Unless you’ve been hiding under a rock for the past two weeks (or actively avoiding the deluge of Dreamforce content on LinkedIn – no shame here, we know it can be a lot), you’ve heard the tale of Agentforce. But what is it? How do you use it? And what does it mean for associations and nonprofits?
What is Agentforce?
Agentforce, formerly Einstein Copilot, is an autonomous AI agent capable of operating independently to provide 24/7 support to employees and customers that far surpasses the functionality and level of service provided by AI-based chatbots of yore.
Agentforce communicates in natural language, responding across channels, and performing tasks proactively within guidelines set by your organization. It can be used in any cloud product and relies on the foundation of the Atlas Reasoning Engine and Data Cloud.
How Does Agentforce Work?
Agentforce has five key building blocks to support functionality:
- Role: The role is the agent’s purpose and the job to be done. Think of this as a job description for your agent that outlines the broader goals it should achieve on your team.
- Data: Like any AI technology, an agent cannot operate without data. This might include internal data like operating procedures, external content like blog articles or websites, AMS data, etc., and will rely on Data Cloud.
- Actions: Actions are predefined tasks an agent executes to fulfill its role and are based on a trigger or instruction.
- Guardrails: These allow your organization to limit what an agent can do and when you would like to involve a human in the process.
- Channels: AI Agents can work in multiple channels including on your site, CRM, Slack, and more.
On a more detailed functional level, agents operate on topics and actions.
Topics are the “jobs to be done” for your agent. You can think of a topic as a job that can be achieved with a consistent set of tools (actions) and rules (instructions) for how to do that job. To create a topic in Agent Builder, start by naming, describing, and defining the scope of the topic in natural language. To finish the topic, create a list of natural language instructions to act as rules that define how the agent should complete its tasks.
Actions are the tools the Agent uses to do the job. Examples include searching knowledge-base articles, looking up information (e.g. order status), or updating data (e.g. changing the address on an account, creating a case record).
In our demo, you can see how the agent takes input from a customer and then applies the instructions and actions to reason a response.
How Can Associations Use Agentforce?
How frequently does your staff handle requests like these?
- I need to renew my membership.
- I can’t log into my account, can you help?
- How do I pay for my outstanding balance on a specific order?
- I want to register for XYZ event, where do I go?
- Can you update my username from to ?
If you’re anything like most associations, your answer is, “often”.
Agentforce is still in the early stages of release, but the potential benefits for associations are obvious and two-fold.
- Association staff can spend less time on tedious tasks like those mentioned above.
- Association members get instant, high-quality service.
Let’s walk through an example of using an agent for membership management.
A member reaches out via your website’s chat to get help with membership renewal. You’ve built your agent to process this request by taking payment, updating membership status, and updating membership renewal dates. You can further customize this process by:
- Sending them to your custom renewal page.
- Providing an order summary.
- Asking the member if they would like to complete the payment using a card on file – “Would you like to renew your membership using the card ending in 1002 or 0000?”
- Asking the member to make a donation based on their history of previous donation behavior.
This highly personalized experience provides your member with instant and comprehensive service while removing the process from your team’s to-do list.
