Two-thirds of IT leaders plan on implementing generative artificial intelligence (AI) within the next year and a half, according to new research from Salesforce. But major questions surrounding the trustworthiness of AI-generated content and the source of the data used to train AI models "are important" to consider, Sanjna Parulekar, Salesforce's senior director of product marketing, told a group of reporters ahead of the TrailblazerDX event.

"There's no doubt that the generative AI revolution is here," Parulekar said. And with the software vendor's introduction of EinsteinGPT, "these are the questions that we're aiming to answer," she said.

Similar to all new technologies, generative AI comes with its share of risks, but "a trusted approach rooted in business outcomes and ethical design" can minimize those risks, Salesforce Service Cloud CEO Clara Shih said. That's why EinsteinGPT, which combines the vendor's proprietary AI models with open ecosystem partners (including OpenAI), prioritizes trust.

"With great power comes great responsibility, and trust begins with data," Shih said. EinsteinGPT is built on Salesforce's public cloud architecture, which ensures access to personally identifiable information (PII) data is protected and compliance with data residency regulations is maintained. The vendor has also partnered "closely across our product and engineering teams and our Office for Ethical and Humane Use to ensure that we're building in ethical guardrails throughout Einstein GPT," Shih added.

"A great example of this is that we have designed into EinsteinGPT a human in the loop, which means that an employee – whether that's a seller, a marketing manager, or a customer service agent – they have final say, and you can have them make any edits on what generative AI creates as a draft before hitting send or submit," she explained.

Salesforce Data Science and Engineering SVP Jayesh Govindarajan added that "truthfulness comes from having both accurate data and also having an expert human in the loop to refine, verify, and supervise the content."

The generative AI offering is broken down into five areas of focus tied to the vendor's cloud offerings: service, sales, marketing, commerce, and IT.

EinsteinGPT for service supports service agents by automatically drafting relevant and tailored responses to lower the average handle time "and get the answer right the first time," Shih said. This offer also takes case notes for agents that can be drafted into new knowledge articles. "On average it takes 10 days for a new knowledge article to be created across industries, and we believe that generative AI and EinsteinGPT will cut this down to a fraction."

The sales-tailored offering generates natural language summaries of account news and updates, identifies contacts and leads for salespeople, and automatically generates drafts of sales emails. And EinsteinGPT for marketing generates dynamic content for "tailored landing pages, email campaigns, [and] ads based on targeted segments," Shih explained.

The vendor is also introducing a version of EinsteinGPT optimized for Slack. This offer allows employees to engage with other EinsteinGPT CRM apps within Slack's interface while collaborating with colleagues. And finally, EinsteinGPT for developers auto generates code snippets, test cases, and comments based on a proprietary language model built by Salesforce Research.

The vendor also shared today that its global investment arm is launching a generative AI fund worth $250 million to support the growing ecosystem of AI startups and responsible generative AI practices.

The new Salesforce Ventures fund is initially investing in four startups: Anthropic, Cohere.ai, Hearth, and You.com. According to the Salesforce, each of these companies is able to meaningfully transform application software and affect end-user workflows with trusted development processes.