← ALL PLAYBOOKS
THE AI-NATIVE GTM SERIES · PART 4 OF 4
✓ APPROVED

Close the loop so your GTM compounds

Episode 4, made runnable. You built a workflow, a context brain, and a signal library with an empty performance column. Close the loop: feed the outcomes back so the system gets smarter every week, instead of running once and forgetting.

[ SALES OPERATOR ]

Close the loop so your GTM compounds

G / Groundthe problem that started all of it
GTM

Episode 4, made runnable.

You built a workflow, a context brain, and a signal library with an empty performance column. Close the loop: feed the outcomes back so the system gets smarter every week, instead of running once and forgetting.

A / Assignthe build, block by block, and who owns each one. Open a step to see it run.
HUMAN + AI, IN THE LOOP
The skills run the work. You stay in the loop on the calls that need judgment.
Measure what happened
Salesforce01
Build funnel metrics the meeting will actually trust.
Details
Pin the metrics and definitions everyone trusts, so the loop runs on real numbers.
Powers up with
AI
Feed it back
Salesforce02
Find what your wins, losses, and churn share.
Details
Feed outcomes back: which signals actually paid, and what your wins, losses, and churn share.
Powers up with
AI
Let it improve
Salesforce03
Watch your analysts, and turn every miss into a lesson.
Details
Watch the system for drift and blind spots, with a weekly digest.
Powers up with
AI
Claude04
Turn feedback and outcomes into proposed skill fixes you approve.
Details
Let the system propose changes to your ICP, scoring, and skills. You approve, nothing self-edits.
Connect to analyze
AI + human
T / Tie backthe result it drove, and how they know
RESULT · 01

The 7th analyst

The judgment layer that makes the system compound.

OUTPUT · 01

Numbers everyone trusts

OUTPUT · 02

Outcomes fed back into your ICP

OUTPUT · 03

Signal library

OUTPUT · 04

A QA layer watching for drift

The Sales Operator
01 / WHY RUN IT

Your GTM runs once and forgets. The campaign ends, the numbers sit in a dashboard, and nothing feeds back into the next play. You repeat the same misses, and the context brain and signal library you built slowly go stale.

RUN THIS AND YOU CAN
Pin the metrics and definitions everyone trusts
Feed outcomes back into your ICP and signal library
Add a QA layer that catches drift before it ships
Let the system propose its own improvements, you approve
02 / THE RUN, STEP BY STEP

The workflow that solves it, one step at a time.

Each step is the plain question you are already asking. The skills answer them in order, one handing its receipt to the next.

1Measure what happened✓ TESTED
Build funnel metrics the meeting will actually trust.

Conversion by stage, velocity, and win rate with every definition pinned, and the leak named.

YOU GETStage conversion, time-in-stage, win rate, and a definition ledger.
OPEN THE SKILL →
2Feed it back✓ TESTED
Find what your wins, losses, and churn share.

Reads closed-won, closed-lost, and churned accounts, then hands back a ranked list of lookalike accounts to work.

YOU GETThe won / lost / churn pattern logic and the lookalike ranking.
OPEN THE SKILL →
3Let it improve✓ TESTED
Watch your analysts, and turn every miss into a lesson.

The check on your other analysts: it surfaces scoring blind spots, CRM hygiene gaps, and workflow drift, then hands back a weekly QA digest.

YOU GETThe six-part audit, the quality gates, and a weekly QA digest example.
OPEN THE SKILL →
4Let it improve✓ TESTED
Turn feedback and outcomes into proposed skill fixes you approve.

Reads QA findings, feedback, and outcomes, then proposes concrete changes to your skills and scoring. You approve, nothing self-edits.

YOU GETThe feedback streams, the four change types, and an approval gate on every merge.
OPEN THE SKILL →
WHERE YOU LAND

Numbers everyone trusts, outcomes fed back into your ICP and signal library, a QA layer watching for drift, and a system that proposes its own improvements for you to approve.

EVERY STEP, TESTED AND PASSED
EVAL RESULTPASS4/4 dimensions
HOW WE EVAL →
Skills resolve
Structure + proof
Example chat
Quality judge
  • All 4 chained skills resolve to real, scored skills.
  • Has a problem, an outcome, 2+ steps, and proof from a real build.
  • Ships an example chat showing the play run in Claude.
  • Reads in plain operator voice, grounded in a real build.
Last eval run 2026-09-12
DEEP EVAL · PASS
THE STANDARD →

We did not just check the structure. A judge read this play against the full text of every skill it chains (composition), then ran it on a real scenario end to end (simulated). It passed both. Test scenario: A Head of RevOps ran last month's signal-driven campaigns from Episode 3 and does not want the results to sit in a dashboard nobody revisits.

03 / WHO ELSE HAS TO MOVE

This play runs across the org, not inside sales.

Every function below owns a piece of the play. Skip one and it stalls where most GTM plays stall: in the handoff.

REVOPS
Pinned definitions and the reporting layer under them

Without it: The loop feeds back numbers nobody trusts.

SALES LEADERSHIP
Honest closed-lost reasons on every dead deal

Without it: The patterns learn from fiction.

CS
Churn and renewal outcomes fed into the pattern read

Without it: The loop only learns from the front of the funnel.

FINANCE
An independent check that the pinned numbers tie out

Without it: One bad definition compounds every quarter.

04 / SEE IT RUN

Watch the skills answer in order.

Describe your situation in plain language. The right skill fires on its own and hands its receipt to the next.

EXAMPLE CHATRavi, Head of RevOps, running the loop in one sitting
R
We ran the signal campaigns from last month. Now what? I do not want this to just be a dashboard we glance at and forget.
M
Measure what happened· Funnel metrics
First, numbers everyone trusts. I pinned the definitions (what counts as a meeting, an SQL, a win), then pulled conversion by stage for the campaign. No more arguing about whose number is right.
Receipt
One agreed scorecard: 6% of touched accounts booked, 28% of those to pipeline, every definition pinned.
R
Okay, so which of my signals actually drove that? I have been guessing.
F
Feed it back· Pattern analyst
Read the wins, losses, and churn from the campaign and fed it back into the repo. The Series B plus RevOps-hire combo converted at 3x; the standalone intent signal converted at nothing. I filled the signal performance log and flagged an ICP pattern: your best wins skew 200 to 500 employees, not 500-plus.
Best signalB + RevOps hire, 3x
Cutstandalone intent
ICP shift200 to 500 emp
Receipt
Signal performance log filled, one signal marked to cut, and an ICP-evolution entry proposed from the win pattern.
R
That is gold. How do I make sure the system keeps catching this without me babysitting it?
L
Let it improve· QA agent
Put a QA layer on it: it watches for scoring drift, CRM hygiene gaps, and signals that stopped firing, and DMs you a weekly digest. It already caught one, a routing rule sending Tier 1 accounts to the wrong sequence.
Receipt
A weekly QA digest running, plus one drift caught: Tier-1 accounts were mis-routed.
R
And the changes themselves, do I have to make them all by hand?
L
Let it improve· Evolution agent
No. The system reads the outcomes and proposes the changes: drop the dead signal, raise the combo's weight, tighten the ICP to 200 to 500. It writes the diff; you approve or reject. Nothing edits itself.
Receipt, loop closed
Three improvements proposed and approved. Your GTM now compounds: every campaign makes the ICP, the signals, and the scoring sharper. It stopped running once and forgetting.
05 / SEE IT IN PRODUCTION

Before you run it, I ran it for real.

This workflow is genericized from a real build. Here is the number it moved, and what I actually did.

THE BUILD · AI-NATIVE THESIS
The 7th Analyst
0.775 AUC
a won account outranks a lost one, on the model the loop rebuilt
1.5x lift
94% of the model Warm-plus accounts are wins, vs a 62% base rate
7 versions
each caught a flaw the last one shipped, by the loop not a customer
0 of 60
lost accounts that reached the top tier, the top of the funnel stays clean
WHAT I DID
01
Give every analyst an audit trail. Each agent logs what it decided and why, every override, every missing field, every enrichment miss. The exhaust of the system becomes its memory.
02
Build the seventh analyst to read them. A meta-analyst reads the other six's audit trails and turns each mistake into a training signal: a rule tightened, a gap flagged, a pattern learned.
03
Close the loop. The corrections feed back into the system, so the next run is better than the last. That is the difference between a tool and a system that improves itself.

AI-native is not how many agents you have. It is whether one of them watches the others and turns their mistakes into improvements. Build the analyst that reads the audit trail. That is the one that compounds.

Dig into the full build
06 / ROLL IT OUT AND INSPECT IT · THE OTHER HALF

The system compounds only if someone closes the loop

The skills hand the team pinned metrics, outcome patterns, a drift watch, and proposed improvements; the compounding comes from a human approving the right changes every week. Here is how a GTM leader runs the feedback cadence so the system gets smarter instead of just older.

THE FRAMEWORK BEHIND IT
The predicted-versus-actual review

Every week, the loop owner reads what the system expected against what actually happened: signals that paid, deals that lied, definitions that bent. Approved changes ship that week; everything else is noted and left alone.

  1. 01
    Teach outcome hygiene first

    The loop is only as honest as the closed-lost reasons and churn notes feeding pattern-analyst. Train the team to record what actually happened, not the polite version.

  2. 02
    Hold the weekly review hour

    Same hour every week: the pinned funnel-metrics, the qa-agent digest, and evolution-agent's proposals in one sitting. A loop reviewed monthly is a report, not a loop.

  3. 03
    Rehearse the approve-or-decline call

    Practice judging an evolution-agent proposal out loud: what evidence backs it, what breaks if it is wrong, how it gets reversed. Nothing self-edits, so the judgment is the safety.

  4. 04
    Inspect the loop's own record

    Quarterly, check which approved changes actually improved conversion and which got quietly reverted. The loop gets audited like everything it audits.

07 / INSTALL THE WHOLE PLAYBOOK

One plugin. 4 skills, bundled.

The whole playbook installs as a single plugin in Claude Code, no copy-pasting 4 times. Or grab any single step above. It runs on what you paste; connect your stack to go live.

OR KICK IT OFF IN CHAT
Here is my situation: [describe your workflow or account in a sentence or two]. Walk me through Close the loop so your GTM compounds step by step, and run each skill as we go.
GET THE NEXT PLAYBOOK

One runnable workflow a week. Free.