The big leap: Getting data AI-ready

An EY online survey revealed that 90% of respondents who work a desk/office job and are at least somewhat familiar with AI say that their organization uses at least one AI technology, with GenAI topping the list. Most say that using AI at work will make them more efficient (82%) and more productive (81%), and will let them focus on higher-value work (81%).

GenAI, in particular, has taken off in areas such as marketing, content management, contract management and the call center. This is on top of applied AI applications that are used to generate predictive insights that lead to more intelligent decision-making.

The survey found that, despite the mounting enthusiasm for these and myriad other use cases, there is anxiety about AI, including legal risks (69%), ethical and moral considerations (67%) and cybersecurity risks (67%). The data being used to fuel AI models is also under scrutiny. Almost half (48%) of respondents to the 2023 Foundry AI Priorities Study said that data quality and quantity issues could impede genAI implementations. They’re also concerned about data privacy and ethical considerations (41%), data variability (37%), and unlabeled or weakly labeled data (29%).

As organizations prepare to leverage AI for competitive advantage, many lack a robust foundation that primes their data for effective AI use at scale. These foundational components include data management and governance practices, AI governance and mature platforms for operationalizing the technology.

“AI needs management and governance to ensure ethical, security, economic and legal concerns are addressed, and associated risks are mitigated,” says Zakir Hussain, EY Americas Data Leader. “All of those need to be understood, planned for, and managed, and many companies are just not ready.”

Getting your data ready for AI

Where do most companies fall short in pursuit of AI-ready data? Maturity levels vary, but most enterprises are lacking in one or more areas of data management, leading to an abundance of proof-of-concept (POC) project failures, Hussain says.

Data that is truly optimized for AI has four key characteristics, he explains:

• The data is reliable.
• It’s accessible and scalable.
• It’s visible with the proper context and consistency.
• It’s trusted and secure.

To get there, organizations need the following technology and process pillars in place:

• An AI-ready data strategy that is focused on business value drivers
• Data governance for data consumers to explore, discover and use data sources efficiently
• Master data management for a single source of truth
• A risk and compliance framework for properly managed risks and access controls
• Data quality and data wrangling capabilities to ensure solutions with curated and validated data
• A robust data architecture for rapidly piloting and scaling solutions

“All of these elements determine whether data is ready for AI or not,” Hussain says. “If data isn’t contextualized, tagged or mastered properly, or if data lineage is broken, it’s bound to generate wrong predictions and drive the wrong decisions.”

Beyond investment in the right technology and data architecture, CIOs can foster data centricity and more mature practices by shifting data and analytics projects and culture toward a product management mindset. The traditional approach of data councils and enterprise data governance has moved the needle on establishing common taxonomies, developing shared data policies and curating multiple data sources. Yet, those practices often don’t go far enough to change culture or solve systematic data problems. Revamping organizational structure to establish product owners, creating practices for managing data throughout its entire lifecycle, and building teams that are focused on outcomes and results will go a long way in creating AI-ready data while increasing demand and funding for data and analytics initiatives.

4 steps to better data management

We offer a number of other recommendations that are designed to promote AI-ready data and data management practices:

Develop a proper roadmap: No data or AI initiative will succeed on its own without a formal business case and an intentional set of business outcomes. CIOs and data leaders need to thoroughly examine the value drivers for their individual organizations and formulate a data and AI roadmap that is aligned to those specific business goals and in cooperation with their business partners.

Place relentless focus on value: Too many AI and data-driven pilot projects fail because they’re too narrow or not tied to a specific business outcome. As opposed to doing a POC with a small data set just to establish relevancy and demonstrate success, Hussain advises CIOs to conduct pilot projects with real data that addresses a high-priority business problem or opportunity. “Stay focused on results and build a pipeline of use cases that are very results-oriented,” he says. “That will pave the way for the company to mature their data readiness.”

Leverage AI (including GenAI) for efficiencies: GenAI can play a role in streamlining data management and data quality practices. For example, it can help to automate data quality checks, bring greater intelligence to master data management and help ensure dynamic data governance policies. “Things are changing so fast – you can’t set a governance policy and let it sit there for the next few years,” Hussain says. “Using GenAI to set up dynamic data policies allows systems and organizations to adapt and evolve policies.”

Move quickly, but prepare for a marathon: As AI looms large, business leaders are hankering to move fast. However, being too aggressive can result in data management missteps that slow down progress. It’s critical to develop a holistic strategy to modernize and ensure AI-ready data. “This is not a sprint,” Hussain says. “Data modernization needs to be carefully thought out, and you have to build a holistic strategy focused on business value drivers, and that’s not just about technology.”