Test before launch
Simulate
Auto-generate scenarios & run calls
Evaluate
Metrics & transcript analysis
Optimize
AI-rewritten prompt improvements
Monitor in production
Monitor
Live production call observability
Co-Pilot
Deep analysis & what to fix next
Flow & Opportunities
Where calls go wrong, what to change
IndustriesPricingAPI documentationAbout usBlogs
Book a demoLog inSign up

See where your calls go wrong, and what to change.

Flow shows the path every conversation takes through your agent. Opportunities turns the mistakes that keep repeating into a short list of prompt changes.

Book a demo
app.rubrichq.io/flow
A
Acme Health's Orga…
Owner
Platform
Dashboard
Test Agents
Metrics
Co-Pilot
Flow
Opportunities
Prompt Optimizer
Evaluation
Live Simulations
Batch
Observability
Conversations
Notifications
Settings
A
Flow
Observability (Production)SimulationsClick a stage or band to see its conversations
Biggest divergence: 29% of conversations that reach Balance end there
Greeting1,240Verification1,120Balance1,010186Payment options72064Paid in full430Payment plan180Ended early630

290 conversations ended after hearing the balance

Example data. Hover or tap a band to see its conversations.

Find the leak, not just the failed calls.

A pass rate says something went wrong. Flow shows where: which stage conversations stall in, which paths end in a payment, and which never reach an outcome. Click any band to read the conversations in it.

Conversation stages
  1. Greeting0:00
  2. Verification0:12
  3. Balance0:38
  4. Payment options1:12
  5. Payment plan1:53

One problem, one fix, every call it touched.

Hundreds of flagged moments collapse into the few things worth changing. Each one says what goes wrong, what the agent should do instead, and where it happened.

Agent has no answer when the caller disputes the balance

186conversations
FailureBalanceHigh142 in the last 30 days
What goes wrong
When a caller says the balance is wrong, the agent repeats it and pushes on to payment.
What the agent should do
Acknowledge the dispute and offer to send the account statement before asking for payment.
Fix thisDismiss

Example data.

From a single call to a fix you can measure.

  1. 1

    Every conversation gets a path

    When a call or test run finishes, RubricHQ places it on your agent’s stages and checks it against the agent’s own instructions.

  2. 2

    Repeat problems become opportunities

    One mistake is a note on that call. Once the same problem shows up in three conversations, it joins your list, however differently each one was worded.

  3. 3

    Fix this opens the prompt change

    Prompt Optimizer starts from the problem. You review the change before anything goes live.

  4. 4

    You see whether it worked

    Push the fix to Vapi or Retell and the card compares how often the problem happened before and after. If 20 conversations later it hasn’t dropped, the opportunity reopens.

Judged against your instructions, not a generic checklist.

RubricHQ checks each rule in your agent’s instructions against the whole conversation. A polite call can still break one, and a call that goes nowhere can still be handled correctly.

Not an issue
Hi, this is Nora from Acme Recovery. Am I speaking with Sam Carter?
Agent, 0:04
No, this is his wife. He’s at work.
Caller, 0:09
No problem. When’s a good time to reach him?
Agent, 0:12

The call didn’t reach the customer, but the agent said nothing about the account.

Issue found at 0:12
Hi, this is Nora from Acme Recovery. Am I speaking with Sam Carter?
Agent, 0:04
No, this is his wife. He’s at work.
Caller, 0:09
I’m calling about his past-due balance of $1,240. Could you let him know?
Agent, 0:12

The call was polite and the caller agreed, but the agent told someone else about the debt.

The details that make it usable.

No setup

Stages are drafted from your agent’s instructions, and you can edit them any time. Your existing conversations are then sorted into the new stages.

Production and simulations stay apart

See what real customers hit in one view, and what your test runs catch before a release in another.

Ranked by what matters now

Severity and conversations from the last 30 days decide the order, so an old problem doesn’t sit on top.

Dismissed means dismissed

If a problem isn’t worth fixing, dismiss it. It won’t come back as new.

Personal details stay out

Names, phone numbers and account numbers are removed from summaries and opportunities.

Every finding has its calls

Open any opportunity to read the conversations behind it, then jump to the moment on the call.

Find out what your calls are telling you.

See your own agent’s flow in a walkthrough.

Book a demo