[ad_1]
Salesforce President Sarah Franklin speaking last week in Chicago about the company’s new generative … [+]
Can you replace a website with a ChatGPT prompt?
Imagine coming to a favorite fashion brand’s website and, instead of logos, graphics, and welcome text, you see yourself in a selfie you never took, wearing the pants you bought last year but paired with a sweater you don’t yet own from the brand’s brand new spring collection. That’s one of the innovations that Salesforce promised companies last week as it launched AI everywhere in an “AI Cloud” for customers that adopt its entire AI-infused software stack for commerce, marketing, analytics, communications, collaboration and more.
“Imagine if I go to a website for a brand and I’m not looking at a bunch of content and copy,” says Salesforce president Sarah Franklin. “I’m looking at a prompt. And this prompt is here with an intelligent conversation that’s grounded in AI and caught coming from my CRM.”
An AI shopping assistant is only a small part of what the company is offering.
All of this is part of a massive vision Salesforce is unveiling that deeply integrates generative AI into almost everything a company does, from sales to service to internal collaboration. For example, in Commerce GPT, Salesforce auto-generates product descriptions that marketers can choose from, edit, and pushes live. It translates texts to dozens of different languages, provides analytics in reaction to natural-language queries in Slack conversations, writes marketing emails in response to prompts, adds store locators to emails, and personalizes messages automagically. If a brand needs something more customized, Einstein GPT can generate custom code in seconds that developers can tailor as they wish. And when customers request help, Einstein helps agents respond by suggesting servicing and usage options that are location-specific and context-aware, including factors such as local weather conditions, public events and locations in addition to customer history and knowledge of a company’s products and services.
Oh, and—of course—it suggests additional purchases based on previous orders.
The AI-first enterprise might be impressive, but it doesn’t come cheap.
Salesforce’s AI Cloud Starter Pack starts at $360,000 annually, and includes “Data Cloud, MuleSoft automation, Tableau Analytics, Slack, CRM and an AI readiness assessment from Salesforce services,” according to Constellation Research’s Larry Dignan.
It’s a bold and comprehensive vision, but there is competition from other major software vendors such as IBM.
“We go after areas where we see quick gains in productivity and time to value,” says IBM SVP and chief commercial officer Rob Thomas. “We see big potential with foundation models—which are reusable and require minimal training—in helping leaders implement AI to radically change how businesses operate.”
The Salesforce vision is one of the opening salvos in an AI arms race that no company can ignore.
While it’s true that all software demos look great initially, there’s always a lag time between new innovations and actual boots-on-the-ground business value, and some glorious promises fail to materialize, it’s also clear that large language model AI systems have demonstrated incredibly impressive results even in early days. What Salesforce demonstrated will no doubt have some birthing pains, but will clearly also significantly speed up many key business processes.
Those who don’t adopt technologies like these risk being left behind.
“Companies that embrace AI-driven opportunities and learn to wield it effectively will outpace their less adaptive competition—competitors that will become extinct,” says Joseph Ours, an AI expert at Centric Consulting.
There are also risks, however.
One in four executive respondents to a recent survey by data analytics and AI vendor Altair said more than half their AI projects fail, and 42% say at least one of them has failed. There’s a lot of change involved in workflows and processes, none of which is without risk.
That said, AI is often best applied in scenarios where there is scarcity or limited reach. An example is the AI-generated minor league baseball stories that the Associated Press started employing in 2016: there was insufficient demand for these stories to justify a human writer’s time.
In much the same way, there’s a current scarcity of great customer service. (Think back to the last two-hour call you had with your mobile phone carrier or TV service.)
What I saw from Salesforce’s new generative AI service and support products look to help service personnel vastly scale both the number of people they can help simultaneously and the quality of the help they can provide. This remains to be proved out in practice in various applications and verticals, but if accurate, it’s literally a game changer.
“End-to-end AI-integrated workflows exponentially improve response times, flexibility, reduce risks, promote faster experimentation, and reduce time to market,” says Balakrishna D., executive VP and head of AI and automation at Infosys.
Not everyone is convinced that AI is ready for prime time. A few executives and founders I talked to mentioned that generative AI in particular still suffers from hallucinations, with one mentioning the fictitious legal precedents that one unfortunate lawyer acquired from GPT-4, used in court, and is being censured for. Others suggested that over-reliance on AI systems risks losing human expertise, or risks allowing AI-driven bias to infiltrate corporate decision or communications. Additional risk factors often cited for companies using generative AI is legal exposure when the models have been trained on copyrighted data.
“While AI is on the road to becoming vastly more intelligent than humans, it’s also an unintended yet frequent liar,” says Mark Weinstein, founder of alternative social media network MeWe.
All of those are real issues.
Salesforce isn’t unaware of the challenges, and emphasized repeatedly that its large language models are all trained on 100% ethically and legally sourced data as well as a companies’ own data. The word “trust” shows up 28 times in a Salesforce transcript from an event announcing many of the new products and features last week in Chicago, and company president Franklin made a point of emphasizing trust and equity, calling the companies’ Einstein GPT “the world’s most trusted generative AI for the enterprise.”
“In the world of AI, it’s never been more important that the first and foremost value is trust,” she said. “Trust in our products, our people, trust that we will always do the right thing together.”
That extends to not co-mingling customers’ data for training purposes and employing solutions such as data masking and zero retention to ensure that enterprises don’t lose control of their own information, or feed a competitor’s success.
Another key aspect: not eliminating humans.
“Most importantly, there’s a human in the loop to give feedback, help the AI continue to learn,” Franklin said. “And that is really so important in this world where trust is paramount.”
Two phrases repeated generously throughout the company’s presentation were “help you, not replace you,” and “edit and take credit,” highlighting that Salesforce sees generative AI as a tool to extend human talent, not replace it.
That rings a bell with many.
“We should not give AI the complete final say on all matters,” says David Brossard, the chief technology officer of automation company Axiomatics. “It can help employees execute, test, and verify existing tasks faster. Ultimately though, the responsibility still lies with a human to make sure the output of AI is still correct and relevant to the company’s needs.”
[ad_2]
Source link