Salesforce Agentforce is measured by what its agents finish, not by how often they are used: cases resolved without a human, leads qualified and routed, time saved for staff and the quality of outcomes that follow. A useful dashboard combines Agentforce’s own analytics with standard Salesforce reports on cases, leads and opportunities, and compares agent-handled work with a human baseline.
What is Agentforce?
Agentforce is Salesforce’s platform for autonomous AI agents that work inside the Salesforce ecosystem. Agents are built in Agent Builder with topics, instructions and actions, and they use data from the CRM and Data Cloud through the Atlas reasoning engine. They can answer customers on web chat or messaging, assist service reps, qualify sales leads, help with marketing campaigns in Marketing Cloud and run actions through Flows and Apex.
The platform includes guardrails through the Einstein Trust Layer, such as data masking and audit trails. Pricing and packaging have changed several times since launch, including per-conversation and credit-based models, so check current terms with your account executive.
Why is measuring agents different?
Classic automation either works or fails. Agents make judgement calls, so they can complete a task badly. A service agent may close a case that the customer reopens the next day. A sales development agent may book a meeting with someone who never had budget.
That means two layers of measurement: activity and volume, which show adoption, and outcome quality, which shows value. Dashboards that only show volume often look good while customers grow frustrated.
Which KPIs matter for service agents?
- Containment or deflection rate: conversations resolved without handoff to a human.
- Reopen rate: cases closed by the agent that are reopened within a set period.
- Customer satisfaction (CSAT) for agent-handled versus human-handled conversations.
- Average handle time for cases escalated from the agent, compared with direct human cases.
- Escalation reasons, grouped into themes such as missing knowledge or policy exceptions.
Reading containment carefully
A high containment rate is only good if customers got answers. Pair it with reopen rate and CSAT, and sample transcripts each week. Customers who give up and leave can look exactly like resolved conversations.
Which KPIs matter for sales and marketing agents?
- Leads engaged by the agent and response time to inbound leads.
- Meetings booked and meeting-held rate.
- Conversion from agent-qualified lead to opportunity, compared with human-qualified leads.
- Pipeline value created from agent-touched leads.
- Opt-out and complaint rates on agent-sent outreach.
For marketing use, such as campaign brief or segment creation, track time from brief to launch and campaign performance against comparable campaigns built manually.
Where does the data come from?
Salesforce offers several sources, and the right dashboard pulls from each.
Agentforce analytics
Agentforce includes built-in analytics and conversation logs, with views on sessions, topics used, actions run and escalations. Agent session data can be stored in Data Cloud for deeper reporting.
Standard objects
Cases, Leads, Opportunities and Messaging Sessions carry the outcome data. Add a field or record type that marks whether an agent handled or touched the record. Without that flag, you cannot compare agent and human work.
Surveys and feedback
Salesforce Feedback Management or post-chat surveys provide CSAT. Thumbs-up and thumbs-down signals on agent replies help find weak topics.
How do you build the dashboard?
Use CRM Analytics or standard Lightning dashboards, and keep the first version to one page.
- Top: volume handled by agents this month, containment rate, CSAT, compared with the human baseline.
- Middle: quality indicators, reopen rate, escalation reasons, conversion from agent-qualified leads.
- Bottom: business value, estimated hours saved and pipeline from agent-touched leads.
- Side panel: cost, conversations or credits consumed against budget.
Estimating hours saved
Take the number of conversations fully resolved by the agent and multiply by the average human handle time for similar cases. Subtract time spent reviewing agent work. Present the result as an estimate with the formula shown, so finance can check it.
How Alien Road applies Agentforce measurement
In CRM projects on Salesforce, Alien Road starts measurement planning during discovery, before any agent is built. We agree with the client which outcomes the agent owns, which human baseline to compare against and which fields will flag agent-touched records.
Setup includes adding those flags, building the one-page dashboard, connecting surveys and setting a weekly transcript review with the service or sales lead. We also record the cost side, so value is always shown against what the agent consumes.
Monthly reports cover the KPIs above, the top escalation reasons and the fixes made, usually to knowledge articles, topic instructions or actions. If an agent’s quality metrics fall below the human baseline, the report recommends narrowing its scope rather than adding more volume.
What mistakes skew Agentforce numbers?
- Comparing agents with humans on different case types; agents often get simpler cases.
- Counting conversations as resolved when customers simply stopped replying.
- Ignoring cost per resolution, which can exceed human cost for complex topics.
- Changing instructions mid-month without noting it on the dashboard.
- Reporting to leadership without transcript examples, which builds false confidence.
How often should the dashboard be reviewed?
Weekly reviews suit the first two months after launch, when instructions and knowledge change often. A short meeting with the service or sales lead is enough: look at the quality row first, read three or four transcripts together and agree one fix. After the agent settles, move to a monthly review with leadership and keep the weekly transcript check as a team habit.
Who should attend?
- The agent owner, usually an admin or operations lead who can change topics and instructions.
- A frontline lead from service or sales who knows what good outcomes look like.
- Someone from finance or revenue operations for the monthly cost discussion.
Where to start this week
Add a simple checkbox or picklist to Cases and Leads that marks agent involvement, and make sure every agent flow sets it. That single field makes every later comparison possible.
Then pull a baseline for the last three months: case handle time, CSAT, lead response time and lead-to-opportunity rate. With the flag and baseline in place, your first Agentforce dashboard can show value instead of activity.
Keep reading: CRM System · Crm System