It's no secret that generative AI technology has spawned tremendous new opportunities for enterprises. Like digital transformation or the move to the cloud in earlier years, the generative AI revolution has become the latest trend that most businesses must embrace if they want to remain competitive.
The question enterprises face now is how to do that. A large part of the answer involves data management. Because data is the foundation on which generative AI tools operate, taking full advantage of generative AI requires businesses to take their data management strategies to the next level.
With that reality in mind, I'd like to walk through five key components of an effective enterprise data management strategy for the generative AI age. As I'll explain, some businesses already have these practices in place, but those that don't will need to invest in new initiatives that ensure they can power generative AI software with the information it needs to work to maximum effect.
Before looking at which data practices enable generative AI for enterprises, let's first go over the ways in which enterprises can use generative AI and the role that data plays in these use cases.
One way that businesses can take advantage of generative AI is to use tools like ChatGPT or Google Bard, which have been at the forefront of headlines about AI technology in recent months. Consumers of these tools have already realized an increase in productivity in areas such as generating process documentation and meeting summaries, gaining competitive insights, or even launching email marketing campaigns.
However, these tools, which are designed for consumers more than businesses, are not likely to be at the center of most AI strategies for the typical enterprise. They are trained only on public data, and on their own, they can't do much to accelerate workflows that require access to private business information. They may be useful for tasks like helping employees to search the internet, but not for much more than that.
Instead, the most innovative enterprise use cases for generative AI are likely to center around the deployment of purpose-built AI tools that train on business's internal information — such as documentation, manuals, private emails, and internal databases — to help complete tasks that require specific understanding of a business's unique operations. For instance, enterprises could build AI-powered chatbots that leverage internal product catalogs to interact naturally with external customers, or that help onboard new employees by interactively explaining to them how business processes work. They could even deploy AI tools that help them generate custom contracts or write code that powers their business applications.
The list could go on — there are many more ways in which enterprises can leverage generative AI — but the point I want to emphasize is that virtually all AI use cases for enterprises require AI tools to have access to internal company data, which developers can use to train the algorithms behind the tools. Without enterprise-specific data, enterprise-specific AI tools just don't work.
The typical enterprise has plenty of data. However, the ability of that data to enable custom generative AI tools depends on how well it aligns with characteristics like the following:
Ideally, enterprise data should be structured in such a way that it's easy for developers to identify relevant data sources for AI training, then connect algorithms to those sources.
Siloed, scattered, or unlabeled data makes it hard to take advantage of generative AI. To succeed, companies need to know where their critical business data exists and how to access it.
Even in organizations where data is well-organized, it's rarely stored in a centralized fashion. Instead, data is typically spread throughout the organization, with different departments or business units maintaining different data repositories.
Moving that data quickly and efficiently into AI tools for training purposes requires data pipelines that can connect disparate data sources together and feed them into algorithms.
It may seem obvious that data must exist in digital form for AI tools to train on it, but the point is worth noting because plenty of enterprise data still exists in non-digital form. Companies may have employee handbooks that exist only in print, for example, or they might manage customer contracts using paper forms.
Data sources like these will need to be digitized so that generative AI tools can train on them.
The old saying "garbage in, garbage out" certainly holds true in the context of generative AI training. You can't train your algorithms effectively if you lack accurate and reliable enterprise data to train them with.
For that reason, now is the time for companies to invest in data quality initiatives that will help ensure their AI tools have access to high-quality training data.
Last but not least, you need a critical mass of data to take advantage of AI. Companies that have not been systematically generating and storing digital data will find themselves at a disadvantage if it means they don't have enough data records to allow AI tools to understand the context of incoming requests from users.
I know based on my experience in the field that the extent to which enterprises are prepared to meet the data challenges that generative AI presents varies widely. Some companies have been making extensive use of AI technologies for use cases like predictive analytics for years, and they already have structured, high-quality data volumes in place. They can make the leap to using custom generative AI tools relatively easily.
But other companies — those that don't have extensive AI or business intelligence processes already in place — have catching up to do if they want to take advantage of generative AI. The good news for them is that there are tools and strategies that can help, if they know where to look. Software vendors are in the midst of unleashing a torrent of new tools — like Google's Gen App Builder and Model Garden, to name a couple cutting-edge low/no code solutions — that can help companies create the type of enterprise-focused AI software I've described above.
Every company has the potential to make great use of AI to accelerate productivity and optimize efficiency. What will separate winners from losers in the midst of the generative AI revolution for enterprises is how effectively companies prepare their data. Virtually every enterprise has the information necessary to enable next-generation AI technologies, but not all of them have information that is AI-friendly in its current form.
Matt Kelberman is the director of data analytics at 66degrees.