program helps you build a roadmap for success, aligning your vision, talent and technology to achieve tangible business outcomes. 

We bring deep expertise in building, scaling and maintaining enterprise-grade data platforms. This means we understand the challenges you face and can provide the solutions you need to succeed.  

Let’s work together to unlock the full potential of your data. Discover more AI and data strategy insights. 

About the author

Ryan Levman has been with Bell for 11 years, starting as a data analyst after graduating from the University of Waterloo’s Management Engineering program. He now leads the consumer data engineering and AI team and is passionate about intersections of people, data, technology and business value.