02 — Sales Twin
Deals that don't stall.
A live model of the revenue funnel — where deals stall, why they stall, and which action moves them — simulated per rep, segment and quarter.
North Star
Revenue per rep · Win rate
The stream
Automation → Capability → Value driver → Financial value
Automation
Automated deal inspection on every open opportunity
Capability
Risk visible before the forecast call
Value driver
Forecast accuracy
Financial value
Fewer slipped deals per quarter
Automation
Next-best-action delivered in the CRM
Capability
Reps act on the highest-yield move
Value driver
Sales productivity
Financial value
Higher revenue per rep
Automation
Quote and proposal assembly
Capability
Hours to minutes on paperwork
Value driver
Cycle time
Financial value
Faster cash conversion
Value system
Value ≠ North Star ≠ KPI ≠ driver ≠ process metric.
The translation layer that connects a corporate objective to something an operating team can actually move on Monday.
Value
Revenue conversion
Economic outcome
Incremental booked revenue per rep
North Star
Win rate × revenue per rep
Outcome KPIs
Driver metrics
Process metrics
Causal model
Value leakage
Where the value goes missing today.
Leakage is multiplicative. Every gate that stays leaky discounts everything built upstream of it.
Qualification
Effort spent on deals that never had a buyer.
Capacity
Mid-funnel
Single-threaded deals stall at legal or finance.
Slipped quarters
Pricing
Discount given where none was needed.
Margin
Forecast
Commit built on optimism, not behavior.
Planning error
Digital twin
Simulate the interventions before anyone funds them.
Instead of implementing ten recommendations and learning the result in six months, the twin stacks them first.
Win rate
Guardrails
- Discount floor held
- No automated outbound to net-new contacts
- Customer contact frequency cap
Constraints
- Headcount plan
- Territory rules
- Legal-approved terms library
Automation
What runs without asking.
Pipeline hygiene loop
Stage, close date and next step validated against real activity.
Deal Coach Agent
Forecast roll-up
Bottom-up probability from behavior, not rep optimism.
Forecast Agent
Proposal assembly
Pricing, terms and references drafted from approved content.
Next-Best-Action Agent
Deal Coach Agent
Diagnoses stalls and recommends the move that closes.
LLM + RAG on wins and playbooks
Forecast Agent
Produces and explains a defensible commit number.
Propensity + survival models
Next-Best-Action Agent
Drafts outreach, quotes and follow-ups in-flow.
SLM routing + LLM drafting
Integrations
Where it plugs into the domain.
Vendor-agnostic by design. The twin reads and writes through whatever stack the domain already runs on.
Systems of record
- CRM
- CPQ / billing
- Contract lifecycle management
Signals
- Conversation intelligence
- Email / calendar
- Product usage telemetry
Data & runtime
- Warehouse
- Metric layer
- MCP connectors
- Model gateway
Domain platforms we connect
CRM
CPQ, billing & contracts
Engagement & intelligence
Data & enrichment
Productivity
Data & runtime
Plus anything else with an API — connectors are added per engagement, not sold as a platform lock-in.
Value
The levers, and what they move.
Growth
Convert more of the same pipeline.
Win rate · Average deal size
Speed
Remove stage decay and admin drag.
Sales cycle days · Selling time %
Trust
Forecast the board can rely on.
Forecast variance
Risk
What could go wrong, and what stops it.
Every risk in this domain has a named containment in the runtime — not a slide.
Agent contacts a customer off-message
Approved content only; outbound requires human send for new logos
Medium
Scoring bias across segments or geographies
Segment-level fairness monitoring on propensity models
Medium
Confidential deal data in model context
Tenant-isolated retrieval, zero-retention inference
High
Compliance · Security
Built into the runtime, not bolted on.
EU AI Act
Limited risk — transparency; not permitted for employment decisions.
- Agent recommendations are advisory; no automated performance decisions on reps
- Explainability record for every scored deal
- Human accountability for all customer-facing output
GDPR
- Call recording consent enforced per jurisdiction
- Purpose limitation on contact data used for scoring
- Retention schedules applied to transcripts and embeddings
Security
- Field-level access mirroring CRM permissions
- Redaction of PII before model calls
- Regional data residency for EU accounts
Controls & oversight
- Discount and pricing guardrails
- Approved-content-only generation
- Immutable action log per opportunity
Agentic execution
Agents earn authority. They are not given it.
No agent in this twin controls anything it has not first proven in replay, evaluation, simulation and shadow.
01
Historical replay
Re-run the last 12 months. Would the agent have been right?
02
Offline evaluation
Scored against held-out outcomes, not opinion.
03
Digital twin
Simulated against the causal model under stress.
04
Shadow mode
Runs live, decides nothing. Divergence is logged.
05
Human recommendation
Proposes; a person executes and rates it.
06
Bounded pilot
One segment, capped exposure, hard rollback.
07
Human-supervised execution
Acts inside thresholds, humans approve exceptions.
08
Progressive autonomy
Authority widens only where evidence widened.
The unit of value
Don't buy transformation. Buy measurable movement.
This twin is contracted the way it is engineered: a baseline, a target, a guardrail, and an attribution method agreed in advance.
Baseline
Win rate 18%, revenue per rep $1.4M
Target
Win rate 24%+ within three quarters
Guardrail
Average discount does not increase
Proof / attribution
Cohort comparison by rep and segment, CRM-audited
Next twin
Financial Twin →