Skip to main content
INDEPENDENT RESEARCH · AGENTIC AI · ENTITY SCORINGTier 1 Flagship

Generic Entity Scoring Engine, Native to the CRM

The hard problem is not adding an agent. It is deciding which agent - and whether any agent - should act.

An independent agent-orchestration study: one deterministic scoring algorithm applied unchanged to any entity, routing to narrow specialists, with every mutation gated behind schemas.

ORCHESTRATORENTITY SCORINGTOOL CALLINGMULTI-AGENT

Executive BriefAgentic
Domain / ScopeAgentic Systems & AI Architecture
Architectural RoleIndependent AI Systems Researcher
Timeline & InitiativeIndependent research · 2025–2026
Key Verified Invariant

Unready fixtures stop before any model call, with a machine-readable reason

01 · The Constraint

What Made the Problem Difficult

Independent prototype, synthetic data. I built this because single-agent demos kept failing the same way in my experiments: one bloated prompt, thirty tools, and no notion of whether the record was even ready. The engine is deliberately generic - the same scoring algorithm runs unchanged across different fixture entities, so readiness means the same thing everywhere instead of being redefined per object. I wanted that boring machinery - scoring, routing, gating - treated as first-class architecture instead of prompt tricks.

Core Platform Constraint

Systemic Friction

What goes wrong without it: models get invoked on incomplete records, contexts bloat across domains, tools fire with unvalidated arguments, and nobody can reconstruct why an action happened. In record-centric work that is unacceptable.

Boundary Invariants & Operational Limits:
  • !No model call on unready records - deterministic triage first
  • !Every mutation behind schema validation and permission checks
  • !One specialist per task with an isolated context window
  • !Full reconstruction: scores, routes, and tool I/O all auditable
02 · The Architecture

What System Was Designed

Three stages, deliberately unglamorous: 1) Score - completeness, urgency, and risk computed deterministically over the fixture record. Below threshold, the orchestrator stops and says why. 2) Route - an intent classifier picks exactly one specialist and hands it a fresh, narrow context. 3) Act - the specialist may only call registered tools whose arguments validate against strict schemas; every call and result lands in an audit envelope.

03 · The Key Decision

Pivotal Architectural Choices

Key Architectural Decisions

ADR // 01DETERMINISTIC CHOICE

Score-first triage vs. immediate generation

Context:

Calling a model on a half-formed record wastes tokens and invites confident nonsense.

Alternatives Evaluated:
  • Prompt directly on the raw record
  • Hardcoded triggers with no model at all
Chosen: Deterministic scoring gate before any model invocation.

Cheap arithmetic filters most cases; expensive inference runs only where it can help. My logs showed fewer wasted calls and clearer refusals.

ADR // 02DETERMINISTIC CHOICE

Narrow specialists vs. one mega-prompt

Context:

One agent holding quoting, SLA, and billing logic degrades at all three.

Alternatives Evaluated:
  • Single prompt with 30+ tools
  • Uncoordinated agent swarm
Chosen: Central router with small, single-domain specialists.

Small contexts stay sharp. Routing decisions stay visible. Debugging stops being archaeology.

04 · The Engineering

What Was Actually Built

01

Built the scorer as pure functions over fixture records - no model involved, fully unit-testable thresholds.

02

Implemented the router as intent classification plus capability matching, with explicit abstention when nothing qualifies.

03

Registered every tool with a strict JSON schema; validation failures return structured errors, never partial writes.

04

Wrapped each run in an audit envelope: input hash, scores, route taken, tool calls, outputs, timestamps.

05

Gave each specialist a fresh context assembled from the envelope - never the full conversation history.

05 · The Trade-offs

Deliberate Architectural Compromises

Trade-offs & Mitigations

Routing hops vs. single-shot latency

Architectural Benefit:

Guardrails, visibility, and far fewer runaway actions.

Associated Cost:

Extra milliseconds per run for scoring and routing.

Mitigation Strategy:

Precomputed score vectors and parallel skill-eligibility checks keep the overhead small.

Schema gates vs. freeform tools

Architectural Benefit:

Invalid mutations die at the boundary with a reason attached.

Associated Cost:

Each capability needs a written schema and mapping.

Mitigation Strategy:

A small tool template keeps new capabilities to one file each.

06 · The Result

Verified Outcomes

✓ VERIFIED RESULT

Unready fixtures stop before any model call, with a machine-readable reason

✓ VERIFIED RESULT

Specialist contexts stay small and task-shaped across multi-step fixture runs

✓ VERIFIED RESULT

Schema gates catch malformed tool arguments before any state change

✓ VERIFIED RESULT

Every run reconstructs end to end from its audit envelope

07 · Architecture Blueprint

System Schematic & Data Flow

System Topology Blueprint
Score-First Orchestrator & Specialist Routing
INTERACTIVE SCHEMATIC
ROOT AGENT · CENTRAL GOVERNANCEEntity Scoring OrchestratorDeterministic triage & context boundary isolation01 · Entity ScoringUrgency, completeness, risk02 · Dynamic RoutingIntent & capability matcher03 · Context BoundaryIsolated prompt scopesSPECIALIZED AGENT RUNTIMES · NARROW TOOL BINDINGSAgent A: Quoting & PricingAgent B: Service EscalationAgent C: Validation & AuditGOVERNED SALESFORCE ENTERPRISE CAPABILITIESSchema-Enforced Tool Calling · Composite API Mutations · Audit Telemetry Envelopes
SCORE-FIRST TRIAGEEvaluates business entity readiness before any LLM execution or tool dispatch occurs.
ISOLATED CONTEXTSNarrow agent contexts prevent prompt degradation and cross-domain hallucinations.
STRICT SCHEMA GATESEvery tool skill verifies typed JSON payloads before touching Salesforce CRM records.

Triage before tokens: score, route to one specialist, and gate every tool call behind schemas.

Text alternative for screen readers: Architecture flow: Record Ingress (synthetic) to Orchestrator via State & Intent Intake; Orchestrator to Scoring & Routing via Readiness Triage; Scoring & Routing to Specialist (A/B/C) via Isolated Delegation; Specialist to Tool Registry via Schema-Gated Execution

08 · Why It Matters

Architectural Conclusion

Agents become trustworthy when the unglamorous parts - scoring, routing, gating - are engineered instead of prompted.

Key Lessons for Platform Scale:
  • Score before you generate; most enterprise records are not ready on arrival.
  • Small routed contexts outperform large shared ones in every run I measured.