Salesforce Introduces Low-Code Tool Einstein 1 Studio for Customizing AI Assistant Salesforce has launched a low-code tool called Einstein 1 Studio, which allows developers and administrators to customize their AI assistant, Einstein Copilot, and embed it flexibly into any customer management application or AI experience. At the annual developer conference, TrailblazerDX, Salesforce announced this new tool as part of its broader strategy to embrace artificial intelligence and achieve deep integration with enterprise data. Salesforce had previously announced Einstein 1 Studio in September this year, which enables the customization and development of secure AI applications using Einstein Copilot, with access to company data in the Salesforce Data Cloud. Clara Shih, CEO of Salesforce AI, highlighted the distinctiveness of the company's AI assistant compared to other co-pilot and consumer AI models. While consumer AI models are trained on public datasets available on the internet, enterprise models are built on trusted commercial data. The biggest challenge faced by enterprises is the abundance of data trapped in silos, such as databases, PDF files, emails, and documents, which are largely inaccessible. "That's why we developed the Einstein 1 platform," said Shih. "The Einstein 1 platform is a way for every enterprise to build future-proof AI applications based on trusted data." While Einstein Copilot AI assistant entered testing mode last week, the new studio brings three different toolsets to the platform. Developers will be able to use the Prompt Builder to define prompts using low-code natural language, the Copilot Builder to configure and customize the AI assistant, and the Model Builder to "import their own models." With the Prompt Builder, both developers and regular users can construct prompts in their own language without needing to be data scientists, programmers, or engineers. Prompts are the content users send to AI to get results, such as asking questions or requesting a summary of a series of data. Prompts can be simple or complex and sometimes need to be written in specific language to trigger specific responses. "If you can write a mail merge email, you can write a prompt," said Alice Steinglass, EVP and GM of Salesforce Platform. The advantage of the Prompt Builder is that it allows anyone, including non-technical individuals, to quickly create prompts without any training and connect them to internal data sources. Once prompts are created, users can add more data sources or iterate and adjust them as many times as needed. Once completed, these prompts can be copied and reused elsewhere. Currently in testing mode, Copilot Builder enables enterprises to customize Einstein Copilot to better meet their business needs. In addition to connecting to a wide range of data sources and being able to read, analyze, and summarize documents and emails, Copilot also offers various capabilities beyond regular conversational AI. Copilot can perform automated actions on behalf of users, such as completing work orders, sending emails, and updating database records. With the powerful capabilities of the Copilot Builder's low-code interface, users will be able to customize the actions Copilot can perform and add custom actions to accomplish tasks that Copilot currently cannot. For example, while Copilot may already be able to send an email to a customer upon request for a booking, it may not be able to send custom text messages. Creating an action to send text messages to customers would be a simple task by entering the builder and teaching Copilot how to send messages. "Artificial intelligence is not only changing how we build, but also what we build," said Steinglass. "Today, without AI, I would have to hardcode every possible action, every button, and every workflow in my code. AI unlocks new possibilities for developers; it allows us to have conversations with customers and take action at runtime." According to Steinglass, the Model Builder will allow companies to choose their own large language models, providing flexibility to connect with any desired third-party provider. This tool enables customers to select a model from Salesforce or import unmodified large language models (LLMs) from other companies such as Cohere, Databricks, Google Cloud's Vertex AI, OpenAI, etc. Developers can also fine-tune these models with their trusted enterprise data within the Salesforce Data Cloud, without leaving the secure zone. Even if enterprise customers choose to use the Model Builder to introduce external models, they will still have full trust and security assurances of the Einstein 1 platform. This includes data masking, additional data security, audit tracking, and additional privacy measures built for model security.