A
The Sales Mastery System A guided tour · No client data, no rubric content
The Sales Mastery Atlas · Methodology only

How Peer measures, scores, and progresses Sales Skills Mastery.

This page is a tour of the system — the assessment architecture, the progression model, the evidence base, and the certification pathway. It does not expose any frameworks, competencies, or rubric content. Those stay inside the engine, and they are the firm's protected IP.

01 · Assessment Architecture

Fourteen specialist AI agents, one calibrated pipeline.

Peer doesn't run a single LLM call against a transcript. It runs a coordinated pipeline of role-specialist agents, each owning a sub-domain of the work. Every observation produced is auditable back to the agent that produced it and the evidence it cited.

The five specialist assessors

Each assessor owns a function family. The router decides which one to invoke based on the framework selected for the meeting.

AGENT · sales_assessor
Sales Assessor
Opus 4.7 · Prompt-cached

Assesses against SALES, SDR and FOUNDER frameworks. Returns atomic observations with grade, polarity, evidence, and competency code.

AGENT · cs_assessor
Customer Success Assessor
Opus 4.7

CUSTOMER_SUCCESS frameworks. Retention check-ins, expansion conversations, QBR motions.

AGENT · support_assessor
Support Assessor
Opus 4.7

CUSTOMER_SUPPORT frameworks. L1/L2 response, incident triage, customer comms quality.

AGENT · management_assessor
Management Assessor
Opus 4.7

MANAGEMENT frameworks. 1:1s, performance conversations, coaching cycle execution.

AGENT · culture_assessor
Culture Assessor
Opus 4.7

Applied to every participant by default. CULTURE frameworks: psychological safety, feedback, accountability behaviours.

AGENT · assessor (fallback)
Generic Assessor
Opus 4.7

Fallback when the framework function isn't yet routed to a specialist. Same rubric structure, broader prompt.

The supporting agent fleet

Around the assessors sit specialist agents for routing, sanitisation, calibration, progression decisions, and narrative.

AGENT · call_purpose
Call Purpose Classifier
Haiku 4.5

Gates the pipeline. Decides whether the recorded meeting is in-scope for assessment (sales call vs. internal sync, etc).

AGENT · framework_router
Framework Router
Haiku 4.5

Multi-framework classifier. Picks the right framework(s) for the call from the org's installed set.

AGENT · sanitiser
PII Sanitiser
Haiku 4.5

Removes personal identifiers before anything is persisted or sampled. PII rules enforced inside assessor prompts as well.

AGENT · ai_dual_rater
Independent Dual-Rater
Sonnet 4.6 (deliberately different)

Blind dual-rating on a sample of assessments. Strips the original grade + evidence before showing the validator. Concordance auto-computed (exact / adjacent / divergent).

AGENT · progression_judge
Progression Judge
Opus 4.7 · 800 tokens

Owns tier promotion decisions. Reviews the evidence snapshot (counts, tier mix, trajectory, sample quotes) and emits promote / hold / not_yet_assessed.

AGENT · coaching_focus
Coaching Focus
Sonnet 4.6

Produces the "this-week focus" narrative for the rep. One mechanic, the why, the cited moment, the next move.

AGENT · coaching_why
Coaching Why-Master
Sonnet 4.6

Generates the career / role / results narrative explaining why mastery of a given competency matters to the rep specifically.

AGENT · peerai_org
PeerAI · Org
Sonnet 4.6

Conversational coach for the enablement leader. Operates over the org's full mastery picture under strict privacy + governance rules.

The end-to-end pipeline

Every observation runs through the same shape, regardless of which assessor handles it.

01 · INGEST
tl;dv webhook delivers the call
HMAC-verified · Realtime + poll fallback
02 · GATE
Call purpose classified
call_purpose agent · scope check
03 · ROUTE
Framework chosen + assessor selected
framework_router → specialist assessor
04 · ASSESS
Atomic observations produced
specialist assessor · prompt-cached · cited evidence
05 · REVIEW
Sampled dual-rating + stamping
ai_dual_rater · concordance recorded

The 10% calibration layer.

A blind independent dual-rater agent reviews a 1-in-10 sample of assessments — every tenth call, every framework, every assessor. The rater is given the transcript and rubric only; the original AI grade and evidence are stripped. Concordance is auto-computed and logged. Divergent ratings trigger calibration review of the assessor's prompt + rubric.

What the dual-rater sees

  • Transcript excerpt for the moment
  • Competency code + rubric criteria
  • Original AI grade hidden — blind rating
  • Original AI evidence hidden — independent reasoning

How concordance is recorded

  • Exact — same grade as the original assessor
  • Adjacent — within one tier (Foundation ↔ Proficient)
  • Divergent — two tiers apart or more (triggers review)
  • Aggregated to per-assessor + per-framework calibration metrics

All atomic observations are also spot-check stamped at 100% — reviewer metadata, human_verified flag, and review_status are attached to every record so any individual observation can be audited later.

02 · The Progression Model

Four-tier assessment, three-tier progression.

The assessment grade scale has four tiers — Foundation, Proficient, Advanced, Expert. The promotion engine currently promotes through three of them (Foundation → Proficient → Advanced). The Expert tier is recognised at the assessment level today and promotes into the formal progression pathway as the certification programme rolls out.

TIER 01 · ASSESSED + PROMOTABLE
Foundation

The rep performs the mechanic at all. The substrate is present. The floor below which there is no mastery to coach against.

Promotion gate Held until ≥3 observations + LLM-judged trajectory + tier-mix evidence.
TIER 02 · ASSESSED + PROMOTABLE
Proficient

The rep performs it consistently and links it to a measurable change in the buyer's state. The substrate methodology needs to land.

Promotion gate Trajectory upward + ≥3 observations at Proficient+ across multiple meetings.
TIER 03 · ASSESSED + PROMOTABLE
Advanced

The rep performs it reliably across motions, ahead of pricing, and recovers from disruption. Top progression tier today.

Top of progression Judge short-circuits further promotion at this tier (no auto-promotion to Expert).
TIER 04 · ASSESSED
Expert

The rep coaches the mechanic in peers — recognises it in others' calls and articulates what good looks like. The tier that produces tribal mastery.

Recognised in assessment · promotion via certification Expert is assigned by Peer Sales Mastery Certified programme review (see §05).

The Progression Judge agent.

Promotion isn't a hard percentage threshold. It's an evidence-based decision made by a dedicated Opus 4.7 agent that reviews a structured snapshot of the rep's recent observations and emits one of three verdicts.

What the judge sees

  • n_observations, n_distinct_calls, n_distinct_competencies
  • tier_mix (Foundation / Proficient / Advanced / Expert counts)
  • pct_at_proficient_plus (% ≥ Proficient)
  • trajectory (rising / flat / falling / insufficient_data)
  • days_since_first_observation
  • sample_recent_quotes (best moments)
  • sample_concerns (worst moments)

How the judge decides

  • Hard pre-check: < 3 observations → not_yet_assessed
  • Top tier rule: at Advanced → hold (no auto-Expert)
  • LLM verdict: promote / hold / not_yet_assessed + reasoning
  • Idempotency: one decision per (rep × framework) per 6 hours
  • Single-step: no tier skipping permitted
  • Audited: reasoning + evidence persisted per decision
03 · The Dependency Principle

Mastery flows along a graph. Coaching order matters.

Most competencies don't exist in isolation. Some are foundational — they need to be in place before downstream mechanics can be coached productively. The standard captures this as a dependency graph. The full graph stays inside the engine — what's shown below is the principle, not the map.

FOUNDATIONAL Skill A UPSTREAM Skill B UPSTREAM Skill C FOCUS Skill D UNLOCKED Skill E UNLOCKED Skill F Coaching Skill D before Skills A, B and C are in place is wasted hours. Once Skill D is mastered, Skills E and F become productive to coach.

What this means in operation. When the system surfaces the "one mechanic this week" for a rep, the recommendation is dependency-aware. It will never point a manager at a downstream skill while the upstream foundation is missing. The graph encodes what an experienced enablement leader knows tacitly — and applies it consistently at the scale of a sales floor.

04 · Evidence Base

What the research says about coaching, observation, and mastery.

The methodology is grounded in two decades of academic and industry research on deliberate practice, sales coaching effectiveness, and observable skill assessment. Specific citations below; full bibliography at the foot of this page.

Sales coaching consistency is strongly correlated with win rate, but most organisations report inconsistent coaching delivery.

CSO Insights / Korn Ferry's annual sales enablement studies consistently report that organisations with formalised, consistent coaching outperform those without — by meaningful margins in both win rate and quota attainment. They also report that fewer than half of frontline managers receive formal coaching training themselves.

SOURCECSO Insights / Korn Ferry · Annual State of Sales Enablement

Mastery comes from deliberate practice with observable feedback against a defined standard — not from time on task.

K. Anders Ericsson's foundational research (1993, restated in Peak, 2016) established that expert performance develops through deliberate practice — specific, feedback-rich, against an explicit standard for what good looks like. The "10,000 hours" popularisation was a misreading of this work; the active ingredient is the structure of the practice, not the volume.

SOURCEEricsson, K. A. et al. · "The Role of Deliberate Practice" · Psychological Review (1993) · Peak (2016)

Manager-rated coaching effectiveness has stayed roughly flat for a decade, while enablement spend has multiplied.

Industry tracker surveys (CSO Insights, ATD, SiriusDecisions/Forrester) consistently report manager-rated coaching effectiveness ratings in the 50–65% range across multiple consecutive years, against a backdrop of substantial growth in enablement budgets and tooling. The gap is what motivated the build.

SOURCECSO Insights · ATD · Forrester (SiriusDecisions) · multi-year industry surveys

B2B SaaS sales onboarding typically takes six to nine months to reach productivity — and has barely improved.

Industry research from Bridge Group, Sales Hacker, Salesforce State of Sales and others has reported median onboarding periods in the 6–9 month range for B2B SaaS AEs for the better part of a decade. Reducing it requires intervention at the layer where ramp actually happens — observable skill mastery — rather than methodology training.

SOURCEBridge Group SaaS AE Metrics · Sales Hacker · Salesforce State of Sales · industry medians

AI-augmented sales coaching is shifting from pilot to category. The methodology questions matter as much as the model.

Gartner's "Future of Sales" research has, for several years, positioned AI-augmented coaching as one of the highest-impact emerging categories in B2B sales technology. The differentiator across vendors is increasingly the methodology — what's being measured, by whom, and with what calibration — rather than raw model capability.

SOURCEGartner · "Future of Sales" / sales technology research

Observable skill assessment beats subjective evaluation across professions — from medicine to aviation to athletics.

A consistent finding across decades of competency-based assessment research is that observable, rubric-anchored evaluation produces more reliable judgments — and more useful development feedback — than holistic subjective evaluation. The shift to standardised observation is what professionalised medicine, aviation, and elite athletics.

SOURCEHBR · Academy of Management Review · competency-based assessment literature
05 · The Certification Pathway

Peer Sales Mastery Certified — three levels, portable for life.

Reps, managers, and enablement leaders earn formal credentials by demonstrating mastery against The Mastery Standard. Credentials live on the rep's personal PeerPal portfolio, survive employer changes, and are verifiable by future employers. Currently in build alongside the ThinkWork Enterprise tier.

IN BUILD
CERT 01 · REP

Peer Sales Mastery — Certified Rep

Rep-level mastery certification. Achieved when a rep reaches Proficient+ across the framework's required competency set and Expert on at least one competency.

REQUIREMENTSProficient+ on full required set · Expert on ≥1 competency · 90 days observed
IN BUILD
CERT 02 · MANAGER

Certified Peer Operator — Manager

Manager-level certification. Demonstrates capability to read Peer output, run a coaching cycle inside the standard, and make defensible tier-progression decisions independently.

REQUIREMENTSCapstone observation review · 1 full cycle run · ≥85% concordance with senior operator
IN BUILD
CERT 03 · LEADER

Certified Enablement Leader

Enablement-leader certification. Train-the-trainer credential from the ThinkWork Enterprise tier. Confers the right to certify internal operators within the leader's own org.

REQUIREMENTS8-week programme · 5 cohort workshops · multi-team operator review

PeerPal — the portable portfolio.

PeerPal is the personal career portfolio that holds a rep's mastery credentials. The PeerPal v1 scaffold is built — magic-link auth, multi-employer aggregation, only Proficient+ moments land in the portfolio. The certification badges and verifiable-credential signing layer roll out alongside the certification programme. The portfolio survives job changes — the credential travels with the rep, not the employer.

06 · The IP Boundary

What's shown, and what stays inside the engine.

The Mastery Standard's content — the frameworks themselves, the competency taxonomy, the tier rubrics, the assessor prompts — is the firm's protected IP. This page deliberately publishes the system, not the content. The line is explicit on purpose.

Shared publicly · This page

The system, the architecture, the principles.

  • The assessment architecture — which agents, what each does, which models they run
  • The 4-tier assessment scale and 3-tier progression model
  • The Progression Judge's evidence inputs and decision rules
  • The 10% calibration layer and how concordance is recorded
  • The dependency principle (concept, not the literal graph)
  • The certification pathway and what each credential requires
  • The evidence base from public research and industry studies
Stays inside the engine · Protected IP

The Standard's content. The rubric. The proprietary work.

  • Specific framework names beyond the small set referenced in the founder's letter
  • The complete competency taxonomy across all functions
  • Tier criteria rubrics — the written descriptions of what each tier looks like per competency
  • Assessor agent system prompts and few-shot examples
  • The dependency graph itself — actual upstream / downstream edges
  • The Progression Judge's calibration data and decision boundaries
  • The fifteen years of in-seat practice and source observations the standard was built from

The principle: teach people to read the system. Never teach them to be the system. The methodology is open because credibility lives there. The content is closed because the work lives there.