AI news

Salesforce today became the latest provider of an application platform to make available a co-pilot that leverages generative artificial intelligence (AI) to automate a wide range of tasks.

A public beta of Einstein Copilot enables users to invoke a natural language interface to invoke multiple large language models (LLMs) capable of, for example, creating a sales plan or running data analytics.

Einstein Copilot also makes use of an Einstein Trust Layer that Salesforce has created to minimize hallucinations by veracity of the data exposed to an LLM as part of an overall effort to ensure security and compliance by, for example, automatically masking sensitive data.

The goal is to make good on a promise to deliver generative AI capabilities using data residing in the Salesforce Data Cloud, says Jayesh Govindarajan, senior vice president for Salesforce AI.

Salesforce is making it simpler to achieve that goal by packaging together prompts in a way that a reasoning engine developed by Salesforce orchestrates to automate a specific set of tasks, he adds. The reasoning engine interacts with an LLM by analyzing the context of the prompt, determining the appropriate actions and generating the output.

Organizations can either make use of the library of actions that Salesforce has built into its copilot or they can build their own, says Govindarajan. “Each action is registered with the engine,” he says.

The ultimate goal is to make it simpler for end users to automate tasks without having to rely on a professional or citizen developer to create an application using, for example, a low-code tool, he notes.

In general, adoption of generative AI tools continues unabated. A global survey of 10,281 workers published today by Workforce Labs from Slack, a unit of Salesforce, shows usage of AI tools in the workplace rose 24% in the last quarter, with 25% of workers reporting they have tried AI tools for work as of January 2024.

Survey respondents report spending 41% of their time on tasks that are “low value, repetitive or lack meaningful contribution to their core job functions.” A full 80% of respondents said AI is already improving their productivity.

However, while 42% said they are excited for AI and automation to handle tasks from their current job, 31% are neutral and another 27% said they are concerned.

It’s now not so much a question of whether employees will be making use of generative AI to automate tasks as much as it is to what degree. Even in organizations that may have attempted to ban generative AI, there is often widespread usage of shadow services to automate tasks such as creating sales pitches. Ultimately, what’s often required most is training on how best to responsibly leverage the capabilities of these platforms in the most transparent way possible.

Regardless of approach, the workplace is evolving in a way that should reduce the overall amount of drudgery workers experience. There will, of course, undoubtedly be changes that will require adjustments to roles and responsibilities, but the probability that AI is going to eliminate large numbers of positions any time soon remains remote at best.