Generative AI has been spreading like wildfire since OpenAI released its groundbreaking generative AI chatbot, ChatGPT, late last year. And permutations of the tech have been cropping up everywhere, from your search engine to your word processor and your phone, to name just a few. And it's poised to change how we do, well, everything.
A different company seems to come out every few days to announce a new generative AI integration. We could call these times generative AI's gold rush. And it's likely just the beginning.
As such, one of the latest organizations to enhance its product offerings with generative AI capabilities is the massive Salesforce, which recently unveiled its Einstein GPT chatbot, now baked into the Salesforce experience. And its impact on the business world will be difficult to understate.
In this post, we're going to provide an overview of Einstein GPT and its integration into the Salesforce platform.
Down to Business
When ChatGPT was first released, millions flocked to OpenAI's website to interact with the shiny new chatbot. The internet was then inundated with ChatGPT's poems, essays, jokes, etc. And that's all fine. But Salesforce had something different in mind.
As Jayesh Govindarajan, Salesforce's Senior Vice President of AI and Machine Learning, recently stated on the Blazing Trails podcast, "The questions companies should be asking [concerning generative AI] are: What is the business purpose? What specific jobs and tasks can generative AI help with? Funny poems and AI-powered search engines are great, but they're not going to impact the bottom line, and they're not going to transform your ways of working."
How Does Einstein Work?
Einstein GPT is the first generative AI model explicitly built for CRM. In Salesforce's implementation, the AI model is integrated into the CRM database, enabling organizations to quickly and easily generate AI-created content across every interaction, whether that’s sales, service, marketing, commerce, or IT.
Merging Einstein's data set with the Salesforce CRM allows it to integrate into all aspects of your business while guaranteeing that it will remain relevant to your company and its use cases. It also provides somewhat of a safeguard that it won't "hallucinate" on you (i.e., start meandering in nonsensical ways, as we've seen with some creative/consumer applications).
A Three-Layered Approach
Einstein's generative AI interactions are grounded in a three-layer approach, where each layer (or step) represents a pivotal phase within the interaction.
The first layer is the grounding of large language models - It's essentially a fancy way of saying, "restrict the AI to your field of interest." And that means that if you ask it to produce an email about business analytics, you want the AI to apply its compositional abilities to the "business analytics" domain. Rather straightforward.
The second layer is instruction-tuning. As the expression states, this is where the user refines Einstein's output based on further instructions and follow-up questions, going anywhere from significant rewrites to minute adjustments.
The third and final layer is reinforcement learning with human feedback. That reiterates the point made above that using AI is training AI. And that ‘human feedback’ part is critical. It enables the AI model to "understand" what sound generative output looks like vs. wrong generative output. As Govindarajan further states, "Human judgment of machine output is a very valuable reinforcing signal."
Poised for Success
Being the world's largest CRM, Salesforce believes it’s in a unique position to successfully integrate generative AI into its platform in a way that is tailored to benefiting business organizations worldwide. And it believes this for three main reasons:
- It feels that generative AI's use cases and benefits in a business setting can essentially be infinite. But because business applications are not open-domain or open-ended as with consumer applications, their scope will be limited and remain intrinsically relevant to the organization's business needs.
- Einstein mixes public data with CRM data. And because of Salesforce's size, millions of customers will be using Einstein GPT. When they use Einstein, they also train Einstein. And so, Salesforce's business-centric chatbot will continuously develop new abilities at a pace that will be hard to follow for any of its competitors.
- The last point comes down to trust. Salesforce believes that its history of building and deploying systems at scale with user trust, customer permission, and human-centric design experiences will also be a vital factor in its success.
We have no doubts about Einstein's and Salesforce's joint success. Nor do we doubt that this integration will bring about transformative business practices - and perhaps even spawn new kinds of businesses moving forward. Because, again, this is just the beginning. These initial generative AI experiences will soon be seen as a baby starting to crawl. Imagine what the tech will look like when it learns to stand up, and then walk, and even run, jump, and do summersaults!
Interesting times, you say? Indeed.
Modev believes that markets are made and thus focuses on bringing together the right ingredients to accelerate market growth. Modev has been instrumental in the growth of mobile applications, cloud, and generative AI, and is exploring new markets such as climate tech. Founded in 2008 on the simple belief that human connection is vital in the era of digital transformation, Modev makes markets by bringing together high-profile key decision-makers, partners, and influencers. Today, Modev produces market-leading events such as VOICE & AI, ESG Tech Summit and the soon to be released Developers.AI series of hands-on training events. Modev staff, better known as "Modevators," include community building and transformation experts worldwide.