In a recent press release, Salesforce announced additions to their Einstein platform that bring Artficial Intelligence (AI) -based solutions to Salesforce developers and admins using a low-code, "point and click" configuration-based experience. The recent additions to the platform include Einstein Translation that allows for the language translation of a Salesforce object or field and Einstein Optical Character Recognition (OCR) which uses computer vision technology to detect and extract objects from documents. In addition, Salesforce has added mechanisms to prevent its AI capabilities from embedding bias within its machine learning models.
Salesforce customers generate more than 6 billion predictions everyday using Einstein AI capabilities. However, using these AI capabilities does not require extensive data science expertise. Salesforce wants to enable business users to be more productive by leveraging intelligent insights generated by the platform:
Companies must continue to accelerate and expand their use of AI to remain competitive, but each has unique processes, data and use cases that require customized AI. With a global shortage of data scientists necessary to build these tailored AI solutions, Salesforce has been bridging the skills gap by empowering Salesforce developers and admins with "point and click" solutions and low-code services in the Einstein platform.
Einstein Translation is one of the new capabilities added to the Einstein platform and allows developers or admins to setup automatic translation of any Salesforce object of field. A use case that Salesforce has identified for this technology is translating inbound customer enquiries to the native language of the customer support agent who is available to service the request.
The other new capability introduced is Einstein Optical Character Recognition which is able to analyze documents and extract contextual information of the object using computer vision. This technology can be used to enable a sales rep to scan a prospect’s business card, extract the text from the business card image, and then load it into Salesforce.
Embedding AI into Salesforce applications is nothing new for the company. Last year, Salesforce introduced Einstein Prediction Builder, which will predict the outcome of any Salesforce field or object and predict a future outcome, such as what is the likelihood of this customer paying their next invoice.
As AI becomes more prevalent, so have concerns about its use in society. Recently, the European Union published Ethics guidelines for trustworthy AI which includes the following:
- Lawful – AI should respect applicable laws and regulations
- Ethical - Respecting ethical principles and values
- Robust – Both from a technical perspective while taking into account its social environment
Salesforce has also made a commitment to democratized and trusted AI and recently issued the following statement:
The implications of AI are unlike those posed by other technologies because of the pace of disruption it will bring and the levels of uncertainty around its potential impact. Salesforce recognizes that it is not enough to just make AI accessible; every developer, admin and end user of AI must also be empowered to use AI in a transparent, responsible and accountable way. As AI technology continues to change and evolve, Salesforce commits to upholding its number one value of trust and delivering the tools and understanding necessary to identify and mitigate harmful AI bias.
As a result of this commitment, Salesforce also announced features that provide explanations about “why” a prediction is being made. Salesforce wants to avoid bias influencing predictions and allows administrators to exclude race or gender biases through alerts when a harmful bias has been detected within a dataset.
John Ball, an evp and general manager at Salesforce, explains why it is important to develop AI capabilities that can be trusted:
Our mission goes beyond just making the technology more accessible; we are committed to democratizing AI that people can trust and use appropriately.
For more information on how Salesforce addresses AI bias, please refer to their blog post.