According to a recent study, 84% of senior IT leaders believe generative AI will help them serve their customers better.
But, with so many gen AI services available, it can be challenging to decide how to best use this new technology effectively.
Say hello to Einstein GPT.
Combining your Salesforce CRM, data, and generative AI improves users’ productivity and impacts the bottom line.
Salesforce has the #1 CRM in the world, containing a robust customer dataset. This allows Einstein GPT to produce highly effective AI generated content at scale within your existing workflows.
Read on for a run-down of how Einstein GPT can transform your business processes.
Einstein GPT is the world’s first generative AI for CRM. It allows you to use Chat GPT-like functionality natively on the Salesforce platform.
Now, for those who don’t know what “GPT” means. It stands for generative pre-trained transformer. It is a type of Large Language Model (LLM).
Salesforce has 3 main focuses for the Einstein GPT product.
The first is the generative aspect of the technology. This means that the technology can power more than automated conversations. (More on that later).
Naturally, trust and security are also a focus. One way that Salesforce does this is by blending public and private data to make the product more trustworthy and valuable for customers.
This means you can use Salesforce’s private AI models (Einstein AI layer) AND the latest LLM models (GPT-4, Watsonx, etc).
As more and more customers use Einstein GPT, the Salesforce-specific model will become fine-tuned for Salesforce use cases. This is a product differentiator, and one reason why using the Einstein AI layer will be beneficial to use over other LLM models.
Personalization is the last focus of Einstein GPT. Leveraging real time data in and outside of Salesforce, Einstein GPT can dynamically generate personalized content based on your business needs.
Anyways, what is the actual business purpose of this new product?
Einstein GPT has the following business purpose:
- Quickly and easily generate AI content in different verticals
- Saves time and improves personalization at scale
- Reduces friction from idea to first draft
- Helps scale SMBs without a huge workforce
- Helps users be more efficient, productive, and effective
- Helps developers lean on the job
- And many more …
Now you know what Einstein GPT is, let’s move on to the nitty-gritty details.
Extremely sensitive customer information is stored in your Salesforce environment. This data must also be leveraged when using Einstein GPT so the content you generate is relevant and personalized.
As a result, you might wonder how secure this new product is. Will your personal information be revealed?
The answer is no.
Salesforce Einstein GPT puts security first using the Einstein Trust Layer. The Trust layer allows you to benefit from Gen AI without worrying about your sensitive data being exposed.
This is a HUGE deal.
But how does it work?
The Einstein Trust Layer pre and post-processes your data to ensure security.
Once you execute a prompt, Einstein securely retrieves your company’s customer data and merges this data into your prompt. The LLM is then grounded to ensure relevant use cases, and information is used to generate an accurate response.
Next, the sensitive data used in the prompt is masked to protect your private information. This is done by replacing the sensitive data with anonymized data.
Before the prompt is sent to the LLM, prompt defense is used to avoid harmful outputs and mitigate hacking.
Your prompt is then sent to the Salesforce LLM, or you can bring your own model (check out IBM Watsonx).
Zero retention is used so the LLMs do not retain customer data. This means your prompts and outputs will not be stored by the LLMs.
The response from the LLM is then moved through a toxicity detection process. Toxicity detection will flag harmful, inappropriate, violent, and hateful outputs so they are never used.
Finally, the data is demasked so your output has relevant and personalized content. An audit trail is also provided for monitoring and compliance needs.
That was a lot of informational text. I’m personally more of a visual person, and you might be too.
So, here’s a diagram of the process just below you can check out 🙂
Einstein GPT is all really cool and all. But how do you actually create prompts?
One of the main ways to do this is Einstein Prompt Studio.
Prompt Studio is like flow builder, but for creating prompts, not flows.
Prompts are used to create personalized content for your users. For example, a cold email to a prospect.
When building a prompt, you specify the following information:
- Text (the prompt itself)
- Merge fields
- Length of the output
- Style of writing
- AI provider and model.
Once created, prompts can be re-used throughout your Salesforce environment. This is super useful because it saves you from typing out prompts whenever you want to generate content.
Salesforce is releasing 6 new products from the Einstein layer of AI cloud that can be integrated into the existing Salesforce clouds.
In the coming sections, I will review those products and illustrate what specific jobs and tasks generative AI can help with.
Sales GPT enables sales teams to generate personalized and engaging sales-related content.
The main features of Sales GPT are:
- Sales emails
- Quickly create and tweak outreach emails to prospects using account data in your CRM
- Call summaries
- Summarize Sales calls to save time and improve follow-through
- Sales assistant
- Prepare for meetings and close more deals
Imagine you are a sales rep and a new account is assigned to you. Sales GPT can give you an account overview and recent news of your new customer.
After this, you can use Sales GPT to generate actions based on the account overview. One option might be viewing contacts to reach out to.
Once Sales GPT has given you the contact information, you can generate an outreach email.
Lastly, use Sales GPT to tweak the tone of the AI-generated email and hit send.
Service GPT helps service agents be more productive improving customer experience to boost customer satisfaction.
The main features of Service GPT are:
- Service replies
- Generate personalized replies based on customer data
- Work summaries
- Wrap up summaries of service cases and customer engagement
- Knowledge articles
- Create and update knowledge articles based on data from support interactions and customer feedback
- Mobile briefings
- Summarize critical information for field service teams
Let’s say you are a customer relations rep at a car business. A customer reaches out via support chat about an issue with their vehicle. Service GPT can answer customers’ questions based on historical data.
Once the interaction is closed, Service GPT can then summarize the interaction. If there is no knowledge article for this specific customer inquiry, Service GPT can generate one in the snap of a finger.
Marketing GPT gives your marketing team an AI interface for campaign creation and audience segmentation.
The core features of Marketing GPT are:
- Email content creation
- Generate email body content and subject lines to use your email marketing campaigns
- Segment intelligence
- Helps marketing understand campaign performance based on segments
- Create landing pages
- Create landing pages, generate copy and images for the landing pages
Einstein GPT combined with Salesforce Data Cloud is a powerhouse marketing combination.
How about an example?
Put yourself in the shoes of a marketing specialist at Larry’s Landscaping. Larry’s Landscaping is offering a new service: Modern Rock Landscaping.
You must create a DRIP email campaign to nurture leads and a new landing page to capture leads.
To create the landing page, Marketing GPT can be used to generate the experience page, copy, and images automatically.
Next, Marketing GPT can be used to create email content with a few clicks. All you need to do is set up the campaign sequences.
And just like that, in 1 day, you have set up the lead capture form and DRIP campaign for the new Landscaping Service.
Commerce GPT is your tool for creating personalized commerce experiences for your customers and better business outcomes.
Features included with Commerce GPT are:
- Goals based commerce
- Generate goals, targets, and AI powered customer insights
- Dynamic product descriptions
- Generate compelling product descriptions
- Fill in missing product catalog data
- Commerce concierge
- Use AI-chat bots and generative technologies
- Helps you create a 1-1 shopping experience
Slack GPT is a Slack App that enhances your internal communication productivity.
The main features are:
- AI-generated channel summaries
- Huddle summaries
- AI-assisted writing
If you are a Slack customer, these features will be super helpful for traceability and getting up to speed on missed items. The Huddle summaries are perfect for focusing on the conversation without taking notes on the side.
Tableau GPT marries Gen AI and the Tableau UX. This results in automated analytics and personalized metrics. Tableau GPT will interpret your data and provide summaries based on the source data. The interpretations will be justified by visually indicating the data used to make the interpretations.
APEX GPT allows you to generate code with generative AI. When creating LWCs or APEX classes all you need to do is type in natural language prompts directly in you code file for what you want the class to do, and APEX GPT does the rest.
As a result, development speed and efficiency are improved, reducing implementation costs. Additionally, APEX GPT will scan for code vulnerabilities and make inline suggestions. This improves your code’s quality and security.
One really cool thing about APEX GPT is that it interprets your code base. This way, APEX GPT can replicate your organization’s coding structure and practices.
Developer productivity has entered the chat. APEX GPT can also create test classes to ensure your code works as expected.
The assumption here is APEX GPT will generate the majority of code, however, you will need to add additional tweaks. This is especially true for test classes.
Similar to APEX GPT, Flow GPT allows you to create Flows using GPT functionality. Users type in a description of what they want to the Flow to do, and Flow GPT will use natural language processing to create it.
Flow GPT helps users automate their workflows.
Funny isn’t it?
You automate the creation of workflows, which then automate manual tasks?
Salesforce sees this as a great way to improve productivity, but also an opportunity for beginners to learn how to create Flows manually.
So far, Sales GPT and Service GPT have been announced, costing $50 per monthly user.
Salesforce customers with Unlimited Edition Licenses get Einstein GPT credits. These credits can be used to generate outputs. The number of credits needed to generate an output will be based on the scope of information the AI model needs to analyze when generating outputs.
This year, Salesforce will release 35 badges related to Einstein GPT. Currently, there are only a few of these Einstein GPT badges available.
There is also the Einstein GPT Quest on Trailhead, where you can read blog posts, watch videos, and do trailhead modules. This trail aims to catch you up to speed on all the Einstein GPT basics. When completing the quest, you get a special badge.
There are currently no certifications specific to Einstein GPT. However, Salesforce does have the AI Associate Certification, which tests your knowledge of AI foundations.
Interested in learning more about Einstein GPT and generative AI?
Here are some helpful and interesting resources you can check out: