Unofficial. A job application for Cresta’s Forward Deployed Product Manager, AI Agent role, built from public pages. Not affiliated with Cresta.

The launch-readiness board.

The gate review from stage 4 of the plan, as a page you can use. You are the sponsor at Pinemoor Lending, a made-up lender. A voice agent is two days from launch. Open a use case, read the turn that failed, and decide what goes live.

Start here

Why I did thisYour posting says this person owns each AI agent and the results it drives. I wanted to show what owning one looks like on the hardest day: the day a customer decides whether it goes live.

What I was thinkingCresta already publishes how it tests an agent, and it is thorough. What I could not find is the page a customer’s executive looks at to say go or no-go. That decision is where trust is won or lost, so I built it.

What it showsA launch board for one AI phone agent. Each type of call is ready, on hold, or has no data. You can open any of them and read the exact conversation behind the status.

Try it in four steps

You play the executive sponsor at a made-up lender. The agent is two days from launch.

  1. Read “The gate” on the right. It says how many types of call are ready to go live.
  2. Open “Make a payment”. It is on hold, because one test the lender called a must-have failed.
  3. Click the failed test. You read the conversation and the moment it went wrong, then say whether the grader was right.
  4. Switch between Run 5 and Run 6, see what changed, then press “Record the decision”.

Why I am the person for this job. I run go or no-go decisions for a living, and my rule is that a gate with no data shows a dash, never green. I also build and test AI agents of my own, and I check their numbers myself. I have not worked in a contact center. I have learned a client’s operation from the inside many times, and this is how I start.

launch readiness · prototype Sample data. Made-up customer. Not a Cresta product.
Pinemoor Lending · account servicing by phoneA made-up lender. Seven use cases, 72 sample test conversations.

The rules this board follows

No dataA dash, never greenA line with no data cannot pass.
HoldA failed must-have holdsWhatever the pass rate.
GoDecided per use caseWhat passes can launch while the rest waits.

Test conversations by use case

Every scenario, count and transcript here is sample data written for this page. Click a use case to open it, then a scenario to read the conversation.

    This is a prototype. Nothing is stored or sent.

    What this replaces

    I do not know what Cresta shows a customer’s executives on decision day. Its public pages describe release criteria and thresholds, and the testing underneath is thorough.

    I found no public picture of the decision itself. If Cresta has one, this is my version of it, and I would use theirs.

    Choices I made in this page

    • No data shows a dash.A green light nobody checked is worse than no light. I carry this rule from the planning board I built for my current program.
    • A grader nobody has checked counts as no data.Nine passes out of nine still show a dash until the lender’s reviewers have checked the grader.
    • The failing turn is two clicks away.An executive should see what failed, not only that something did.
    • The decision is per use case.Four can launch while three wait, and the date holds.
    • A person’s entry is marked as a person’s entry.Every line says where its answer came from.