Oracle users share challenges, successes with cloud, AI

Dollar General's multi-cloud operation

Discount retailer Dollar General runs its operations on Oracle Cloud and Google Cloud. The company's use of the two services exemplifies how a multi-cloud strategy evolves and why Google launched an Oracle database service this year.

Dollar General uses BigQuery, Google's fully managed data warehouse, for analyzing the store's data, said Anil Garg, senior manager of database management services at Dollar General.

The company uses Google's machine learning models to forecast the products most likely to sell well in its 20,000 stores during different times of the year. The company will ship products based on predictions that are up to 90% accurate, according to Garg.

Dollar General is in the early stages of its migration to the cloud. More than 95% of its data is in on-premises Oracle databases.

Now that the Oracle database is on Google Cloud, Dollar General will eventually migrate the data to Google. Having an Oracle database on both ends means Garg only has to back up the data, migrate it and test the performance of the database.

"This is a much easier lift," Garg, who expects the migration to take a few months, said.

Dollar General executives preferred to stay with Google because "they have very good faith in Google AI," Garg said. When the company chose Google, Oracle didn't have a competitive product.

"Last year, Oracle AI was a little bit behind," Garg said. "They are just catching up with [Microsoft] Azure and Google."

Nevertheless, some of Dollar General's business has gone to Oracle. Last year, the company opened stores in Mexico and operates them through Oracle Cloud retail software, Garg said.

Building AI into business operations

TD Bank, which uses Oracle Cloud software for financial planning, budgeting and reporting, has been using AI since acquiring research lab Layer 6 AI in 2018. The bank is gradually rolling out AI services across its operations, said Morgan Klein-MacNeil, vice president of finance and risk.

"For us, it's really focusing on human-in-the-loop use cases," Klein-MacNeil said. Also, the bank's AI development has prepared it for using Oracle AI features as they become available in the company's software, she said.

Core to the bank's use of AI is the quality of the underlying data, she said. "When we look at the opportunities with GenAI, none of that is possible without a clean, trustworthy dataset." Klein-MacNeil added that her team does not intend to build and maintain its own large language model.

Sy Choudhury, director of AI partnerships at Meta, joined Klein-MacNeil in a panel discussion led by professional services firm PwC. Meta uses Oracle Fusion software for its business operations.

Choudry addressed the security side of GenAI, saying Meta uses four AI models to power its services. One is a safety model that acts as a filter to prevent inappropriate responses. Meta offers an open-source security model for GenAI named Llama Guard.

"A safety model should be an integral part of your safety design," Choudhury said. "Safety models are to automate the enforcement."

Choudry and Klein-MacNeil agreed that no matter the size of the company, deploying AI will require hiring an outside consultant to help tame the complexity of the technology.

"The [AI] tooling is becoming so complex that it's better to lean on a set of companies that know how to advise you and can actually help you in utilizing that tooling for the best," Choudry said.

Benefits of cloud-based software

Hearst Communications has found many efficiencies after moving to Oracle Cloud for its ERP system and financial and human resource software. Hearst runs all its businesses on Oracle, including its information service and traditional media, including television, newspapers and magazines.

"There's a big investment to get to the cloud," said David Hovstadius, Hearst senior vice president of finance operations. "You have to do the initial investment to get to the starting gate."

Once in the cloud, most of the efficiencies come through its process automation.

"The best work is stuff you don't do and just get rid of," he said.

Hovstadius warned against trying to customize cloud software, saying it's best to adapt to the provider's forms and system.

"If you have the desire to customize things, then you tend to do it over again," he said. "I would have to say that we're almost 100% vanilla across our cloud footprint."

Getting help with cloud migration

Ferrara Candy Company hired Peloton Consulting Group to move its Oracle on-premises software for planning and budgeting to the Oracle Cloud. Earlier, Peloton helped Ferrara fix a botched on-premises deployment of the software.

"It was kind of a disaster," said Michael Gunter, financial applications director at Ferrara.

The lesson Gunter learned was that a professional services company must be as involved in the planning as in the deployment of enterprise software. The first consulting firm did whatever Ferrara asked it to do.

"There was no thought as to where we might be going in the future and working with us to figure out what was the best path," Gunter said. "They just built what we asked them to."

Ferrara plans to use Oracle's cloud-based software for profitability tracking, cost management and financial consolidation while sticking with SAP's ERP system.

Using AI vector search in Oracle Database 23ai

PayPal and Deutsche Bank use Oracle's AI Vector Search in Oracle Database 23ai. The technology searches through structured and unstructured data, with the latter including images and video.

PayPal uses vector search to recommend products to online shoppers who are customers of the financial service, said Akash Guha, director of database engineering at PayPal. The company builds customer profiles based on their online shopping activities and stores the information in the Oracle database.

When customers buy something online through a PayPal merchant, the company provides a recommendation for other items based on the person's profile. PayPal adds the recommendation to the bottom of the digital receipt people receive immediately after a purchase.

PayPal reported issuing more than 300 million purchase receipts a month.

"I specialize in scale," Guha said. "That's what we live and breathe every day."

Deutsche Bank uses AI vector search internally when an employee submits a trouble ticket to IT staff because of software or computer problems. The bank vectorizes all submissions and stores them in the Oracle database.

When an employee submits a ticket, the bank's system searches the database. It then offers recommendations based on similar problems reported in the past. The system significantly reduces the number of tickets that reach the IT department, said Marcus Praetzas, global head of database services at the bank.

"The fundamental idea is that a problem seldom happens for the first time," he said.

Praetzas' team is at the trial stage for chatbots that people would use to describe a technical problem in natural language and get possible solutions before submitting a helpdesk ticket. The chatbot would retrieve the information quickly through a backend vector search.

"The vector search itself is actually rather fast," Praetzas said.

Antone Gonsalves is an editor at large for TechTarget Editorial, reporting on industry trends critical to enterprise tech buyers. He has worked in tech journalism for 25 years and is based in San Francisco. Have a news tip? Please drop him an email.