#SalesforceAI10 discutindo

Hi @Galla Rajeswari Agentforce is powerful, but there are some challenges we need to consider in real projects.

1. Response Time: Complex requests can take longer to process.
2. Action Timeouts: Apex, Flow, or API actions may fail if they exceed time limits.
3. Subagent Selection: Sometimes the agent may choose the wrong subagent.
4. Response Accuracy: Even with correct data, the agent may give incorrect or incomplete answers.
5. Data Access: Permissions and sharing settings can affect agent responses.
6. Testing: Testing different user questions and scenarios is not always easy.
7. Cost: Usage costs can increase with frequent agent interactions.

In my view, clear instructions, proper testing, and monitoring are very important before production deployment.

Agentforce is useful, but we should not depend on AI alone for critical business decisions.

Hey Jiyaulla,

Here’s how I would approach the solution design:

1. Agent Boundaries & Routing

• Split by persona: Don't build one monster agent. Keep Employee Desk (IT/HR Q&A via Knowledge), Coworker (sales/service reps on record pages), and OOTB Service Agent in clean, separate buckets.

• Topics & Hand-offs: Limit each agent to 4–6 tightly scoped Topics. Write explicit "When to use" instructions so the reasoning engine doesn’t cross wires. Always map human escalation via Omni-Channel, carrying the chat summary over.

2. Actions & Grounding (The Engine)

• Actions: Use Autolaunched Flows for standard CRM updates and Invocable Apex for external API lookups. Keep JSON outputs short; feeding raw multi-page payloads burns tokens and slows reasoning down.

• Flex Prompt Templates: Stick to a 4-part prompt: Role, Grounded Context, Guardrails (what NOT to do), and exact Output Format.

• Knowledge & Data Cloud: If grounding with PDFs or policies, ensure your Data Cloud search indexes chunk properly. Watch Field-Level Security—if the running agent user can’t see the field, the LLM hallucinates.

3. Trust, Guardrails & Security

• Einstein Trust Layer: Decide upfront if PII needs data masking and where audit logs get retained.

• Hard Refusals: Put explicit negative prompts inside each topic (e.g., "Never disclose salary bands; route user to Workday").

4. Licenses & Consumption Reality Check

• Cost control: Agentforce bills per conversation, and custom prompt calls burn credits. Don't use the LLM for things a simple Flow lookup or formula can solve deterministically.

• Token limits: Heavy record grounding causes latency spikes. Keep grounding payloads minimal.

5. ALM & Deployments (Where Projects Trip Up)

• Metadata API: Topics, Actions, and GenAiPromptTemplate deploy through standard DevOps pipelines (Copado, Flosum, SFDX).

• Manual Post-Deploy: Playbooks always break here. You must manually:
  1. Assign permissions and FLS to the Agent Execution User.
  2. Activate the Agent version in Target Org.
  3. Validate Data Cloud search index syncs.

•  Metered: Agentforce usage is charged based on consumption; for agentic usage, the main unit is generally actions, not simply tokens or conversations.  
• Flex Credits: A standard Agentforce action currently consumes 20 Flex Credits, while custom actions are also measured by the number of actions called. 
• Unmetered: Certain licensed users/features can use qualifying Agentforce capabilities without consuming Flex Credits, when the required licensing, permissions, and execution conditions are satisfied. 
• Important: “Unmetered” does not mean everything in the agent is free; underlying services such as Data 360 can still consume credits. 
• Employee Agents: Agentforce add-ons provide unmetered Agentforce usage for employees; the Agentforce User License, by contrast, provides metered employee usage and requires Flex Credits. 
• Customer-facing Agents: Salesforce currently offers both Flex Credits (pay per action) and Conversations (pay per conversation) for customer-facing agents. 
• High-volume usage: Consumption-based pricing makes the cost depend on actual usage, so action/conversation volume and the specific features used need to be estimated rather than assuming one fixed cost. 
• Monitoring: Use Digital Wallet to monitor Agentforce consumption and usage in the org. 
• Forecasting: Salesforce recommends reviewing actual Digital Wallet consumption and working with your Salesforce Account Executive for contract-specific pricing and usage estimates. 
• Bottom line: Metered = pay according to usage; Unmetered = qualifying usage included through specific licenses/permissions. Which is more cost-effective depends on your users, volume, features, and contract—not simply whether the agent is internal or external. 

Also, go through the doc it helps you--->

 Agentforce and Generative AI Usage and Billing | Salesforce Help

Considerations for Agentforce Employee Agent | Salesforce Help

Metering for Agentforce and Generative AI Usage | Salesforce Help

Salesforce Agentforce Pricing | Salesforce 

Hi All, 

I am planning to take the Salesforce Agentforce certification. If anyone has vouchers available, I would greatly appreciate it if you could share them with me via email:

Thank you.