Agentforce: What Salesforce ISVs Need to Know

Agentforce was all the buzz at Dreamforce this year. Unlike other announcements from Salesforce that felt like pre-announcements, aka available in 12-18 months after the announcement, Agentforce seems to be more real and more of a present focus than announcements of years past.

As an ISV, here are a few things to know about the Salesforce 2024 Agentforce announcement. 

What is Agentforce?

At a super high level, Agentforce is a chatbot that can be deployed on a website, your Salesforce CRM and elsewhere. The Agentforce bot is powered by actions and flows powered by Salesforce and third-party ISVs. 

For Salesforce customers, you can configure an agent from scratch on Agentforce with existing Salesforce tools like prompts, flows, Apex, and APIs.

For a Salesforce ISV, you have the opportunity to create pre-built actions or templates for Salesforce customers to deploy.

Agentforce Partner Network 

Salesforce has launched the Agentforce Partner Network. They are calling it “The World’s First Agent Ecosystem”. 

If you’re an ISV looking to build an agent or template for Agentforce, you can do so via Agent Builder.

Salesforce Agent Builder

In Agent Builder, an ISV can add or configure jobs to be done for their agent by topic or action. It is the low-code builder for Agentforce where users can leverage existing tools like Flows, prompt templates, Apex, and APIs to configure their Agent on Salesforce. 

• Topics = define the guardrails, in natural language, and the set of actions that can be used within those guardrails. 
• Actions = can be simple sets of secure API calls, actions engineered for a specific industry, complex multi-step workflows, or secure integrations to third-party agents. 

ISVs can also build and distribute pre-built agents and agent templates. These Agentforce templates assemble a set of actions and topics together into a pre-tested agent for specialized industry, line of business, or functional use cases. These partner-built, out-of-the-box agents can be deployed from the AppExchange. You can see these agents up in minutes to work just like the Salesorce-provided, out-of-the-box-agents.

To Set Up: Create a job to be done by the Agent by defining topics, giving natural language instructions for that topic, and creating a library of actions for it to choose from. Easily monitor an Agent’s plan of action and test its responses right in Agent Builder.

Ready to get started? You can join a virtual Agentforce workshop to help in building your first agent. Sign up here. You can take a Trailhead as well to kickstart your Agentforce build. The sessions is 20 minutes long and requires a custom Agentforce Playground.

Examples of agents available today on Agentforce

Agentforce agents can be broken down into the following areas:

• Content creation and management agents
• Sales and customer insights agents
• Industry automation agents

Content creation and management agents

• Google Workspace: Create, find, and share Google Docs and Slides, based on CRM data.
• Box: Surface the right Box content for an account, lead, or opportunity and use natural language to interact with documents.
• Docusign: Generate partnership document, route for signature, and track doc and signature status.
• Honeywell: Assemble compliant training materials for life sciences workers based on complex regulatory standards.

Sales and customer insights agents

• Zoom: Access analytics and key insights from calls, meetings, and other productivity solutions from Zoom Workplace.
• Moody’s: Compile customer insights from company overviews, news articles, and financial filings.
• Korn Ferry: Suggest personalized seller actions and schedule coaching to improve.
• Sprout Social: Summarize customer cases with social data to understand what drives their satisfaction.
• Tribble: Increase sales team productivity by automating RFPs and speeding up rep onboarding.

Industry automation agents

• Copado: Generate developer and testing materials to deliver Salesforce Platform apps faster.
• Coupa: Optimize budget spend, monitor compliance, and find budget efficiencies.
• OpenText: Enable insurance adjusters to verify claim-specialist findings and expedite a claim.
• GoMeddo: Optimize appointments across patient, doctor, technician, and exam room calendars.
• Asymbl: Identify candidates based on job needs and schedule interviews with top prospects.

Agentforce terms to know

Role – The role is the agent’s purpose. It defines the job to be done and the broader goals the agent should achieve on your team.

Trusted data – Trusted data is the information the agent needs to carry out its role. This could include company knowledge articles, CRM data, external data via Data Cloud, public websites, and so on.

Actions – Actions are the predefined tasks an agent can execute to do its job based on a trigger or instruction. For example, it could run a flow, prompt template, or Apex. By exposing metadata about defined workflows and AI automations, this makes the agents capable of high-value tasks. These may be actions taken on the employee’s behalf –  for example, these AI agents can help a sales rep draft an email to a customer, create a close plan, and kick off a 1:1 nurture cadence for the customer that includes all their upcoming marketing activities.

Guardrails – Guardrails define what the agent can and cannot do. These can be natural-language instructions to escalate to a human, or could come from built-in security features in the Einstein Trust Layer. Guardrails can help identify when to escalate or hand off a task to a human as well. 

Channels – Channels are the applications where agents can do work. This can be your website, CRM, mobile app, Slack, and more.