Go to market

The rollout

What it would take to get from this prototype to something a whole agent network uses: what it is called, what has to change to run at scale, how it is tested on the way, and the order it reaches people.

Status: a plan. None of this has happened, and the business would reshape it.

The name

First Call. Named for the moment it exists to protect: the first call a new agent makes on a lead they paid for.

It is a working name. If it shipped, it would take whatever name fits the platform it lives inside, and agents would reach it from there, not from a separate site.

Who it is for

First: New agents in their first weeks, before they work a paid lead.
Then: The trainers and team leaders who are responsible for getting them productive.
Later: Experienced agents picking up a new product or lead type.

What changes to run at scale

The prototype proves the interaction. These are the parts that are deliberately missing from it.

Part	Prototype today	At scale
Sign-in	Anyone with the link.	The agent's existing login, so every call belongs to an agent, a team and an onboarding class.
Storage	Nothing is kept.	Transcripts and scores are kept with a stated retention period, so progress and cohort comparisons exist.
Leads	Three invented people, written by me.	A library written with the best sales coaches, one set per lead type, versioned so a change is traceable.
Rubric	One rubric, six items, my weights.	Owned by the training team, reviewed by compliance, one per product line, versioned like the leads.
Voice	Typed, with optional browser dictation.	A spoken call, because the skill being trained is speaking, not typing.
Grader checks	Quotes verified, total computed in code.	Plus a fixed set of coach-graded calls that is re-run on every prompt or model change. A change that moves those scores does not ship.
Cost	Unmeasured.	Cost per practice call measured in the pilot and reported next to the result, with a cheaper model on the lead and the stronger one on grading if quality holds.

How it is tested

Three gates, in order. Each one is cheap compared with the one after it, and failing any of them stops the work.

Gate 1. Can the grader be trusted?
With
One or two experienced coaches
Test
They grade a sample of calls blind. Their scores are compared with the AI grader item by item.
Pass
Agreement at a level the training team sets in advance. Until then no agent sees a score.
Gate 2. Will agents use it?
With
A handful of new agents and one trainer
Test
They use it for a week with someone watching. Do they finish calls, do they come back, where do they get stuck.
Pass
Most finish the calls they start and the trainer would assign it. If not, fix the product before measuring it.
Gate 3. Does it move the number?
With
Two matched cohorts from one onboarding class
Test
The 30-day pilot on the experiment page, with the decision rule agreed beforehand.
Pass
Advance, redirect or stop, by the thresholds set before it started.

How it rolls out

Phase 1
One class, one lead type
The pilot itself. Small enough to watch every call and fix things the same day.
Phase 2
Every new agent
A step in onboarding, completed before the first paid leads. Team leaders see who has practised and which rubric item their team is weakest on.
Phase 3
Experienced agents
New lead types and new products get their own leads and rubric, so an agent can rehearse before selling something for the first time.
Phase 4
Partner agencies
Offered to other agencies as part of the software they already license, with their own leads and rubric.

Launch target

Announce Phase 2 at the next Engage conference.

Engage is where thousands of agents and agency owners are in one room for training and product demos, so it is the natural place to put this in front of the whole network: a live practice call on stage, then every new agent gets it the following week.

Working back from that date sets the schedule. The grader check, the usability week and the 30-day pilot all have to finish, and the pilot has to clear its decision rule, with time left to harden it. If the pilot says stop, nothing is announced.

Each phase starts only if the one before it held up. Phase 2 is where it stops being an experiment and becomes the product team's to run.

What gets handed over

When the pilot clears its decision rule, the product team receives a case, not a demo.

Working code and the lead and rubric library
The coach-graded call set used to check the grader
The pilot result against the agreed decision rule
Measured cost per practice call
The open risks below, with what was tried

What could go wrong

Agents learn to please the grader, not the lead.: Rotate leads, and keep checking scores against real appointment rates, not against themselves.
Coach mode becomes a crutch.: Coached calls are labelled and never count toward a required score.
The grader drifts when the model changes.: The fixed coach-graded set is re-run before any change ships.
A suggested line turns out to be non-compliant.: Compliance reviews the rubric and the coach instructions before agents see either.