A lot of the technology industry’s giants are scrambling to cash in on the meteoric rise of ChatGPT’s popularity across a diverse range of users. Given the chatbot’s surprisingly widespread adoption, big tech companies have been immersed in developing competing products. But as we saw with Google’s failed launch of Bard, not all attempts have been stellar or positively received. In a race against time, quality is often sacrificed. That’s why Salesforce’s announcement of Einstein GPT is captivating. The CRM leader is partnering with OpenAI, the research laboratory responsible for ChatGPT. By integrating its Einstein AI solution with OpenAI’s technology, Salesforce can launch an application with proven capabilities to enhance the specific functionalities of its various clouds. And for businesses that use Salesforce, there’s much to look forward to.
What Is GPT and How Does It Work?
If there’s one thing everybody’s chatting about, it’s ChatGPT. The generative artificial intelligence (AI) model capitalizes on advances in deep learning and natural language processing to create an interactive and conversational system for answering questions, analyzing data, and even generating original content or code — all based on simple inputs.
A Generative Pre-trained Transformer (GPT) is an autoregressive language model that uses deep learning to produce human-like text. Given an initial text as a prompt, it will produce content that addresses or answers the input included in the prompt. Improved algorithms, increasingly powerful computers, and an increase in digitized data have fueled a revolution in machine learning, with new techniques resulting in rapid improvements in tasks, including language manipulation.
As OpenAI states, “The dialogue format makes it possible for ChatGPT to answer followup questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests.” Still, it’s a lot to take in. Let’s try to simplify it.
The whole “generative pre-training” thing boils down to a system that predicts “tokens” based on previous tokens. A token is a basic building block or component in computer code. To put it another way, think of it as a word. You can’t write a sentence without words, just as you can’t write code without tokens. In training the model, the AI is programmed to crawl through billions of tokens contained in web pages across the Internet. It’s collecting and analyzing all of that content. A group of human engineers and researchers evaluate the results and train the AI to produce more accurate, more human sounding, and less toxic (offensive, biased, untrue, et al.) content in its output.
These software models are also trained to “learn” by exposing them to thousands or millions of examples and then synthesizing that data in a structure loosely based on the neural architecture of the brain. Of course, it’s more complicated than that, but we’re not AI scientists, so this will have to suffice. Examples of real ChatGPT interactions do a better job of illustrating its capabilities.
In one scenario, ChatGPT was asked to compose “a piano piece in the style of Mozart.” And it did, starting with an Allegro in the key of C. Having already combed through billions of related tokens, the AI had an understanding of music theory as well as Mozart, his composition style, chord progressions, and more. Another user asked ChatGPT to reply as though it had stuffed its mouth with marshmallows. It came back with, “Srr mshllws muh muh muh muh mmmph.” Essentially, the AI is accessing and processing the same information available to humans, but at incredibly accelerated speeds and immense scales.
Salesforce Einstein AI Evolves to Einstein GPT
When Salesforce incorporated AI into its offerings, it initially focused on sales intelligence and productivity. Einstein AI was really a tool for business insights, data analysis, and decision making.
- Einstein AI could analyze customer data to derive recommendations for increasing win rates.
- It could automate the prioritization of opportunities to maximize the efficiency of sales reps.
- Einstein AI’s lead scoring algorithms would analyze data, predict which leads and opportunities were most likely to convert, and display intuitive factors behind the scores.
- It could sync a user’s email and calendar to Salesforce automatically to provide a prebuilt activity dashboard or create new contact records.
- Einstein AI could also uncover historical pipeline information, business trends, and other advanced analytics to drive sales or forecast trends through machine learning.
Einstein AI was highly useful. As Sean Michael Kerner noted in VentureBeat, “AI is not a new thing for Salesforce, which has been working on its Einstein AI platform since 2016 as a tool to help improve customer relationship management (CRM), marketing and sales processes. In 2020, the company claimed that Einstein was serving more than 80 billion predictions per day across the Salesforce cloud platform.”
“With Einstein GPT,” Kerner further explained, “Salesforce is looking to take its generative AI capabilities a step further, integrating initially with OpenAI in a bid to help users automatically generate content, respond to emails, create marketing messages and develop knowledge base articles to help improve customer experience.”
How Businesses Will Benefit from Einstein GPT
Einstein GPT is designed to help companies better connect with their customers through a unified platform that can automate and tailor recommendations with the help of the generative AI tool. The combination of real-time Salesforce data and the GPT will allow businesses to enhance the personalization of shopping experiences and deliver more meaningful recommendations to their customers. It can also personalize product listings or engagement for individual customers without product descriptions being preloaded on the back end.
Citing a briefing with Jayesh Govindarajan, senior vice president for AI/ML at Salesforce, Kerner illustrated how Einstein GPT will work using a combination of customer data and generative AI models: “Govindarajan said that Einstein GPT is a combination of natural language processing (NLP) components for understanding what the user or organization wants to achieve and then helping them to execute those tasks. He noted that from a technical perspective OpenAI’s GPT model is a large language model (LLM), and the goal for Salesforce is to have layers on top of it to fine-tune the model.”
Quickly Generate Impactful Content
According to Salesforce’s release, Einstein GPT will infuse Salesforce’s proprietary AI models with generative AI technology from an ecosystem of partners and real-time data from the Salesforce Data Cloud, which “ingests, harmonizes, and unifies all of a company’s customer data.”
With Einstein GPT, customers can connect that data to OpenAI’s models (out of the box) or choose their own external model and use natural-language prompts directly within their Salesforce CRM to generate content that adapts to customer needs in real time. Here are some examples of how businesses can leverage Einstein GPT.
- Automatically generate personalized emails for salespeople to send to prospects or customers.
- Generate specific responses to customer queries.
- Create marketing content targeted at increasing campaign response rates.
- Einstein GPT can even produce code for developers.
Salesforce also demonstrated some of the AI’s capabilities in a video.
- A user asked to see their top contacts. Einstein GPT instantly displayed a list of names, titles, and options to create a contact or generate an introductory email.
- When choosing to write an email, Einstein GPT immediately composed this: “Dear Mr. Smith, could we arrange a meeting to discuss your growth plans.” A customized signature for that salesperson was also included.
- However, the salesperson felt the message was a bit stuffy and asked Einstein GPT to “make it less formal.” The AI complied and wrote, “Hi John, I’d love to connect on your growth plans. Next week?”
In the demonstration, Einstein GPT was also able to create a lead gen page on its own and add a signup form — with no work required of the user except to ask the AI to perform the task. It can also create reports, deliver insights, handle tickets, and much more, with no more effort than typing in a request.
Einstein GPT Will Support Multiple Clouds
As of the March 7 announcement from Salesforce, Einstein GPT will launch for Sales, Service, Marketing, Developers, and Slack Customer 360 apps.
- Einstein GPT for Service will automatically draft relevant responses to customer inquiries, assist agents with taking case notes, and turn them into knowledge base articles. It will also auto-generate personalized chat replies to increase customer satisfaction and engagement.
- Einstein GPT for Sales generates natural language summaries on account updates and helps identify key contacts for sales outreach. And as we saw in the demo, the AI can compose email drafts for salespeople to review, alter, and send. In terms of productivity, the AI can automate other sales tasks such as scheduling meetings and preparing for the next interactions.
- Einstein GPT for Marketing dynamically generates personalized content to engage customers and prospects across email, mobile, web, and advertising. This includes automated content creation for landing pages, email campaigns, and ads based on targeted segments.
- Einstein GPT for Developers taps into Salesforce’s large language model (LLM) to automatically generate code snippets for Salesforce applications to improve developer productivity by using a chat assistant to produce code and ask questions for languages like Apex.
- Einstein GPT for Slack Customer 360 apps will deliver AI-powered customer insights in Slack. Examples include smart summaries of sales opportunities and end user actions such as updating knowledge articles.
A GPT Solution for Every Salesforce Business, Not Just Giant Enterprises
Naturally, when advances like Einstein GPT hit the market, the flurry of excitement sometimes overshadows more practical concerns. Smaller and mid-sized organizations may assume that these groundbreaking developments and innovative tools will be out of their reach. “I’m not Mercedes-Benz, so I probably won’t be able to incorporate this,” is sort of how the thinking goes. But Salesforce executives have said that Einstein GPT will “open the door for companies — from small business to the very largest enterprises, across every industry and across every region in the world — to completely reimagine how they engage with their customers.”
And Salesforce has shown genuine commitment to its customer experience by partnering with OpenAI. Instead of trying to design a competing model from scratch with potentially lackluster or disastrous results, Salesforce has aligned with a proven, tested product to ensure that its customers achieve their goals.
Oloop is a Salesforce Ridge Partner and proud member of the Salesforce Talent Alliance. Our business was predicated over a decade ago on supporting Salesforce customers, and this niche has differentiated our organization from other workforce solutions companies that lack direct experience in a wide spectrum of Salesforce clouds. We’ve tackled complex enterprise and federal engagements, developing custom configurations and repeatable processes. Despite our small business size, our stature in the Salesforce ecosystem is large. We have specialized in consulting and implementation, continuous integration, custom development, Lean Agile solutions and program management, and creating Salesforce omnichannel customer experiences for shared services.