Don’t let your GenAI project fail before it begins

The excitement around generative AI (GenAI) over the last year has led the C-suite to drive use within their organizations, but the harsh reality is that many enterprises simply aren’t ready for the technology on a wide-scale basis. Thanks to rapid consumer adoption and the ease of tools like ChatGPT (see Exhibit 1), many executives wrongfully assume leveraging GenAI will be a simple, straightforward path to value creation but overlook the importance of addressing their data foundation. Building a proper data foundation will increase the quality and reliability of insight from GenAI while minimizing exposure to significant business risks.

We connected with enterprise data-management provider Syniti and its clients and partners to discuss the importance of building a solid data foundation before adopting GenAI and to gather some top implementation insights for enterprises embarking on the journey. There was a key takeaway thread throughout our conversations: Data is the key to unleashing the full potential of GenAI, and the success of an enterprise’s project almost entirely depends on its data quality.

Exhibit 1: Consumer adoption and ease of access drive rapid growth of GenAI

Sample: 104 enterprises actively exploring and deploying GenAI across the Global 2000
Source: HFS Research, 2024

Despite the excitement, many enterprises have yet to realize the true value of GenAI

The explosive growth of GenAI has left countless executives worried that if they don’t adopt the technology at the same pace as their competitors, they will lose a competitive advantage. As Naveen Gupta, Global Data Leader at IKEA, a global furniture manufacturer and retailer and a Syniti client, summarized, “We wanted to be in this space yesterday—it’s a challenge.”

Naveen Gupta isn’t alone in facing this challenge; more and more executives are finding that adopting GenAI isn’t as straightforward as initially seemed. The outcome is often a careful balancing act of deploying the technology in a handful of siloed use cases while developing a defined strategy for implementation at a much higher level.

While it is imperative that building a long-term data strategy remains the top priority, small introductory projects using small samples of reliable data can be excellent quick wins that demonstrate success to the C-suite and help lock in further investment.

That’s exactly the approach IKEA is taking. According to Naveen Gupta, IKEA has already been using GenAI for simple tasks, such as collecting and providing product information in an understandable format, while working with Syniti to develop a long-term data strategy to ensure future large-scale GenAI projects. Naveen explained:

The biggest challenge we’re facing in IKEA is having data management practices in place. We don’t have practices for data cleansing, strategy, and governance. We need all of that to make sure GenAI is a success.

– Naveen Gupta, Global Data Leader, IKEA

Building a robust data foundation must be the first step in your GenAI adoption journey

When it comes to GenAI, the familiar adage “garbage in, garbage out” could not be more true today. The quality of an enterprise’s AI model is nearly entirely dependent on the quality of the data fed into it. Large GenAI projects should only be tackled once a proper data foundation is established. Deploying the technology can be complex and lengthy, spanning model development, training, and testing. Enterprises that overlook their data set themselves up for failure before they even begin.

To build a proper data foundation and set yourself up for GenAI success, we recommend enterprises follow these key steps:

• Assess your landscape: Understand the current state of your organization’s data and build a robust plan spanning data collection, storage, integrity, and quality.
• Foster a data-driven culture: Create a mindset shift by aligning business and IT teams with shared business goals that place value on data-driven decisions, ensuring relevant stakeholders understand the importance of a proper data foundation.
• Establish data governance: Develop internal standards that comply with external regulatory requirements, highlighting how data should be gathered, stored, and processed.
• Continuous monitoring: Recognize that building a solid data foundation isn’t a one-time task, and review your practices regularly to ensure your organization maintains good data practices to maintain data integrity.

Lazy GenAI might be suitable for small initiatives, but enterprises should turn to data management firms for large projects

Not addressing your data and its quality might be acceptable for simpler, small-scale GenAI projects with minimal investment, but larger enterprise projects require significant investments. Subpar returns are unacceptable. We connected with Lenno Maris, Chief Data Officer of leading chemical manufacturing firm Caldic, a Syniti client:

GenAI technology comes back with answers based on what the user asked. In that sense, it’s a straightforward, lazy tech. If your organization is reluctant to put in the effort to assure data quality, the GenAI will be lazy in return. Proper AI requires a proper data foundation.

– Lenno Maris, Chief Data Officer, Caldic

A recent HFS survey, in partnership with Syniti, found that one-third of executives believe less than half of their organization’s data is actually consumable, highlighting just how many organizations aren’t ready for GenAI. The good news is that to ensure the right infrastructure is in place to break down silos and deliver trusted, usable data to fuel GenAI models, data management firms like Syniti are ready to support enterprises in addressing their data quality.

However, the importance of establishing a proper data foundation extends far beyond just improved outcomes.

An improper data foundation could lead to significant business consequences

The fallout of deploying GenAI without a proper data foundation can extend much beyond poor-quality outputs. If biases exist within the data fed into GenAI models, such as gender or racial biases, these biases will be quickly replicated at scale within an organization. Biases could lead to real-world consequences, for example, enterprises discriminating against certain individual job or loan applicants and healthcare providers giving incorrect advice to minorities. In turn, this could cause reputational damage, have regulatory implications, and concern investors.

Nobia, a European kitchen manufacturer, highlighted these concerns in our conversation. Tobias Nilsson, Group Data & Information Architect of Nobia, explained:

One of the hardest parts of AI is bias and privacy. It’s hard to know if we can trust the data and that it’s not going to create something harmful. There are good frameworks for detection of bias—we need to use them.

– Tobias Nilsson, Group Data & Information Architect, Nobia

The Bottom Line: Generative AI isn’t a quick solution to unlocking new value. Enterprises must first ensure they have a solid data foundation.

The true value of GenAI remains unrealized by many organizations thanks to their legacy data estate, and industry leaders at Caldic, IKEA, and Nobia confirmed it throughout our conversations. Leading executives must not get swept up in the excitement of tools like ChatGPT and instead work with data management experts to develop a strategic approach to establish a proper data foundation.

Your GenAI journey truly begins when tackling your data foundation and ensuring it can deliver trustworthy and unbiased insights to demonstrate real return on investment and scale with business needs.