Where We Are With AI
Salesforce has been utilizing AI for more than a decade in the form of predictive AI with functionality such as lead scoring, whitespace analysis, churn analysis, market shift detection, predictive analytics, and more.
More recently generative AI that uses Large Language Models (LLMs) to process natural language interactions has come to the forefront of everything AI. Most of us are familiar with ChatGPT, which is an example of generative AI, where you input some text in a prompt and ChatGPT replies to you with an answer. A very simplified way to understand how this is done is to think of it as autocomplete functionality. The LLM guesses, with high probability based on massive training, what the next word or phrase should be in its response, given its context, and then replies with that, which when taken as a whole appears to be a well thought out response.
Salesforce has taken the next step beyond simple generative AI in the form of autonomous agent processing, which is known as Agentforce. At the core of Agentforce is Salesforce’s proprietary Atlas Reasoning Engine which is responsible for determining what actions should be taken in what order and then executes them. Salesforce admins set up Agentforce by specifying jobs that an agent can do and specifying which actions (e.g., update a record, initiate a return, answer a question with knowledge, generate a sales POV, etc.) can be executed as part of fulfilling an individual job. It is very important to set up the jobs (a.k.a. topics) with detailed and correct instructions so that the agent follows your business rules as closely as possible.
No Need to Build AI Yourself
If you were to try to build your own agents, it would be prohibitively time-consuming and costly. The beauty of Salesforce is that you get all the infrastructure along with many standard agent functions out-of-the-box, no matter the size of your company. You don’t have to be a Fortune 50 company to have access to and benefit from the latest cutting-edge AI technology. Additionally, you get all the updates and innovations that Salesforce will make to Agentforce in the future. This has always been one of the strong arguments for why to choose Salesforce in the past and it holds true with Agentforce.
Permissions and Trust
The topics of permissions and trust cannot be understated. One of the major concerns that customers have about this new type of AI is data leakage. They worry that, with access to their company’s data, an AI could accidentally expose it to someone and that hackers will find a way to get that sensitive information by interacting with an agent.
Because Agentforce is built on top of Salesforce, all the record access controls and object permissions that you’ve been using for years to secure your data in Salesforce work with Agentforce out of the box. If you are rolling your own AI solution outside of Salesforce, you must code your own data access rules, but with Agentforce you use the same trusted way of setting up data access. No matter what a person asks of your agent, you can be assured that the agent will never include any data to which you have not given it access.
The other concern around security that customers have is that they are worried that their data will be sent to an LLM such as ChatGPT and that LLM will then store and have access to their data. Salesforce has thought of that and there is a zero-retention policy in place so that no LLMs keep any of your data for training or any other purposes. Additionally, sensitive data can be masked before it is sent to the LLMs and unmasked and merged back into the response, so that it never even leaves Salesforce. This is all done through the Agentforce Trust Layer that is a core part of the agentic workflow.
How to Get Started Now
It seems like AI appeared out of nowhere just the other day with ChatGPT and has been at the top of mind in everything that we do. At first, it was not totally clear how this technology could be leveraged in our organizations and there were not any concrete business implementations. It was safer to take the wait and see approach for many. Now that Salesforce has made Agentforce available to everyone, there is no excuse for sitting on the sidelines and risk falling behind your competitors that are embracing AI.
Luckily, Salesforce has introduced Salesforce Foundations which is available for free to all Enterprise Edition and above customers. Among other features, it includes free Data Cloud credits and Agentforce conversations so that you can immediately begin working on Agentforce initiatives.
Trifecta offers Agentforce implementation services to get you up and running very quickly on solid grounding with Agentforce.
After reading this article, hopefully you have a better understanding of how powerful Agentforce really is and will plan on an implementation in the near future!
Agents for Sales Cloud and Service Cloud
What jobs could you assign to the Agents? For those with Sales Cloud, Service Cloud, or both, join me at Trifecta’s upcoming webinar where I am a speaker covering more information about Agentforce and various applications for Agents within these specific clouds.
Trifecta’s webinar: Unpacking Agentforce for Sales Cloud and Service Cloud. It will take place on December 17th at 2 PM EST. RSVP Here for the link to join!
