Sales AI v3 — Mission Control

Stage 1 — Behaviour Genome

Pattern Genome Trigger × Response Coverage Simulator Account Profiles

Stage 2 — Market & Simulation

Market Model Simulation Preview

Stage 3 — Expert Baseline

Expert vs AI Score

Stage 4 — Monte Carlo

Stage 5 — RL Agent

Stage 6 — Visualization

Sales AI v3 — Mission Control

Stage 4 active · Monte Carlo complete ✓ · Stages 1–3 complete ✓

Current stage

3 / 6

Expert Baseline

Gates passed

10 / 10

Stages 1 & 2 ✓

Patterns defined

60

SP30 + BP30 ✓

Client archetypes

18

8 industries ✓

Behaviour Pattern Library

+ Add pattern

↓ Save to dataset

Reset

60 patternsSP30 seller · BP30 buyer0 confirmed

ID	Name (RU)	Behaviour type	Trigger / Condition	Action	Status

English name/label>

—

Behaviour type/label>

—

When this fires/label>

—

Recommended action/label>

—

Relationship impact/label>

—

Explanation template (decision rationale)/label>

—

Trigger × Response Matrix

↓ Save to dataset

Reset to defaults

20 triggers12 response types— active links

Strength:

None

Weak

Medium

Strong

Click cell to cycle strength

Coverage & Balance Check

Gate: waiting

— coverage— balance

Simulation days 1000

Account count 15

Seed 42

Pattern activations (top 20) Balance gate

Run coverage check to see results

Coverage — all 60 patterns Coverage gate

SP triggered BP triggered

—

Account Profiles

+ Add archetype

— accounts— industries

Card view

Relationship editor

Market Model — LME Aluminium Price

Recalculate

LME — USD/t— market shocks

Parameters

LME base price (USD/t)

2450

Price trend (% per month)

+0.3%

Noise σ (% daily)

0.8%

Seasonality amplitude (%)

6.0%

Market shocks (count / 26 months)

3

Seed

42

Min—

Max—

Avg—

Volatility σ—

LME price — Jan 2024 to Mar 2026 (26-month simulation) Price Black swans

Account Timeline — Daily Decision Log

Run 100-day simulation

No account selected— events logged

Select account

Data access boundary

Run to verify

Event log —

Select an account and run simulation to see the decision log

Stage 2 gate checks

Reproducibility — same inputs produce identical results

Accounting identity — revenue − costs = margin (always)

Plausibility — 100-day account log reviewed and confirmed

Data access boundary — agent never sees hidden state

Edge cases — zero budget, inactive accounts, all silent

Margin range — values within realistic PLN range for aluminium B2B

Expert vs AI Scoreboard

Export CSV

Expert baseline: — AI system: pending Delta: —

Expert model parameters

Known seller patterns

20 / 30

Pattern miss rate

15%

Expert bias type

Accounts managed

15

Daily capacity (hours)

8h

Monthly budget (PLN)

25,000

Performance comparison — 26-month simulation Simulated · not real data

Method	Total margin (PLN)	vs Random	vs Expert	Accounts retained	Status

Cumulative margin over 26 months Random Expert AI system Upper bound

Outcome distribution (50 runs)

Statistical significance

Run scoreboard to see results

Monte Carlo Explorer

Complete ✓

🎲

Stage 4: Monte Carlo — Complete

245,700 records in replay buffer · 30 trajectories · CHURN 22.7% / STABLE 50% / EXPANSION 27.3% · All gates passed

Visualization coming in Stage 6

RL Agent

Planned

🔒

Stage 5: RL Agent

Belief state estimator · Policy network (PPO/DQN) · Training loop · LLM layer 7

Gate: unlocks after Stage 4 ✓

Results & Visualization

Planned

🔒

Stage 6: Visualization & Demo

AI vs Expert margin chart · Portfolio Σ · Belief state evolution · 5-min investor demo

Gate: unlocks after Stage 5 ✓