Agentforce Agents: Checkpoints, customization & configuration

Choosing and successfully deploying the right enterprise-grade AI solutionfor business can prove to be a tall order in the wake of new developments. New tools in the market exemplify predictive and generative AI’s potential that empower businesses to innovate and enhance customer engagement.

In this blog, we look at the following:

Check points to consider first:
Now, here’s the autonomous agent
Getting into action: 5 key parameters
Use cases: customized agents
Tools: One size does not fit all
Try it out at the workshop by Movate

Not all the features of AI agents would be relevant for the business. Let’s look at the checkpoints to consider.

Before diving headlong into an enterprise-grade AI solution, start by chalking out the desired business outcomes—whether it’s user friendliness, improving customer service efficiency, productivity, increasing sales conversions, or streamlining operations. Consider the solution’s features that align with specific needs of the business and how they impact customer experience and employee productivity.

What specific business problems do you want to solve through AI agents?

Assess the processes where AI can be integrated. Instead of a big bang approach, taking a strategic approach ensures that the chosen AI tools are not just advanced technologies, but also practical solutions tailored to the organization and its customers. Look for repetitive/monotonous tasks that consume significant time and resources, such as data entry or customer follow-ups. By mapping out these processes, effectively integrate AI solutions to enhance productivity and drive meaningful engagement with customers.

Prioritize requirements by identifying processes that would benefit the most from AI agent deployments. Take the case where the support team is overwhelmed with inquiries, deploying an autonomous AI agent (agentic AI) can provide 24/7 support and handle routine questions, freeing up human agents for more complex issues.

Piloting shortlisted tools is essential to know how they work with the existing technology stack and thwart any negative impact on customers, employees and the business.

Having clarity on the above checkpoints brings us to the next topic of Agentforce by Salesforce.

An Agentforce Agent is a proactive, autonomous application that provides specialized, always-on support to employees or customers.

Agentforce agents are equipped with the necessary business knowledge to execute tasks according to their specific role.

Getting Agents into action

Identify the business functions, processes, and use cases for Agentforce agents under these 5 specific parameters.

Role definition: Each agent is tailored to fulfill specific roles within the organization, ensuring that they can address unique business needs—think of it as the agent’s purpose on the team.
Trusted data: Agents operate based on a defined context, utilizing both structured and unstructured data from the CRM and external sources via the Data Cloud.
Actions: These are goals an agent can fulfill—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.
Channels: Agents can interact through various channels such as web, mobile, WhatsApp, and Slack, facilitating seamless communication.
Guardrails: These act as the bylaws or guidelines under which the agents operate and what they can or 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.

Identify the various sources of data and actions of those agents.

Exhibit 1

Use Cases

Movate has identified several unique use cases for custom agents across different personas. Spin up out-of-the-box agents across Customer Service, Sales, Marketing, Commerce and more with Agentforce—deploy industry agents easily.

In the Sales Coach example below, it gives a very personalized pitch for a customer and the problem-statement. Here are some examples of use cases across personas.

Exhibit 2

Customization tools

One size does not fit all and Movate provides tools to configure & deploy customized agents for specific roles (see exhibit 2 above).

These include Prompt Builder, Agent Builder, and Model Builder.

Prompt Builder: Customizes agent responses with precise prompt engineering (good prompt versus a bad prompt). A well-defined robust model understands customer intent despite unexpected or irrelevant inputs. The tool looks into all aspects such as actions to take, inputs to ignore, and invoking relevant answers.
Agent Builder: Low-code tool for quick custom agent creation. It integrates flows, Apex, APIs for specialized actions; builds different types of agents.
Model Builder: Registers, tests, & activates custom AI models. It simplifies LLM integration into Salesforce workflows.

Despite the availability of standard agents with standard actions that are pre-built, customization and configuration is needed. Identify the boundaries or limits within which the agents should operate, and which agent will do what. For instance, in the case of insurance where the agent has to pose a specific question to the customer.

Work with a trusted partner

Here are the key areas where Movate’s Salesforce consultants can help:

Implementation: Build and deploy agents, create and unify knowledge and data, vectorize unstructured data.
Advice: AI & data strategy, AI model optimization, continuous optimization.
Change management: Process reengineering, training and adoption.
Managed services: Ongoing support and performance optimization.
GTM solutions (support of custom solutions): custom applications, and AI-powered tools.

Movate proposes a 5-week discovery & implementation workshop where experts are ready to deploy any 1 agent with 2 predefined use cases working on CRM data. Additional deployments can be taken up on a ‘need-by-need’ basis.

Exhibit 3

Here is the breakdown of activities under each week.

Movate’s week 1 activities:

Conduct a project initiation meeting with key stakeholders.
Define project scope, objectives, and success criteria.
Identify and assign roles and responsibilities.
Perform initial data review to understand required data points and sources.
Outline initial requirements for the predefined use case.

Week 2 activities:

Conduct a detailed risk assessment for the selected use case.
Define a mitigation plan for identified risks.
Assess the current state of data and systems for agent configuration.
Finalize design specifications for the RAG (Salesforce Data Only) agent.

Week 3 activities:

Develop and configure the agent with up to two topics.
Implement up to 2 standard actions and 1 custom action (without integration).
Schedule and conduct feedback sessions with stakeholders on initial configurations.
Refine agent actions and responses based on feedback.

Week 4 activities:

Establish governance and operational frameworks for agent use.
Outline policies for agent performance monitoring and updates.
Finalize operational guidelines for deploying and managing agents.
Prepare training material and guidelines for end-users.

Movate’s final week activities:

Conduct a thorough validation and testing phase to ensure agent readiness.
Deploy the agent in a production environment.
Monitor agent performance and gather feedback from users.
Schedule ongoing feedback sessions and define improvement areas.

Partner with Movate to transform data into actionable insights with Agentforce by Salesforce. Contact us for the workshop.

**Indranil Sengupta – Senior Solution Architect**

Indranil Sengupta has 14+ years of experience as a Senior Solution Architect for Digital Transformation using Salesforce. He specializes in Salesforce org assessment and consolidating, revenue cloud and consumer goods cloud. LinkedIn.

Related information