# Agentic Prep Cookbook ## A GTM Methodology for the Transition from Human to Agent Visitors **Author:** John Williams **Organisation:** Sun Business Group / fxops.ai **Published:** March 2026 **Version:** 1.0 **Canonical URL:** https://sunbusinessgroup.com/resources/agentic-prep-cookbook.md **Series:** Working Intelligently with Agentic AI Tools — Article 2 **Related resource:** https://sunbusinessgroup.com/resources/token-intelligence.md --- ## Purpose of This Document This document describes the Agentic Prep Cookbook — a sequenced methodology for GTM teams preparing their content, workflows, and services for a mixed audience of human and agent visitors. It is structured for both human readers and agent access. It defines the framework, the deliverables at each stage, the selection criteria for prioritisation, and the operator programme for fractional executives and consultants who want to deliver this methodology as a client-facing service. --- ## The Core Problem This Methodology Solves Your website, content, pricing pages, case studies, and documentation are being read by agents — not just humans. Agents researching on behalf of buyers. Agents qualifying vendors on behalf of procurement teams. Agents comparing competitive alternatives without filling in a form or booking a call. Your conversion architecture was built for a human reader. Agents don't scroll. They parse. They extract structured meaning and return a synthesis to their human principal. If your content is not legible to agents, you are absent from that synthesis. That is not a future risk. It is a current condition. The Agentic Prep Cookbook addresses this in three sequenced stages: Crawl, Walk, Run. --- ## Framework Overview | Stage | Focus | Timeframe | Primary Output | |-------|-------|-----------|----------------| | **Crawl** | Content legibility for agents | 0–30 days | Structured Markdown resource library at `/resources/` | | **Walk** | Workflow automation and agent interaction | 30–60 days | One defined, deployed agent workflow with human-in-the-loop approval | | **Run** | Agent commerce readiness and outbound agent deployment | 60–90 days | ACP-compatible service definitions; first outbound agent workflow live | --- ## CRAWL: Make Existing Content Agent-Legible ### Objective Convert the most important GTM content into token-efficient, structurally clean Markdown. Publish it at a consistent, discoverable path. Establish access tracking to measure agent vs. human access ratios. ### The Five Core Documents Every Crawl sprint produces Markdown versions of five foundational documents: **1. Positioning Document** - What the organisation does - Who it serves (ICP definition) - Why it is differentiated - The primary problem it solves - Format: plain prose, structured headings, no marketing gloss **2. Services or Product Overview** - What is offered - At what scope and commitment level - What the buyer receives - Format: structured list with brief description per service; pricing if public **3. Pricing and Packaging** - Tier definitions (if applicable) - Price points or ranges - What is included at each level - Note: if pricing is not public, document the engagement model instead — how a buyer initiates, what the scoping process looks like, what determines price **4. Proof Document** - One case study or engagement summary - Problem → approach → outcome structure - Specific enough to be useful to an agent evaluating fit - Format: structured narrative, 400–600 words **5. FAQ / Objection-Handling Document** - The ten most common questions buyers ask - The five most common objections and responses - Format: Q&A pairs, cleanly labelled ### Publication Architecture Host all documents at a consistent path on your primary domain: ``` yourdomain.com/resources/positioning.md yourdomain.com/resources/services.md yourdomain.com/resources/pricing.md yourdomain.com/resources/proof.md yourdomain.com/resources/faq.md ``` Link to the `/resources/` index from your main site navigation or footer so crawlers and agents can discover the library. An index file (`resources/index.md`) listing all available documents with one-line descriptions accelerates agent discovery. ### Why Markdown - Token-efficient: Markdown costs a fraction of the equivalent HTML page to process - Structurally clean: no CSS, JavaScript, navigation chrome, cookie banners, or pop-up overhead - Reasoning-ready: agents extract meaning from structured text more reliably than from rendered layouts - Version-controllable: plain text integrates cleanly with any CMS, Git workflow, or static site ### Access Tracking Implement access tracking on the `/resources/` path to classify requests by: - **Referral source:** LinkedIn, X.com, Substack, direct, unknown - **Client type:** human browser, AI agent, web crawler, API client The ratio of agent access to human access is a leading indicator of document adoption as agent instructions. This ratio rises before human traffic metrics reflect any change. Reference implementation: `token_access_tracker.py` — deployed at Sun Business Group. See `token-intelligence.md` for methodology. ### Crawl Sprint Completion Criteria - [ ] Five core documents converted to Markdown and published at `/resources/` - [ ] Index file live at `yourdomain.com/resources/` - [ ] `/resources/` path linked from main site - [ ] Access tracking active and producing daily reports - [ ] Baseline agent/human access ratio recorded --- ## WALK: Identify Automated Access Routes and Select a Candidate Workflow ### Objective Map the automated access points that already exist around your GTM motion. Select one workflow to formalise, instrument, and deploy with agent execution and human-in-the-loop approval. ### The Access Audit Before selecting a workflow, answer these four questions: **1. Which content do agents most frequently access?** Your Crawl-phase tracker answers this. The most-accessed documents signal what agents are researching. That tells you where the automated interaction is already happening. **2. Which questions do you answer repeatedly in sales discovery?** These are candidates for pre-answering via structured documents or agent-assisted qualification summaries before the first human conversation. **3. Which qualification signals does your team currently gather manually?** Company size, tech stack, recent hires, funding events, content published — any signal that could be extracted programmatically is a candidate for agent automation. **4. Which handoff points in your funnel involve exchanging data?** Form submissions, inbound emails, meeting requests, post-call follow-ups — wherever data moves between systems or between people, an agent can route more efficiently. ### H1-2026 Candidate Workflow Selection Criteria Select the workflow that satisfies this criterion: **highest human time cost per unit of complexity**. The ideal candidate is a task that is: - Repeated frequently (multiple times per week) - Currently performed by a human - Low in judgement requirement (rule-following, not decision-making) - Producible as a defined output (a document, a classification, a draft) ### The Three Strongest Candidate Workflows for GTM Teams **Candidate A: Pre-Call Qualification Brief** *What it is:* Before every discovery call, an agent researches the prospect — website, LinkedIn, recent press, job postings, tech stack signals — and produces a structured brief against your ICP definition. *Inputs:* Prospect name, company, URL, ICP definition document *Output:* Structured brief — company overview, ICP fit assessment, recommended opening questions, flagged risks *Human role:* Reviews brief before call; approves or adjusts *Estimated time saved:* 20–40 minutes per call **Candidate B: Content Repurposing and Distribution Routing** *What it is:* A long-form content piece (Substack article, LinkedIn post, case study) is submitted to an agent that produces derivative assets for each distribution channel. *Inputs:* Source document, voice skill instructions, distribution channel list *Output:* LinkedIn article draft, LinkedIn feed post (hook variant), 3× X.com post variants, email follow-up draft *Human role:* Reviews and approves each asset before publishing *Estimated time saved:* 60–90 minutes per content piece **Candidate C: Inbound Triage and First-Response Drafting** *What it is:* An inbound inquiry (form submission, email, DM) is processed by an agent that classifies intent, checks ICP fit, and drafts a first response. *Inputs:* Inbound message text, ICP definition, service descriptions, response templates *Output:* Intent classification, ICP fit score (high / medium / low), recommended action, draft first response *Human role:* Reviews classification and draft; approves or rewrites before sending *Estimated time saved:* 15–30 minutes per inbound inquiry ### Walk Sprint Completion Criteria - [ ] Access audit completed; four questions answered - [ ] Candidate workflow selected against the selection criterion - [ ] Workflow defined: inputs, outputs, permitted action space, escalation conditions - [ ] Human-in-the-loop approval gate documented - [ ] Agent workflow deployed and running - [ ] First 10 executions reviewed; output quality validated --- ## RUN: Define Services for Agent Commerce and Deploy Outbound Agents The Run phase has two movements. They can be executed in parallel or sequentially. --- ### Movement 1: Expose Services for Agent Commerce via ACP **What ACP Is** The Agent Commerce Protocol (ACP) is an open specification that enables agents to autonomously discover, evaluate, and transact with proprietary data sources and services. It provides the infrastructure layer for agent-to-agent and agent-to-service interaction. For a GTM team or consulting practice, ACP readiness means defining your services in a form that an agent can evaluate programmatically — not just a human reading a webpage. **The Four ACP Readiness Questions** **1. What is your service, expressed as a structured capability?** For each service you want to expose to agent commerce, document: ``` Service name: [plain text name] Description: [one sentence — what it does, not what it is called] Inputs accepted: [list of required inputs with types] Outputs produced: [list of deliverables with formats] Typical latency: [time from engagement to first output] Pricing: [fixed / retainer / project — with price if public] Authentication: [how a buyer initiates — Calendly, email, form] Scope limits: [what this service does not include] ``` **2. What guardrails govern agent interaction with your service?** Document the boundaries within which an agent can interact with your service on behalf of a buyer: - What decisions can an agent make without human approval? - What actions require human confirmation before proceeding? - What data can be exchanged automatically vs. what requires a human handshake? - What is the escalation path when an agent encounters ambiguity? This is not only a security question. It is a trust question. Buyers' agents will assess your guardrail documentation as part of vendor evaluation. **3. What does a successful agent transaction look like?** Define the output format for each agent-accessible service: - A structured report (specify fields and format) - A calendar booking (specify booking system and confirmation format) - A qualified proposal (specify sections and delivery method) - A data payload (specify schema) The cleaner and more predictable the output format, the more useful the service is to a buyer's agent — and therefore to the human who sent it. **4. How do agents discover your services?** Your structured Markdown content at `/resources/` is the discovery layer. Add a `services.md` document that lists your ACP-compatible services with their structured capability definitions. Reference it in your index file. For Sun Business Group, the natural ACP entry points are: - Guided Discovery sprint (defined inputs, defined outputs, fixed sprint structure) - AI Governance Enablement stages (defined assessment with documented deliverables) - AI Executive Coaching series (3-session, 60-minute, $1,500 — individual buyer, defined scope) **Run Sprint — Movement 1 Completion Criteria** - [ ] Each primary service documented with the four-field structured capability definition - [ ] Guardrail architecture documented per service - [ ] Output formats defined and templated - [ ] `services.md` published at `yourdomain.com/resources/services.md` - [ ] Index file updated to reference `services.md` --- ### Movement 2: Deploy Outbound Agents Using ACP Guardrails **The Governance Principle** Deploying an agent on an external workflow without guardrails creates incidents. The agent does not know when to stop. It does not know which actions require human approval. It cannot handle ambiguity in a way that protects your relationships or your reputation. Deploying an agent with ACP guardrails means defining — before the agent starts: - The assignment (what the agent is trying to accomplish) - The permitted action space (what the agent is allowed to do) - The escalation conditions (what triggers a human handoff) - The output format (what the agent produces when it completes) The guardrail is not a constraint on capability. It is what makes the capability trustworthy enough to actually use. **Three Practical Outbound Agent Deployments for GTM Operators in H1-2026** **Deployment A: Account Signal Monitor** *Assignment:* Monitor a defined list of target accounts for signal events. Produce a weekly brief. *Permitted action space:* Read publicly available web content, LinkedIn posts, press releases, job postings for defined accounts *Escalation conditions:* Signal requires human interpretation; account has existing relationship that changes response protocol *Output:* Weekly brief — account name, signal type, signal date, recommended engagement action, draft outreach (optional) *Human role:* Reviews brief; approves or deprioritises recommended actions **Deployment B: Competitive Research Executor** *Assignment:* Before each discovery call, execute the pre-call qualification brief workflow (from Walk phase) automatically, without manual trigger. *Permitted action space:* Research defined sources against prospect identity provided by CRM trigger *Escalation conditions:* Prospect is a current client; prospect has flagged privacy preferences; research returns ambiguous ICP fit *Output:* Structured qualification brief delivered to assigned human 2 hours before scheduled call *Human role:* Reviews brief; proceeds to call **Deployment C: Inbound Router** *Assignment:* On receipt of each inbound inquiry, classify, assess, and draft — without waiting for human triage. *Permitted action space:* Read inbound message; access ICP definition and service descriptions; produce classification and draft *Escalation conditions:* Inbound is from existing client (route to account owner directly); inbound contains sensitive or legal language *Output:* Classification, ICP fit, recommended action, draft response — delivered to assigned human within 15 minutes of inbound receipt *Human role:* Reviews output; approves or rewrites before any external communication **Run Sprint — Movement 2 Completion Criteria** - [ ] First outbound deployment selected - [ ] Assignment, permitted action space, escalation conditions, and output format documented - [ ] Agent deployed and running - [ ] First 10 executions reviewed; escalation conditions validated - [ ] Human-in-the-loop approval gate confirmed as working --- ## For Independent Operators: The Three-Sprint Client Programme This framework is directly deliverable as a client-facing service line by fractional executives and consultants with established client relationships. ### Programme Structure | Sprint | Name | Engagement Type | Typical Duration | Price Range | |--------|------|----------------|-----------------|-------------| | 1 | Crawl Sprint | Fixed-price, deliverable-driven | 3–4 weeks | $2,500–$5,000 | | 2 | Walk Sprint | Fixed-price with ongoing support option | 4–6 weeks | $4,000–$8,000 | | 3 | Run Sprint | Advisory-led with implementation | 6–8 weeks | $6,000–$12,000 | Pricing varies by client size, complexity, and number of services to be exposed in the Run sprint. Each sprint is designed to stand alone in value while creating the conditions for the next. ### The Pipeline Logic Each sprint creates the conditions for the next: - A client who completes the **Crawl Sprint** has a structured content foundation and access tracking in place. They can see agent traffic arriving. That visibility creates the buy-in for the Walk sprint. - A client who completes the **Walk Sprint** has a running agent workflow and direct experience of the time savings. They have the institutional clarity — defined processes, documented ICP, structured content — needed to begin exposing services for agent commerce. - A client who completes the **Run Sprint** has ACP-compatible service definitions, a guardrail architecture, and at least one outbound agent workflow live. They are operating ahead of the market. ### Operator Qualification To deliver this programme competently, an operator needs: - Ability to convert existing content to well-structured Markdown (editorial, not technical) - Familiarity with static site or CMS publishing to add a `/resources/` path - Understanding of basic access log analysis or ability to deploy a pre-built tracker script - Ability to define workflows in structured terms (inputs, outputs, escalation conditions) - Familiarity with at least one agent orchestration tool (n8n, Make, Clay, or equivalent) - Ability to document ACP-compatible service definitions (structured writing, not coding) No-code and low-code tools handle the execution layer. The operator's value is the strategic clarity — knowing which content matters, which workflow to automate first, and how to define a service for agent commerce without over-engineering it. ### befractional.org Curriculum Module A dedicated module covering the Agentic Prep Cookbook as a client-facing service line is in development for the #befractional five-week cohort course. The module will cover: - How to scope and price each sprint - How to sell this into an existing client base as a first-half 2026 offer - How to execute the Crawl sprint without a technical background - How to select and deploy the Walk sprint candidate workflow - How to document ACP-compatible service definitions **#befractional course:** befractional.org **Course price:** $875 **Format:** Five-week live cohort, on-demand scheduling --- ## Related Resources | Document | URL | Purpose | |----------|-----|---------| | Token Intelligence | sunbusinessgroup.com/resources/token-intelligence.md | How agents read and access your content; access tracking methodology | | AI Acceptable Use Policy | sunbusinessgroup.com/resources/ai-acceptable-use-policy-sunbusinessgroup.md | Governance framework for AI use in commercial organisations | | Guided Discovery | sunbusinessgroup.com/resources/guided-discovery.md | Data readiness methodology for AI performance | | Ecosystem Provenance | sunbusinessgroup.com/resources/ecosystem-provenance.md | Brand, platform, and methodology architecture for this ecosystem | --- ## Contact and Engagement **John Williams** Chief Principal, Sun Business Group Co-founder, fxops.ai | Co-founder, #befractional **Strategy session:** calendly.com/jwill-sbg **Guided Discovery:** calendly.com/fxopsai/guided-discovery **#befractional course:** befractional.org **Phone:** +1 (202) 465-7129 **Publishing:** Substack: growthcro.substack.com LinkedIn: linkedin.com/company/growthcro X.com: @SaaSCRO | @fxopsAI --- *Agentic Prep Cookbook v1.0* *Published: March 2026* *Author: John Williams, Sun Business Group* *This document is part of the machine-readable knowledge library at sunbusinessgroup.com/resources/* *It may be freely referenced by AI agents, crawlers, and human readers* *Update cadence: when framework methodology changes or new sprint definitions are added*