Four Questions to Consider When Navigating the Rapid Evolution of Generative AI

Generative AI’s (gen AI) capabilities seemed startlingly novel a year ago, when ChatGPT’s release led to an explosion of public usage and, simultaneously, intense debate about its potential societal and business impacts. That period of initial amazement and suspicion has given way to business urgency, as companies scramble to adopt gen AI in ways that leverage its potential for maximizing workforce productivity and profitability.

However, intentions to implement differentiated gen AI solutions can quickly lead to roadblocks when a few core realities come into view:

• Where do we get the talent?
• How do we wrap our heads around the scope of possibility?
• How do we build trust in AI?

I’ve spent over 20 years helping large corporations gain significant market footholds by optimizing data, analytics and AI – and have seen firsthand how a holistic data strategy is foundational to avoiding wasted resource investment and achieving success in a new competitive landscape. Of course, it’s easier to understand the value of a modern data infrastructure than it is to build one, including adapting your workforce accordingly, avoiding common pitfalls and retaining customer trust throughout the process.

Here are four questions firms must first grapple with to help ensure generative AI solutions boost their success instead of risking their reputation and standing.

In the past year, it seems everybody’s become an expert in generative AI. The data market is fragmented from an architectural perspective, thanks to the emergence of separate data management architectures, and the impact of generative AI on talent and technology is unprecedented. When organizations manage their influx of data on their terms, according to their goals, the solution space grows, but so does the fragmentation of data management tools. Running a host of different initiatives to address this data sprawl further dilutes business density, squeezing out room for revenue. 

The data strategy is the baseline which establishes a firm’s overall business strategies, priorities and investments. The noise of generative AI tends to invite what often seems like a frenzy of hasty business investments in its capacities. The noise is tempting but can also be naive and immediate. A holistic data strategy outlines what is actually important to an organization and how it paves the way for proper infrastructure investment.

The widespread accessibility and ease of use of generative AI can also lead to plenty of mediocre output, whether that’s  a script, an image or an interview. How do we adjust existing talent skill sets to guide gen AI to produce usable content the technology could not produce without the help of human creativity? 

Even the automation of business functions requires talent to materialize it. New engineering skills (for example, the responsibility of finding the right AI “prompts”) and AI-governors are needed to effectively manage and govern generative AI.

Talent also requires fluency in AI usage or at least in understanding it, especially when generative AI’s potential virality impacts all verticals and branches of the organization. The pace at which changing capabilities potentially disrupt certain business functions can create rifts between those too slow to adapt and those well-equipped to track generative AI’s changes and effects — the winners and losers of the changing ecosystem, in other words.

Fluency courses like those offered by Snowflake are possible solutions to help understand what generative AI even is and, more important, how its ceaseless changes might alter business functions or data strategies. While generative AI can be used by anyone, the question remains: does everyone with access to these tools in the organization know how to use them?

Trust in AI will be hard won, considering the potential for bias depending on the quality and diversity of the data that models are trained on. Yet training trustworthy generative AI models comes with a significant cost: The potential risk of bias in available training data could compromise business integrity. Marginalized demographics unrepresented in available training data could be excluded from models designed for generating credit card approval requirements, for example. Or in some cases the historical data used in training may simply be too old to produce any equitable results. 

Ethical governance and regulation of generative AI training emerges in this speculative and frenzied ecosystem as a safeguard against deregulated misuse of the technology and the potential damage that could result from trusting AI hallucinations as fact. As effective as generative AI might be in realizing higher workforce productivity, its advantages should not come at the cost of trust and public safety.

To learn more about preparing for generative AI’s impacts and how Snowflake can help with its adoption, check out my full interview on DCN’s channel.