Salesforce Aims to Plug ‘AI Trust Gap’ With New Tech Tools
Salesforce says its new technology can erase user prompts—or what employees input into AI models—and keeps the models from retaining customer data. Photo: Cam Pollack/The Wall Street Journal By Belle Lin and Angus Loten June 13, 2023 5:23 pm ET Salesforce says its new technology will help client companies use generative AI tools such as ChatGPT safely by protecting proprietary data, among other things. For Salesforce, convincing information-technology leaders that generative artificial intelligence is safe to use within their digital corporate walls is crucial as it seeks to sustain a turnaround plan that began this year. Its most recent earnings and revenue topped Wall Street expectations, with Chief Executive Marc
Salesforce says its new technology can erase user prompts—or what employees input into AI models—and keeps the models from retaining customer data.
Photo: Cam Pollack/The Wall Street Journal
Salesforce says its new technology will help client companies use generative AI tools such as ChatGPT safely by protecting proprietary data, among other things.
For Salesforce, convincing information-technology leaders that generative artificial intelligence is safe to use within their digital corporate walls is crucial as it seeks to sustain a turnaround plan that began this year. Its most recent earnings and revenue topped Wall Street expectations, with Chief Executive Marc Benioff highlighting the company’s cost-cutting efforts as it copes with a recent sales slowdown and pressure from activist investors.
Newsletter Sign-Up
WSJ | CIO Journal
The Morning Download delivers daily insights and news on business technology from the CIO Journal team.
Subscribe NowThe software company announced the new platform Monday. The move comes as chief information officers and other business leaders are increasingly questioning how they can use generative AI without their proprietary data winding up as training data for widely available large language models. Companies including JPMorgan Chase and Verizon have banned the publicly available version of ChatGPT for employees. Verizon said that it worried about losing ownership of customer information or source code that its employees typed into ChatGPT.
At an event held in New York City on Monday, Salesforce executives addressed generative AI’s risks for companies—including the difficulty of controlling who gets access to corporate data—while introducing its new “Einstein GPT Trust Layer,” a tool the company said can help resolve such concerns.
Benioff said during the event that every CEO he has spoken to has brought up the “AI trust gap” between their “desire to rapidly move forward” with the technology, and the problems that large language models introduce in corporate environments.
“The most important thing we’re doing with AI is trusted and ethical AI,” Benioff said. “From our perspective it’s all about trust.”
Salesforce CEO Marc Benioff at Monday’s event announcing the new platform.
Photo: Angus Loten / The Wall Street Journal
Salesforce said the new tool can erase user prompts—or what employees input into AI models—and prevents the models from retaining customer data like personally identifiable information. The tool can be applied to a company’s own large language models or those from providers such as Amazon.com and Anthropic. The new technology is part of Salesforce’s AI Cloud suite, which embeds generative AI into its flagship customer-relationship-management software, Salesforce-owned messaging app Slack, and other data visualization, marketing and customer-service tools.
“Trust is an underrated but hyper important topic,” said Daniel Newman, a founding partner and chief analyst of IT research and consulting firm Futurum Research. “Salesforce realizes generative AI apps themselves aren’t as differentiated as having users that trust the data and tech.”
Still, some tech executives remain cautious. Scott duFour, global CIO of FleetCor Technologies,
Other big tech companies have also introduced tools to help clients use generative AI. Oracle on Tuesday said it plans to release a generative AI service with AI startup Cohere that the database company said will include data-ownership assurances and tools for data provenance. Microsoft, which embedded its Copilot generative AI into its business software earlier this year, has said its models preserve a company’s security and privacy controls.
Among major technology vendors, the hope is that generative AI will revive IT spending, which has taken a hit as companies have cut technology budgets amid economic uncertainty. CIOs have also been under pressure from chief financial officers and boards to streamline spending.
Revenues of global generative AI-based technologies are expected to reach as much as $3.7 billion this year, according to estimates from S&P Global Market Intelligence, growing to $26 billion by 2028.
At Monday’s event, Salesforce executives addressed generative AI’s risks for companies.
Photo: Belle Lin / The Wall Street Journal
For CIOs, Salesforce’s trust tool can be thought of as a “central control plane” to manage multiple large language models, or LLMs, said Kyle Davis, a vice president analyst at IT research and consulting firm
Chetna Mahajan, chief digital and information officer at app analytics company Amplitude, said that before using any LLM she would need to know how the vendor and its partners handle customer data.
IT leaders need to explore generative AI now or risk being left behind, said William Parks, senior vice president and chief information officer at Infinera,
Infinera is working with Microsoft’s generative AI technology, but has paused the internal use of ChatGPT until an updated usage policy is in place, Parks said. His concerns include having the company’s data inadvertently shift to the cloud as well as working with incorrect data or “hallucinations” generated by generative AI tools, he said.
Compared with startups and smaller vendors that offer solutions to help mitigate some corporate data risks associated with generative AI, established enterprise-technology companies have an advantage in gaining customer buy-in, especially for emerging technologies, analysts said. Salesforce’s reputation, plus access to large swaths of critical customer data, is likely to help it get customers on board, Newman said.
Salesforce said its AI Cloud suite will cost $360,000 a year with an annual contract. Its trust tool will be available to use this month, while other generative AI-embedded tools will be available this year or next year.
Write to Belle Lin at [email protected] and Angus Loten at [email protected]
What's Your Reaction?
