Mnemos AI
AI Interview Agent

The interviewer that never gets tired.

A voice-led agent that sits down with every employee and asks the questions you wish your best operator would ask. It probes for edge cases, captures the why, and turns spoken context into structured organizational memory.

Session with Daniela R.
Senior AR Analyst · 17:42 elapsed
Recording
Interviewer

You mentioned customers occasionally pay by ACH and short the invoice. Walk me through what triggers a short-pay review.

Daniela

Anything more than $250 under invoice gets flagged. We open a ticket in NetSuite and check the contract first because some accounts have rebate clauses that look like short-pays but aren't.

Interviewer

Which contracts have those rebate clauses today?

Auto-paused at 18:00 for a follow-up
3 SOPs in draft
How it works

A patient operator on every workflow.

The agent runs on a tuned interview policy. It picks topics from the graph, asks role-aware questions, and knows when to stop.

Role-aware question trees

Tenure, scope, and prior coverage shape every question. A new manager gets a different session than the operator who has run the same close for six years.

Probes for the edge cases

When an answer is general, the agent asks for an example. When the example is happy-path, it asks what breaks. The why is captured, not just the what.

Background mode

The agent can listen during live work — a pricing review, an incident, a customer call — and quietly stitch what it learns into the memory graph.

Sample transcript

Edge cases are where the value hides.

Interviewer

When you say a renewal is at risk, what signals do you actually look at?

Maya, CSM

Three things, in order. License utilization under 40% in the last 60 days. Champion role change in the last 90. And whether they have an active integration that broke and didn't get reopened.

Interviewer

What if utilization is high but the champion changed?

Maya, CSM

Higher risk than the inverse. We've lost three accounts in the last year where utilization was healthy but the original buyer left. Lesson — usage is a lagging indicator on enterprise.

Interviewer

What action do you take when both fire?

Maya, CSM

Schedule a 45-day exec sync, loop the AE, and re-validate the original business case with the new champion within two weeks.

Extracted artifacts
  • SOP: Renewal risk triage (revised)
  • Heuristic: Champion change outweighs utilization
  • Workflow: 45-day exec sync on dual-signal
  • Entity link: Maya → 12 mid-market accounts
Agent flow

From the first hello to the final SOP.

Step 1

Listen

The agent opens with the employee's stated owned scope from the graph.

Step 2

Probe

It asks for specifics, examples, and edge cases until the topic settles.

Step 3

Verify

It reflects back what it heard and confirms each procedure before extraction.

Step 4

Route

Drafts are sent to the right reviewer based on the entity ownership graph.

Capture before you lose it.

Pilot the interview agent with a single department in under an hour. No instrumentation, no calendar invites, no extra meeting.