Assessment Methodology & Scoring Principles
How we evaluate operational feasibility, enforce non-negotiable risk gates, and calculate net realizable economic returns. Completely open and auditable.
The Four Operational Evaluation Pillars
The assessment calculates a normalized score from 0 to 100 based on four balanced operational dimensions:
Pillar 1: Operational Feasibility
Evaluates whether workflow inputs are structured, predictable, and clean enough for an AI model to process reliably. Looks at the presence of written SOPs, document layout consistency, and digital accessibility versus physical paper or closed legacy databases.
Pillar 2: Economic Leverage & Value
Evaluates whether the task consumes enough human specialist time to justify pilot setup and maintenance overhead. Analyzes monthly task volume, cycle time per unit, backlog delays, high rework costs, and key-person operational risks.
Pillar 3: Verifiability & Human Control
Evaluates how rapidly and reliably a qualified specialist can audit the model's output before it executes. Fast spot-checking (< 30 seconds) scores highly; laborious re-reading reduces feasibility. Requires a dedicated triage queue for ambiguous cases.
Pillar 4: Safety & Risk Posture
Evaluates the confidentiality of input data (Public, Internal, Confidential, Regulated PII) and the worst-case financial, legal, or safety consequence of an undetected hallucination or extraction error.
The 4 Non-Negotiable Safety Override Gates
In real operations, an aggregate score of 82/100 is completely meaningless if a single error triggers a catastrophic regulatory fine or client lawsuit. We enforce four deterministic safety gates that override favorable mathematical scores:
High Downstream Consequence + Low Verifiability
Trigger: Consequence is "Significant Financial" or "Severe/Legal" AND verifiability is "Laborious" or "Unverifiable".
Operational Rationale: If an undetected AI error causes significant financial or legal harm and a human cannot easily spot it in seconds, automated or assisted execution creates an unacceptable business liability.
Regulated Data Without Isolated Infrastructure
Trigger: Input data contains Personally Identifiable Information (PII), banking credentials, or regulated medical records without dedicated enterprise-grade zero-retention infrastructure.
Operational Rationale: Ingesting protected customer or employee data into standard commercial AI tools risks severe regulatory fines (GDPR, HIPAA, CCPA) and breaches customer trust.
Zero Human Oversight in Production
Trigger: User selects "Unverifiable / Blind Execution" (expecting autonomous execution without human sign-off).
Operational Rationale: In practical small-business operations, AI is an assistant, not an autonomous agent. Workflows without active human oversight are inappropriate for pilot deployment.
Unstructured Process + Poor Data Quality
Trigger: Process consistency is "Unstructured / Ad-hoc" AND data accessibility is "Siloed / Physical".
Operational Rationale: AI cannot fix an undocumented, chaotic process with missing source records. Teams must write standard operating procedures and digitize source documents before introducing automation.
The Economic Calculation Model
We explicitly distinguish Gross Theoretical Hours from Net Realizable Financial Savings. Saving 5 minutes per day across 10 people rarely reduces payroll. Our model reflects operational reality:
The Core Mathematical Formulas
Current Hours = (Monthly Volume × Manual Minutes Per Unit) / 60
Assisted Hours = [(Volume × (AI Time + Human Audit Time)) + (Volume × Exception Rate × Exception Time)] / 60
Gross Hours Freed = Current Hours - Assisted Hours
Net Realizable Savings = (Gross Hours Freed × Realization Factor × Hourly Labor Rate) - Monthly Tool Costs
Industry Framework Alignment & References
Our operational assessment and pilot kit reference established industry standards as informational guidelines to promote disciplined governance:
NIST AI Risk Management Framework 1.0 (NIST SP 1270)
Our pilot charter and risk checklists reflect the four core NIST functions: Govern (pilot charter, kill thresholds), Map (process boundary template), Measure (baseline time studies, calibration logs), and Manage (human-in-the-loop review SOPs).
ISO/IEC 42001:2023 Principles
Informed by ISO/IEC 42001 international management standards regarding risk-based operational impact assessment, continuous human oversight, and data provenance verification.
Evaluate Your Own Workflow
Put this transparent scoring methodology to work on your team's repetitive tasks in less than 4 minutes.