SAN FRANCISCO — Business software makers are rapidly integrating generative AI into their products, triggering numerous questions among customers enticed by the potential for significant productivity gains.
Last week, Salesforce answered the queries of roughly 5,000 developers attending its TrailblazerDX conference to tackle the company’s latest GenAI technology, including the new Einstein 1 Studio. The software suite comprises a trio of no-code tools for using and training AI models and developing a simple path for accessing their services.
While eager to embed AI in Salesforce software, customers acknowledged that it requires planning, time, and work to make it useful in the company’s sales, marketing, and customer service applications.
A challenge for Carnegie Learning is ensuring that Salesforce software can access data in Oracle’s cloud-based ERP system. Carnegie, a provider of K-12 educational services to school districts, has spent more than two years mapping Oracle data to the Salesforce format.
About a year ago, Carnegie licensed the Snowflake data warehouse as a go-between to ensure Salesforce applications can get information from Oracle customer data, such as product usage, order history, and returns.
Carnegie wants to add Salesforce’s Data Cloud to the mix and take advantage of the platform’s capabilities for managing metadata across Salesforce applications and other sources. Data Cloud is critical for getting the most out of Einstein 1, but harmonizing a multi-vendor environment isn’t easy.
“There’s a lot of decisions that are being made on our operational side that we’re still working toward,” said Marissa Scalercio, vice president of sales operations at Carnegie. “There’s going to be some path forward with probably both systems [Data Cloud and Snowflake].”
AI challenges
Carnegie’s difficulties in preparing its environment for AI are not unusual. In a survey of 339 companies published in December, TechTarget’s Enterprise Strategy Group found that the top five challenges were finding quality data for use in models, the high cost of implementation, determining the return on investment, ensuring data privacy and security, and making AI available across an organization.
Independent software vendors embedding products in Salesforce software are also working hard to understand the limitations, strengths, and business value of Salesforce’s AI tools.
“This is exciting stuff,” said an ISV developer who asked to remain anonymous because he wasn’t authorized to speak for his company. “Customers are asking about generative AI, so we’re trying to see how we can deliver it.”
BACA Systems, a midsize manufacturer of stone-cutting and polishing equipment, avoided Carnegie-like complexity by gradually making Salesforce the center of its business operations over the last 10 years. Today, BACA uses Salesforce sales and customer support software, the Rootstock ERP system built on the Salesforce Cloud Platform, and Salesforce partner Natterbox for communications.
BACA has tested tools within Einstein Studio for roughly half a year. Initially, the capabilities were basic, but in October, Salesforce granted the company access to Einstein’s Prompt Builder. The tool enabled BACA to train Salesforce’s version of the OpenAI GPT-4 model to create custom emails by pushing a button on an account page and choosing from a list of subjects.
“They felt almost scarily too good,” said Andrew Russo, BACA’s Salesforce architect, of the emails. “If you were the reader, you would have no clue that it was not written by a person.”
BACA stores personal information on its customers, such as whether they are married or have children; their occupation; recreational activities they enjoy, like golf or tennis; and their aspirations, such as buying a house. The information that the company has gathered for several years lets Salesforce AI bring a personal tone to emails.
“It [the data] helps you connect at a deeper level,” Russo said.
Carnegie also tested Prompt Builder for building an email-creation button on contact and account pages. The emails recommend products or services based on buyer history and product usage. The feature “saves an immense amount of time,” Scalercio said.
Approaching AI cautiously
Unlike BACA, Carnegie isn’t ready to become a Salesforce-only company. Carnegie uses HubSpot for marketing, and some business leaders within the company prefer to keep it that way.
“It’s a conversation we have every time we try to grab another tool from Salesforce,” said Scalercio, who said she favors using as much of Salesforce as possible.
Another concern is the proliferation of AI tools across all of Carnegie’s business software. The company uses Google’s Workspace productivity suite, so it tested the software’s Duet AI technology.
“We’re hoping to see some really big ROI gains from using [Salesforce AI],” Scalercio said. “That being said, we’re moving slowly, purposefully, so we don’t overwhelm our sales reps.”
Overall, the integration of generative AI into business software is presenting both challenges and opportunities for companies looking to leverage the technology for increased productivity and efficiency. As more organizations navigate the complexities of AI implementation, the industry is poised for further innovation and growth.