APV Framework V1.0

The discipline of knowing
which AI agents to build

The Agentic Process Value (APV) Methodology is the first structured, repeatable methodology for determining whether an AI agent should be built — and if so, how to make it profitable.

5
Phases
From strategic discovery to operational handoff, each ending with a Profitability Gate.
5
Feasibility Axes
The Five-Dimensional Feasibility Model — score, prioritise, decide.
1
Governing Principle
The Profitability Gate. No phase is exited without measurable evidence of business value.
01 — Introduction

Why most agent initiatives fail before they start

The enterprise AI market is flooded with agent-building toolkits, multi-agent frameworks, and model marketplaces. What is conspicuously absent is a structured, repeatable methodology for determining whether an agent should be built, how it should be designed within its business context, and what architecture will make it reliable, compliant, and profitable.

Most organisations begin their agent journey by asking the wrong question: "What can we build?" The right question is "What process, if transformed by an intelligent agent, would generate disproportionate economic value?" The Agentic Process Value (APV) methodology inverts this. Every phase begins with a business question and ends with a gate that demands measurable evidence.

"Creating agents is one thing. Making them relevant, efficient, and profitable is another."— APV Framework

Ecosystem-Agnostic

Works with any LLM provider, any agent framework, any deployment target. Technology selection is deferred until the business case is sound.

Deliverable-Driven

Every phase produces a concrete artefact — scorecard, BPMN diagram, architecture blueprint, governance checklist — not another slide deck.

Profitability-Gated

A gate failure in Phase I is not a failure of the methodology — it is the methodology working, preventing investment in an agent that should never have been built.

61%
Of marketers report biggest disruption in 20 years
17%
Of organisations have deployed AI agents in production
83%
Of agent POCs fail to reach production
~0
Structured methodologies for agent feasibility exist
02 — The Framework

Five phases. Five gates. One question per phase.

Sequential in logic, iterative in practice. Findings in Phase IV may force a return to Phase II. Each gate demands evidence, not PowerPoint.

I

Strategic Opportunity
Mapping

Score every process candidate across five feasibility axes. Identify where the economics of agentification are defensible.

Gate: Candidacy Verdict
II

Process
Modeling

Map the process in BPMN. Identify automation boundaries, exception paths, and human escalation points.

Gate: Agent Scope Defined
III

Agent Architecture
Design

Tool use, MCPs, HITL orchestration, model selection, trust tiers, and guardrail configuration.

Gate: Architecture Frozen
IV

Governance &
Audit

Guardrails, compliance checks, monitoring, drift detection, audit trail specification.

Gate: Compliance Sign-off
V

Operational
Handoff

Deployment, alerting, performance baselines, escalation procedures, and ongoing monitoring.

Gate: Production Go-Live
03 — The Scoring Model

The Five-Dimensional Feasibility Model

Every process candidate is scored on a 1–5 scale across five axes. A single axis below threshold disqualifies the candidate — the weakest-link principle of agent feasibility.

5

Decision Complexity SWEET SPOT

Does the process require nuanced judgment, pattern recognition, or contextual reasoning from multiple ambiguous inputs?

4

Input Variability

How much do inputs vary in structure, format, and content? Unstructured, multi-format inputs score higher.

5

Transaction Volume SWEET SPOT

What is the annual volume of process instances? Above 100,000/year scores highest.

3

Error Cost GUARDRAIL

What is the business impact of an incorrect output? Higher scores mean stricter guardrails required.

3

Latency Tolerance

What is the acceptable response time? Sub-second processes require different architecture than batch.

DECISION COMPLEXITY INPUT VARIABILITY TRANSACTION VOLUME ERROR COST LATENCY TOLERANCE 5 4 5 3 3
Candidate: "Claims Processing" · Score: 20/25 · PRIME CANDIDATE

Scoring heuristic: A score of 3 or below on Decision Complexity suggests traditional automation (RPA, rule engines). High Error Cost combined with high Decision Complexity signals a high-risk candidate requiring aggressive guardrails.

Sweet spot: DC ≥4 · IV ≥3 · TV ≥4 · EC ≤3 · LT ≥2

04 — Try It Live

Score a process in real time

Adjust the five sliders to see how the APV Feasibility Model determines whether an agent should be built — and why.

Decision Complexity4

Nuanced judgment, pattern recognition, contextual reasoning

Input Variability4

Unstructured, multi-format, domain-specific inputs

Transaction Volume5

High annual volume of process instances

Error Cost3

Regulatory or financial impact of errors

Latency Tolerance3

Acceptable response time window

DECISION INPUT VARIABILITY TRANSACTION VOLUME ERROR COST LATENCY TOLERANCE 4 4 5 3 3
19/25

PRIME CANDIDATE

High ROI potential. Build this agent.

Try the full methodology →
05 — Why APV

What makes APV different from other frameworks

While toolkits and platforms focus on how to build agents, APV focuses on whether to build them at all — and how to measure the outcome.

  • Business-firstNot "what can we build?" but "what process would generate disproportionate value?"
  • Profitability GateEvery phase ends with a measurable check. No evidence, no progression.
  • Ecosystem-agnosticWorks with OpenAI, Anthropic, Google, Meta, Mistral, DeepSeek — any LLM, any framework.
  • Deliverable-drivenEach phase produces a concrete artefact (scorecard, BPMN, checklist, report).
  • False-negative safeA gate failure is the methodology working. Better to abort early than waste millions.
  • Audit-readyGovernance Phase IV produces compliance evidence regulators recognise.
  • RepeatableUse the same scoring model across 15+ process candidates, compare objectively.
  • Designed for the middleBuilt for the architect between the C-suite and the engineers — speaks both languages.
06 — For Whom

Built for the architect
between the C-suite and the engineers

APV is not for developers writing agent code. It is for the person who must answer "why this agent?" before anyone asks "which LLM?"

Independent ConsultantClient Engagements

Independent Consultant

Deliver agent feasibility assessments that are structured, repeatable, and defensible. Differentiate your practice with a proprietary methodology — not another opinion deck.

Enterprise Architect / BAInternal Assessment

Enterprise Architect / BA

Your organisation is exploring agents. You need a framework to evaluate candidates objectively across business units, prioritise billions in investment, and justify decisions to the CFO.

Solopreneur (One-Person Company)Service Packaging

Solopreneur (One-Person Company)

You're building an AI-powered consulting practice. APV gives you the rigour of a McKinsey methodology — at 149 dollars. Package it, brand it, use it with unlimited clients.

07 — The Product

The APV Framework Toolkit

Everything you need to apply the APV framework tomorrow morning. Digital download. Pay once, use forever.

APV Instruction Guide

Complete step-by-step guide: how to score, assess, and present any process

PDF

APV Process Canvas

One-page A3 vector template — map any process through the 5 phases

SVG / AI / PDF

Governance Checklist

50-point audit: guardrails, compliance, monitoring, security, HITL

PDF

Client Report Template

Executive-ready report skeleton — 80% pre-written in C-level language

Docs / PDF

Three Case Studies

Claims, Mortgage, Onboarding — scored end to end with full methodology

PDF

APV Skill File

Drop into Claude Code, Cursor, or Hermes — your AI becomes an APV-certified assessor

MD

Client Presentation Deck

12 slides, C-level tone. Insert your scores, present immediately

PPT / PDF

APV Scoring Engine

Auto-scoring Google Sheets with candidacy matrix, profitability gate, portfolio heatmap

Sheets

APV Prompt Library

12 copy-paste prompts for any LLM — score, analyze, draft reports in seconds

PDF

Claude Project Config

Pre-configured Claude Project — paste instructions, upload files, start scoring

MD

ChatGPT Custom GPT Config

Pre-built Custom GPT — import instructions, enable Code Interpreter for ROI calcs

MD

Mistral / DeepSeek Config

System prompt optimized for Mistral Large and DeepSeek V4 — EU and APVC ready

MD

Standalone System Prompt

Universal system prompt for any LLM API — paste and deploy

MD
One-time purchase · Digital download

$149 USD

Lifetime updates · Commercial use · Immediate access

  • 13 files: scoring engine, templates, prompt library, and AI configs
  • Designed for repeatable use with unlimited clients
  • Ecosystem-agnostic — works with any LLM provider
  • Free updates when the methodology evolves
  • Commercial license included
Get the Toolkit — $149
stripe Pay by card · Delivered by email
08 — FAQ

Frequently asked questions

Everything you need to know about the APV framework and toolkit.

What is the APV Framework?

The Agentic Process Value (APV) Methodology is a five-phase framework for determining whether an AI agent should be built for a given business process, how it should be designed, and what architecture will make it reliable, compliant, and profitable. It is ecosystem-agnostic, deliverable-driven, and governed by the Profitability Gate.

Who is this for? Consultants, developers, or executives?

APV is designed for independent consultants advising clients on AI strategy, enterprise architects evaluating agent candidates within their organisation, and solopreneurs building their own AI-powered consulting practice. It bridges the gap between the C-suite and engineering teams.

How does the Five-Dimensional Feasibility Model work?

Each process candidate is scored on a 1–5 scale across five axes: Decision Complexity, Input Variability, Transaction Volume, Error Cost, and Latency Tolerance. A score below threshold on any single axis disqualifies the candidate. The sweet spot requires Decision Complexity ≥4, Input Variability ≥3, Transaction Volume ≥4, Error Cost ≤3, and Latency Tolerance ≥2.

What is the Profitability Gate?

The Profitability Gate is the governing principle of APV. No phase is exited without measurable evidence that the agent will generate business value. It prevents investment in agents that should never have been built.

What is included in the toolkit?

Thirteen files: the APV Scoring Engine (Google Sheets), the APV Process Canvas, the Governance Checklist, the Client Report Template, three Case Studies, the APV Skill File (load into Claude Code, Cursor, or Hermes), the Instruction Guide (PDF), the Prompt Library, the Client Presentation Deck, and four pre-configured AI platform settings (Claude, ChatGPT, Mistral/DeepSeek, and a standalone system prompt).

Can I use APV with any AI platform?

Yes. APV is ecosystem-agnostic by design. It works with any LLM provider (OpenAI, Anthropic, Google, Meta, Mistral, DeepSeek, xAI), any agent framework, and any deployment architecture.

Can I use the toolkit with my own clients?

Yes. The toolkit includes a commercial use license. You may apply the methodology with unlimited clients, produce reports, and present findings using the provided templates. No royalties, no per-client fees.