Service and sales agents
Agents that answer customers, qualify leads or resolve cases by reading and updating CRM records. What they change matters as much as what they say.
SERVICES / ENTERPRISE SOFTWARE
Evaluate Agentforce agents before they touch your CRM.
Independent testing and evaluation for AI agents built with Salesforce Agentforce: routing, actions, record updates and grounded answers, measured against your business outcomes.
Common Agentforce patterns
Subagents and actions
Agents organized into subagents (formerly topics) with actions such as flows, Apex and prompt templates. Each routing and action choice is a point where behavior can go wrong.
Grounded in your Salesforce data
Answers drawn from CRM records, knowledge articles and Data 360. Permissions, freshness and completeness of that data shape every response.
Quality issues we look for
Wrong subagent, wrong action
The request is routed to the wrong subagent, or the right subagent calls the wrong action or passes the wrong parameters.
The wrong record changed
A case, opportunity or contact is updated that wasn't the one the user meant, or an update is reported but never saved.
Inconsistent across runs
The same utterance routes differently from one run to the next, especially after instructions or Agent Script changes.
What we assess
Routing and action accuracy
Expected subagent and actions for each scenario, compared with what the agent actually chose, across realistic and ambiguous phrasing.
Outcome verification
For actions that change data, we check the resulting records in a sandbox, not only the agent's confirmation message.
Grounding and permissions
Whether answers stay faithful to the records and articles available to the agent user, and whether access limits are respected.
Working with native Agentforce testing
Batch test suites
Where they fit, we use Agentforce's built-in testing (the Tests tab in Agentforce Studio, or Testing Center for legacy agents) with your own test cases, expected subagents, expected actions and expected responses.
Preview and session traces
We use builder preview and session tracing to explain failures: which subagent, which action, which inputs and what came back.
Calibrated, repeated and recorded
Native results are compared with your reviewers' judgment, key cases are repeated to measure variability, and agent versions are recorded with every run.
Related services
Evaluation Assessment
A baseline for one Agentforce agent or workflow, with prioritized findings and a release recommendation.
Explore serviceHallucination Testing
Measure how often your Agentforce agent invents facts, policies or actions, and which fixes reduce it.
Explore serviceManaged Evaluations
Recurring tests as your Agentforce agent, prompts, models and data change, with a monthly quality report.
Explore servicePlan the engagement
Your inputs
A sandbox with representative data, an agent user with agreed permissions, version details, business rules, example conversations and a domain reviewer.
Your deliverables
An agreed benchmark, prioritized findings with reproducible cases, reusable test suites and recommendations for the next iteration.
Support after handover
We can retest fixes, extend coverage to new Agentforce workflows or move to recurring Managed Evaluations as your agent changes. Platform usage and licenses stay in your account.
Make your next decision with evidence.
Start with one workflow, clear success criteria, and an evaluation plan your team can use.