Business efficiency a place to start with generative AI

Business benefit of generative AI

Improved efficiency is one of the great promises of generative AI.

Organizations are driven by data. It's what informs the decisions that lead to action. But many organizations make poor use of their data by leaving much of it untouched because they have more data than they can manage or by failing to use data in real time because processes are manual and take too long.

The result often is decisions based on only partial data or decisions based on potentially outdated data, which lead to outcomes that fail to maximize an organization's potential.

Generative AI has the potential to eradicate many of the problems that organizations face when attempting to derive maximum value from their data.

It can automate many of the data management processes that previously required manual supervision and operation. It can also virtually eliminate the need for users to know code, which opens data exploration and analysis to any business user rather than only those data experts trained in coding languages such as SQL and Python.

So inefficiencies are where generative AI should be applied, according to Dalloul.

Anywhere within a business that data is used to inform processes and decisions -- for example, customer service, marketing, fraud detection or content creation -- generative AI can be used to improve those processes and decisions.

"[Generative AI] models are really good at summarization, content creation, understanding and generating code, text and images," Dalloul said. "When you add those together and complement them with other AI services ... enterprises are delivering remarkable results."

Jonathan Cohen, vice president of applied research at Nvidia, similarly said improving efficiency with automation is an obvious place for enterprises to begin applying generative AI to their business.

He noted that organizations have extensive amounts of data related to their operations. Given the capabilities of generative AI, all that data can be made meaningful.

Data management and analytics vendors are not yet providing extensive tooling to enable organizations to build generative AI models and applications. Many features have been introduced, but the vast majority are in preview.

But just as an organization's data engineers and data scientists can build their own data management and analytics capabilities, they can train their own generative AI models with public tools, develop their own language models from the ground up and build their own generative AI-powered applications while they await capabilities from vendors.

"Enterprises collect data -- that's what they do," Cohen said. "The real use case is to take that data and turn it into automated systems that can make decisions or recommend actions based on the patterns of the way a business has historically functioned."

In particular, unstructured data can be used with generative AI to improve efficiency and drive automated processes better than in the past, he continued.

Unstructured data is data such as text, audio or video that does not have a predefined number or table and therefore can't be classified and organized as easily as structured data, which is predefined by numbers or tables.

To use unstructured data along with structured data for a thorough view, unstructured data has to be given structure. Historically, that meant data engineers had to assign numbers to unstructured data -- an overwhelming manual process because most data is unstructured.

Generative AI, however, can be programmed to automatically give structure to unstructured data by assigning it numbers, suddenly giving organizations significantly more data that they can use to train models and inform decisions.

"Now, you can take all that unstructured data and customize a model, build AI that has access to all that information, and formulate answers and responses and look for patterns," Cohen said. "That's accessible today and very achievable."

Generative AI governance

While targeting inefficiency is a clear place for enterprises to start applying generative AI to their business, it has to be done carefully, the panelists noted.

Just as organizations carefully control which employees have access to what data, they must put access controls on their generative AI models and applications to ensure their data and AI remain secure.

In fact, the relationship between data and AI is so intertwined that there is no AI governance without data governance, according to Christian Kleinerman, Snowflake's senior vice president of product.

"There is no AI strategy without data strategy, and data strategy includes governance," he said. "That has to carry through."

So as enterprises begin applying generative AI in an effort to improve their business, it's imperative that they do so with AI governance measures in place that enable users to derive value from the AI while simultaneously protecting the organization, according to Dalloul.

"You're dealing with client data, you're dealing with your own data, and that content should not be made available to someone who [is not supposed] to have access to that information," he said. "You have to partition the data and make sure that role-based access control is in place and implemented within the [generative AI] architecture."

And that applies whether an organization customizes a public LLM such as ChatGPT or Google Bard to its own needs or develops its own language model, he continued.

"You need to have a very deep understanding of the roles, of the service, of the use case and how you manage the data," Dalloul said. "That's the baseline."

Eric Avidon is a senior news writer for TechTarget Editorial and a journalist with more than 25 years of experience. He covers analytics and data management.