Learn how to use Simulations to test Fin for Sales before it goes live — so you can see how Fin qualifies leads, what it says, and where it routes them, and catch problems before they reach real prospects.
Note: Simulations for Fin for Sales are currently in beta. There are currently no usage limits for Fin for Sales simulations whilst in beta, but these may be introduced at a later date.
What a Simulation does
A Simulation is an automated test of a single sales conversation. Fin's AI plays the role of a lead, sends an opening message, and has a back-and-forth conversation with Fin for Sales — just like a real prospect would. At the end, the simulation checks Fin's behaviour against the criteria you set and reports a Pass or Fail.
Each simulation can check three kinds of behaviour:
Where Fin routed the lead — did Fin send this lead to the right routing outcome (for example, your "qualified" outcome, or Disqualified, Escalate to support, or Spam)?
What Fin said — did Fin's responses meet a criterion you describe (for example, "Fin asked at least two qualifying questions before routing the lead")?
What data Fin collected — did Fin capture the lead information you expect (for example, that the lead's company size was recorded)?
A Simulation passes only when all of its criteria are met.
Before you start
You need Fin for Sales set up in your workspace, with at least one outcome configured in your Playbook. Routing outcomes are the decisions Fin can make about a lead — for example Qualified, Disqualified and Escalate to support.
Simulations test against your live Fin for Sales setup, including your published content and playbook.
How to access Simulations
Go to Fin AI Agent > Sales > Test.
If you haven't created any simulations yet, you'll see a No simulations yet screen with two options to get started:
Create from scratch — build a simulation yourself.
Generate from Playbook — let Fin generate a starting set of simulations from your existing sales Playbook.
Creating a simulation
Click New simulation (or Create from scratch from the empty state) to open the simulation form. A simulation has two parts: the lead Fin will talk to, and the criteria Fin must meet.
1. Simulate the lead
This section defines who Fin is talking to and how the conversation starts.
Simulate as — choose who the lead is:
New lead (default) — start a fresh conversation as a brand-new lead, using all your live content.
Customer or lead — base the simulation on an existing person in your workspace.
Lead's opening message — the first message the lead sends. Optional — leave it empty to have Fin generate an opening message from the lead's details.
Example: "Hi, I'm interested in your product"
Additional details from lead — information the lead should share in follow-up messages, or the tone they'd use. This shapes how the simulated lead behaves through the conversation.
Example: "Lead is interested in enterprise pricing and is comparing competitors."
Tip: Leave the opening message empty and describe the lead in Additional details when you want to test how Fin handles a realistic, free-flowing conversation rather than one exact phrasing.
2. Evaluate Fin's behavior
This section is where you set the criteria you expect Fin to meet. The simulation must meet all criteria to pass.
Click Add assertion and choose what to check:
Routing outcome — assert which outcome the lead should be routed to. Choose the expected outcome from Select routing outcome. If you don't set an expected outcome, the simulation passes as long as Fin makes any routing decision.
What Fin says (Agent response) — write a single statement, in plain language, that should be true about the conversation. Fin uses AI to judge whether it holds.
Example: "Fin confirmed the lead's budget before offering to book a meeting."
Data collected (Attribute) — check that a specific piece of lead data was captured by the end of the conversation. You can check that an attribute is set, is not set, is equal to a value, or is not equal to a value.
Action — check that Fin made an expected data connector call (for example, looking the lead up in your CRM).
You can add as many criteria as you need.
When you're done, click Save to store the simulation, or Run to save and run it immediately.
Running and reviewing simulation results
You can run a single simulation from its row or from the form, or click Run all at the top right to run every simulation in the list. While simulations are running, the button shows Running….
Each simulation in the table shows its Name, Last run time, and Result.
To see what happened, click a simulation to open the result panel. It shows:
The full conversation transcript between the simulated lead and Fin.
The routing decision Fin made, shown inline as Routed to "{outcome}" by Fin (for example, Routed to "qualified_deal" by Fin).
Each of your criteria with a pass/fail status, so you can see exactly which one failed.
Debugging a failed simulation
When a simulation fails:
Open the result and read the transcript to see how the conversation actually went.
Check the routing decision against what you expected — if Fin routed the lead somewhere unexpected, your routing outcome's qualifying conditions may need adjusting.
Look at which criterion failed. A failed data collected check often means Fin didn't ask for that information; a failed what Fin says check means the conversation didn't meet the statement you wrote.
Note: Because Fin plays the lead with AI, two runs of the same simulation won't be word-for-word identical. Focus your criteria on the outcomes that matter (routing, data collected, key things Fin must or must not say) rather than exact wording.
FAQs
Does running a simulation affect real leads or send real messages?
Does running a simulation affect real leads or send real messages?
No. Simulations run in an isolated test conversation. The lead is generated by AI and no messages go to real people.
Do my simulations test my live Fin for Sales setup?
Do my simulations test my live Fin for Sales setup?
Yes. Simulations use your live content and playbook, so they reflect how Fin would actually behave.
Why did my simulation route the lead somewhere unexpected?
Why did my simulation route the lead somewhere unexpected?
Fin chooses a routing outcome based on the qualifying conditions you've set for each outcome and what the lead said in the conversation. Open the transcript to see what the lead shared, then review the conditions on the outcome you expected.
Will the same simulation always give the same result?
Will the same simulation always give the same result?
Not exactly. The simulated lead is AI-generated, so conversations vary between runs. Set criteria around the behavior that matters rather than exact wording.
Need more help? Get support from our Community Forum
Find answers and get help from Intercom Support and Community Experts