Inside Agentforce: Our experience with Salesforce’s game-changing AI tool
Agentforce is here — an AI assistant that works round the clock, handling routine customer service tasks, providing real-time insights to your team, and making every interaction more efficient. As Salesforce’s latest innovation, it’s not just another tool — it’s a game-changer for customer service teams looking to work smarter, not harder.
When Salesforce announced Agentforce, it created a buzz that couldn’t be ignored. But we didn’t just read about it; we already applied it to our CheckMyNumber app to explore its capabilities and see its impact firsthand. Spoiler: it exceeded our expectations. And while we prepare for a public release of this extension, here are some things we learned along the way.
What is Agentforce?
Agentforce is Salesforce’s AI platform, which allows you to create autonomous agents to accelerate operations and improve customer experiences. These agents operate 24/7, assisting with handling customer service requests, supporting sales teams, and managing marketing activities.
Key features of Agentforce include:
- Multi-channel availability across websites, CRM, Slack, mobile apps and more;
- Customizability for specific roles or industries;
- Native integration with the Salesforce ecosystem.
How does Agentforce work?
Agentforce operates using Salesforce’s low-code Agent Builder, allowing companies to build, customize, and deploy agents quickly. These agents can be designed to handle specific use cases, ensuring they meet the business's needs. Here’s how it works:
- Agent development
Utilize Salesforce tools like Flows, Prompts, Apex, and MuleSoft APIs to design and deploy agents. You can modify pre-built agent templates or create them from scratch.
- Access to trusted data
Agents can access real-time, secure data from Salesforce, ensuring they perform tasks accurately. They can pull customer records, analyze data, and provide insights during interactions.
- Built-in guardrails
You can set up clearly defined boundaries to ensure agents operate within set guidelines. Tasks beyond the agent’s scope are escalated to human representatives, ensuring sensitive or complex issues are handled appropriately.
- Seamless integration
Agents work with existing Salesforce tools, ensuring compatibility with CRM systems. They are easy to scale and adapt as your business needs grow.
- Task automation
Routine operations like answering FAQs, scheduling meetings, or managing simple transactions are automated. Agents reduce the workload for human employees and allow them to focus on more critical tasks.
Applying Agentforce to the CheckMyNumber app
Our CheckMyNumber application is designed to validate and format phone numbers, ensuring clean, accurate data for businesses. When Salesforce introduced Agentforce, we saw an exciting opportunity to integrate its capabilities into our app to automate and enhance several processes.
To make the most of Agentforce, we tapped into its core functionality while adding custom features tailored to our needs. Here’s how we implemented it:
1. Automating customer query responses
With Agentforce, we developed custom functions that allowed the chatbot to:
- Validate phone numbers in real time.
- Prompt users for additional details, like a country code, when required.
- Offer immediate responses based on the app’s functionality and Salesforce data.
For instance, if a user asked, “Can you check this phone number?” and the default Agentforce lacked this capability, we created custom agent actions to enable it. These actions included validation flows and invocable objects that Agentforce could execute autonomously.
2. Real-time recommendations
Agentforce wasn’t just limited to answering queries. It provided real-time suggestions to guide users through complex processes. For example:
If a number cannot be validated due to missing information, Agentforce suggests adding a country code or switching to a specific format (e.g., international). It supported dynamic interactions by interpreting partial user inputs (e.g., recognizing “Britain” and auto-filling the correct country code).
3. Enhancing backend processes
Agentforce significantly improved backend workflows by automating ticket categorizationquery categorization. Using topics and agent functions, we:
- Grouped similar queries under predefined categories, reducing the manual effort for our support team easing the communication between customer and AI agent.
- Created business logic that mapped each action to a specific topic, ensuring smooth operation and clear instructions for the agent.
Real-life example
One of the most impactful scenarios we tested involved bulk phone number validation. A user requested validation for several numbers in a single message. Initially, Agentforce responded with default functionality, unable to process multiple numbers simultaneously. We addressed this by:
- Adding a custom bulk validation action;
- Implementing a flow that allowed the user to upload a list of numbers or reference a Salesforce ListView;
- Ensuring that Agentforce handles each number efficiently and provides real-time updates on progress.
For instance, when a user entereds “Validate these numbers: +441234567890, +331234567890,” Agentforce quickly validatedvalidates and formattedformats them as international numbers, then updatedupdates the corresponding Salesforce records. This automation reduced response times significantly and eliminated the need for manual intervention.
Noltic’s perspective on Agentforce
Integrating Agentforce within the CheckMyNumber app wasis a journey filled with valuable lessons. While the tool showcased impressive capabilities, it also revealed areas that required optimization. Here’s what we learned from our hands-on experience:
1. Strengths of Agentforce
Agentforce brought several standout benefits to our app, making it an invaluable addition:
- Contextual understanding
Agentforce excelled at maintaining conversational context, enabling smoother interactions with users. For example, if a user provided partial information like “Britain” instead of a full country code, Agentforce could interpret the input and autofill the correct data.
- Ease of customization
Salesforce’s low-code environment allowed us to tailor Agentforce to meet our specific requirements. By creating custom agent actions and grouping them under topics, we ensured the AI could perform key functions like validating and formatting phone numbers or handling bulk operations.
- Scalability
With its ability to engage users autonomously across channels like SMS or WhatsApp, Agentforce scaled effortlessly to meet increasing demands without adding workload to our support team.
- Real-time efficiency
Tasks like real-time validation of phone numbers or providing immediate recommendations to users became seamless, saving time for both users and our team while improving overall accuracy.
2. Challenges we encountered
While Agentforce proved powerful, there were areas where it required additional input or development:
- Limited default capabilities
Out-of-the-box, Agentforce’s functionalities were somewhat basic. For instance, it couldn’t validate or format phone numbers independently until we built custom flows and invocable actions. We did some extra work to unlock its full potential.
- Occasional context issues
Although Agentforce generally understood the conversational context, it sometimes struggled with more complex or ambiguous inputs. For example, processing multiple phone numbers in a single request required fine-tuning to handle such bulk operations effectively.
- Dependency on business logic
Agentforce relied heavily on well-defined topics and business rules to perform accurately. While this gave us control, it also required careful planning and constant updates to ensure the agent functioned as intended.
- Salesforce limitations
Some features, like deep analytics on how users interact with Agentforce or seamless file uploads, are currently not supported. Salesforce indicated that such features might become available in future releases.
3. Lessons learned
Our experience highlighted the importance of collaboration and customization when implementing Agentforce:
- Custom development is key
By building custom actions and designing intuitive topics, we extended Agentforce’s capabilities and empowered the agent to handle tasks far beyond its default scope.
- An iterative approach is essential
Agentforce’s performance improved significantly as we applied and refined its configurations step-by-step. Regular feedback loops helped us identify and address gaps quickly.
- Documentation matters
We created comprehensive documentation for all customizations, ensuring our clients and teams could easily adapt or extend the functionalities if needed.
- Future-proofing is a must
Knowing that Salesforce plans to expand Agentforce’s features, we avoided over-packing the agent with Managed Package configurations. Instead, we focused on scalable, modular designs that can evolve as the platform matures.
4. Opportunities for improvement
Looking ahead, we see the potential for Agentforce to become even more impactful:
- Enhanced context awareness
Improving the agent’s ability to process bulk inputs or multitask efficiently will make it even more user-friendly.
- Bulk validation features
Integrating functionality for bulk operations directly into Agentforce would streamline processes further, especially for apps like CheckMyNumber.
- Advanced analytics
Adding detailed user interaction analytics would provide insights to fine-tune configurations and improve the overall user experience.
- Broader file support
Allowing users to upload CSVs or other file formats for data processing would unlock new use cases.
We’re excited to see how Agentforce evolves and apply it to our projects and other AppExchange products. In the meantime, look out for the official public release of Agentforce for CheckMyNumber on our blog or LinkedIn page!
together