Salesforce introduces Einstein GPT. As a result, the software giant’s existing AI offering is evolving toward a model built on top of generative AI. OpenAI is an unsurprising launch partner.

Microsoft announced today that it is integrating generative AI into both Power and Dynamics 365. On Trailblazer DX, Salesforce announces a similar goal, although integration of GPT-based AI into the company’s products appears to be well advanced. After all, during the announcement of Einstein GPT we see a number of concrete demos that demonstrate the added value of the new functionality. Whether Einstein GPT is actually the world’s first generative AI for CRM, as Salesforce claims, we find it a bit more difficult to assert given Microsoft’s announcement.

Einstein GPT is an evolution of Einstein. If you’re somewhat familiar with Salesforce, you know that Einstein isn’t a product in itself. Salesforce used Einstein as a collective term for various AI skills. This AI functionality then takes the form of very specific AI-driven functions in the Salesforce products. Einstein GPT is no different.

Einstein for the entire Salesforce cloud

Salesforce rolls out Einstein GPT for its large clouds, for Slack, and for developers. Specifically, the announcement includes the following introductions:

• Einstein GPT for service: AI for the service cloud, enabling customer services to use AI to quickly suggest automated responses to customer questions.
• Einstein GPT for sales: Here, AI can create helpful summaries about prospects or generate email suggestions that sales reps can send out.
• Einstein GPT for Marketing: Einstein helps here to automatically generate content, for example to fill in personalized landing pages.
• Einstein GPT for Slack: This integration allows employees to work with Einstein GPT applications from the communication platform.
• Einstein GPT for developers: Based on a natural language question and code already written in the company, AI can generate suggestions for new code here.

“AI should make people more productive in their jobs,” says Jayesh Govinddarajan, SVP of Engineering for Einstein and Bots at Salesforce. “All generated content is contextual, personalized and based on data in the Salesforce environment. With our implementation, there is always a human in the loop, which makes the use of AI reliable.”

Einstein GPT leverages Genie – a data cloud foundation that Salesforce introduced last year. Genie merges a company’s data into one only source of truth. As part of Einstein GPT, the AI ​​algorithm accesses this data in a secure manner. This allows the system to generate relevant content without disclosing sensitive internal data to the outside world. The Salesforce Hyperscale architecture supports these cloud-based capabilities. The algorithm can also incorporate external data into an answer to a question.

AI Customer Service

During the demos, we see an illustrative example where customer service for a fictional apparel company receives a question about the temperature class of a jacket via the Service Cloud. The AI ​​sees what the question is about and can retrieve the correct data for the jacket in question from the company’s internal product data. The customer service representative thus receives a ready-made answer that he can check or adapt, but otherwise send immediately.

In a more complex variant, a customer asks whether the jacket in question is good enough for a planned ski trip in a certain area. In this context, Einstein GPT can include external temperature data in the problem and prepare an answer again. If changes are necessary, the person behind the computer can implement them immediately. The AI ​​will also learn from the adjustments.

A second demo shows a scenario where a programmer asks a technical question in human language. Einstein GPT understands this question and immediately generates a few lines of code that apply to the business environment and are therefore ready for implementation.

Safe and trustworthy

For Einstein GPT, Salesforce combines its own data and AI capabilities with a generative model. The CRM specialist works with OpenAI for this. Over time, Salesforce plans to make other GPT models available so the customer can choose which model they prefer.

During the presentation, Salesforce emphasizes the importance of proper data, but also the security of that data and the reliability of AI. Accuracy and security are said to be high on the priority list. One of the aims of the mandatory human intervention is to ensure that ultimate responsibility rests with the human expert, who currently mainly uses the input of the AI ​​to facilitate their own work.

Revolution with advantages and disadvantages

If you look at it a little more pessimistically, you can also see how certain profiles improve the AI ​​today and therefore continue to train, so that their input is worth a lot less tomorrow. “The generative AI revolution is here,” said Sabjina Parulekar, Senior Director of Product Marketing, during the presentation. “AI offers every company the opportunity to optimize productivity in all departments.” Optimization does not usually mean hiring more people, and Salesforce knows that.

Nvidia CEO Jensen Huang called the launch of ChatGPT the “iPhone moment for AI”. Almost four months ago, OpenAI released ChatGPT to the world. Microsoft and Salesforce are already showing concrete integrations that can have a direct and very concrete impact on the workflow of companies worldwide. Huang seems right.