Since its presentation at the last Dreamforce 2024 event, there has yet to be any talk of anything else: Agentforce represents a leap in what is possible to do with generative AI in Salesforce.
Agentforce translates into the 3rd phase of AI evolution within Salesforce.
At the time of the official presentation of Agentforce, Marc Benioff posed the question "What if your workforce had no limits? This is one question that we could not do not only in the last 25 years that I have worked with Salesforce but also in the 45 years that I have worked in the software sector,"- he says.
After testing it, we were challenged to build an agent and realised why the expectations were created. Unlike all the solutions presented so far, it is now possible to find use cases that are very easy to design and quick to implement.
In addition, it allows us to know dozens of use cases where we can access what has been implemented in some companies and even use the prompts used in these use cases.
What do we need to take into account to implement Agentforce in a real environment?
Before implementing an Agent, it is essential to take the same care in the use of any AI, as we illustrate in this article, and only after this crucial work, we set out to define the 5 attributes of the Agent:
Once these attributes have been defined, we are ready to configure the Agent, which is divided into three blocks:
- The definition of the topic.
- The definition of the instructions, that is, the creation of the prompt.
- The definition of the actions that this Agent can perform.
After these steps, we are able to test our Agent and see if its behaviour is as expected. To do this, Salesforce provides a console where we can test the solution before activating it for users or customers.
In addition to these tests, we can monitor the usage and behaviour of agents, allowing them to be optimised over time.
Evaluation and continuous improvement are key factors that ensure that Agents keep up with technological and business evolutions.
Finally, and because I couldn't talk about AI without talking about one of the topics that most concern many companies, the entire development of Agentforce takes into account one of Salesforce's values: security. For this, Salesforce created the Einstein Trust Layer:
After all the announcements from several companies in recent years about AI for enterprise solutions, it was the first time that, in addition to finding several use cases, we were able to understand how they can be implemented quickly, see examples of that, even the prompt used, and, perhaps the biggest differentiator, it allowed us to test and create an agent.
After testing, it became more evident to everyone that this is the most effective AI path, in which we will be able to find the best models to be applied in our companies and the path still to be taken to implement them.
In a few months, Agentforce will be the best friend of teams already working with Salesforce, now without limits.
