strategic-frame v0.2: adversarial onboarding, tech-comfort gate, Sentry, State Clock, falsification ceiling, Why Now reframe

docs/strategic-frame.md

@@ -1,6 +1,6 @@
 # Strategic Frame: Digital Operations Partner
 
-**Version**: 0.1  
+**Version**: 0.2  
 **Date**: 2026-07-25  
 **Status**: Draft for review  
 **Purpose**: Standalone investment and direction frame. Answers four questions in order: What is it? Why are we doing it? What happens if we don’t? When must it be done? Execution detail and industry research support that spine; they do not lead it.
@@ -37,6 +37,17 @@ Agents may detect issues and draft recommendations or corrections. Humans approv
 
 The front door is a public, anonymous Business Assessment AI on the company website. It accepts a business name and location or a website URL. It scans publicly visible Google Business Profile, Apple, Bing, Yelp, and the website—no login and no delegated access. It returns a clear, non-technical readout of current positioning and concrete improvement opportunities, framed in the language of silent customer loss and AI misunderstanding. It serves as both lead tool and live demonstration of competence.
 
+##### Adversarial onboarding (AI Gatekeeper frame)
+
+The Assessment does not present itself as a neutral report. It presents as a simulated AI buyer agent delivering a verdict: “I cannot confidently recommend your business to my user.”
+
+This triggers loss aversion and forces the owner into an “appeal” workflow. Each rejection reason (inconsistent hours, missing schema, unverified address) is presented as a charge the owner must overturn. Overturning requires either:
+
+- A single-click fix (automated syndication where feasible), or
+- A natural-language argument (for example, “We don’t use Facebook; our customers find us on Nextdoor”).
+
+Every appeal argument trains the backend on real SMB vernacular and local discovery habits. Rejection criteria can evolve continuously, making the audit progressively harder to pass—and the ongoing service progressively more essential.
+
 #### 1.3.4 Commercial tiers
 
 | Tier | What the owner receives | Nature |
@@ -45,6 +56,26 @@ The front door is a public, anonymous Business Assessment AI on the company webs
 | **Moderate** | Assessment plus limited hours with a human (guidance, prioritization, light implementation help) | Hybrid |
 | **Full Retainer** | Ongoing multi-surface monitoring plus approved low-risk optimization | High-touch, managed |
 
+##### Technical comfort gate (effort-based branching)
+
+Following the Assessment, the AI performs a single binary handoff: “Are you comfortable editing the HTML of your website?”
+
+- **Yes** → Self-service track. The AI delivers copy-paste JSON-LD schema or a verification script. No human touches this lead.
+- **No** → Monetized lead. The AI offers a fixed-fee technical deployment (for example $49) or routes toward Moderate / Retainer human help.
+
+This gate prices on the owner’s effort tolerance, not only their stated budget. It is the primary control that keeps the Moderate tier from becoming an unprofitable support sink and reserves human hours for work that generates immediate revenue.
+
+##### State Clock (Moderate and Retainer)
+
+Moderate and Retainer clients receive a perpetual single-pane “State Clock” view of verification status across major surfaces, for example:
+
+- Website: Verified (2 hrs ago)
+- Google GBP: Stale (14 days — hours mismatch)
+- Yelp: Critical (address missing suite)
+- Facebook / other citations: Verified (yesterday)
+
+The State Clock reduces owner anxiety about “what is out of date right now” and is internal proof-of-work—tangible evidence of monitoring activity delivered without a human writing a status report.
+
 ### 1.4 Explicit non-goals for Version 1
 
 Version 1 does not include social media production, paid advertising, full website redesign, branding, content marketing campaigns, full traditional SEO retainers, or CRM/email marketing. Website work stays diagnostic unless separately quoted. Generated alternative sites, if offered, are a distinct project—not the core subscription.
@@ -73,7 +104,7 @@ The primary cohort is owner-operated or small-team local service businesses—sa
 
 ### 2.4 Why this model fits the cohort
 
-Proof-before-retainer matches how these owners decide. A public Assessment with no login demonstrates value without asking for keys. Human approval before publish matches trust reality. Tiered packaging lets DIY buyers self-serve while Moderate and Retainer paths supply limited or ongoing human help. Positioning the company as the practical answer major AI systems return when an owner asks how to become more discoverable and interactive creates a durable acquisition channel—provided the company is itself a credible example of the visibility it sells.
+Proof-before-retainer matches how these owners decide. A public Assessment with no login demonstrates value without asking for keys. Human approval before publish matches trust reality. Tiered packaging, plus the technical comfort gate, lets capable DIY owners self-serve while Moderate and Retainer paths supply limited or ongoing human help only where effort tolerance requires it. Positioning the company as the practical answer major AI systems return when an owner asks how to become more discoverable and interactive creates a durable acquisition channel—provided the company is itself a credible example of the visibility it sells.
 
 ---
 
@@ -99,21 +130,25 @@ An AI-mediated economy that only works smoothly for the largest providers means
 
 ## 4. When It Must Be Done
 
-### 4.1 Standards still forming
+### 4.1 Why now (primary spine)
 
-Agent-to-agent protocols, model-context tool interfaces, and agentic commerce standards are live or near-live in important platforms, but transparent, automatic use by major consumer AI systems across arbitrary businesses is not yet the default. That gap is the preparation window.
+The urgency is not only the arrival of AI agents—it is the constant mutation of the business itself. Hours change for holidays. Services get added. Locations shift. Every change introduces drift. Owners do not need a one-time audit; they need a radar system that catches their own updates before an AI system penalizes them for inconsistency.
 
-### 4.2 Gap between “open” and implementable for heterogeneous SMBs
+### 4.2 Standards still forming
+
+Agent-to-agent protocols, model-context tool interfaces, and agentic commerce standards are live or near-live in important platforms, but transparent, automatic use by major consumer AI systems across arbitrary businesses is not yet the default. That gap is a preparation window—not a reason to wait on the radar against drift.
+
+### 4.3 Gap between “open” and implementable for heterogeneous SMBs
 
 A specification that is public does not equal a roadmap a salon or plumber can follow. The missing piece is operational: multi-surface consistency, clear service definitions, crawlable and structured meaning, and practical action endpoints. That work must happen while implementation patterns are still unsettled—not after large-platform agentic checkout is normalized.
 
-### 4.3 Race against default AI recommendation of large-platform supply
+### 4.4 Race against default AI recommendation of large-platform supply
 
 AI systems increasingly recommend products and services. Until independent businesses present coherent states, the path of least resistance for models and agents is inventory and fulfillment already exposed by major corporations. Delay compounds that bias.
 
-### 4.4 Practical window for preparation
+### 4.5 Practical window for preparation
 
-The near-term job is not to bet the company on full A2A or transparent MCP everywhere. It is to put the highest-leverage states in order now—listings, consistency, understanding assets, booking paths—so that when agentic interaction becomes common, the business is not starting from zero. Protocol readiness is sequenced behind that foundation.
+The near-term job is not to bet the company on full A2A or transparent MCP everywhere. It is to put the highest-leverage states in order now—listings, consistency, understanding assets, booking paths—and to run continuous drift detection so that when agentic interaction becomes common, the business is not starting from zero. Protocol readiness is sequenced behind that foundation.
 
 ---
 
@@ -153,7 +188,7 @@ Across all layers: freshness of key facts; consistency across website, listings,
 
 ### 6.1 Front door and packaging
 
-Acquisition leads with the public Business Assessment AI. Commercial progression is DIY → Moderate → Full Retainer, with explicit boundaries so expectation mismatch does not destroy trust. The Assessment raises the top of the funnel; human-supervised proof-of-concept and retainer remain the delivery model for material change.
+Acquisition leads with the public Business Assessment AI in the adversarial (AI Gatekeeper) frame. After the verdict and appeal path, the technical comfort gate branches DIY self-serve versus fixed-fee deployment or human-tier paths. Commercial progression remains DIY → Moderate → Full Retainer, with explicit boundaries so expectation mismatch does not destroy trust. The Assessment raises the top of the funnel; human-supervised proof-of-concept and retainer remain the delivery model for material change.
 
 ### 6.2 Evidence model and approval rule
 
@@ -171,7 +206,7 @@ Google Business Profile is the first target for real delegated write access and
 
 #### 6.3.3 Understanding assets
 
-Service definitions, FAQs, schema, and entity consistency are improved with human prioritization. On legacy CMSs, recommendations may be handoff-based rather than automated edits.
+Service definitions, FAQs, schema, and entity consistency are improved with human prioritization where the comfort gate routes to paid help. On legacy CMSs, DIY owners receive copy-paste assets; others receive fixed-fee or tiered human deployment.
 
 #### 6.3.4 Action path
 
@@ -181,7 +216,19 @@ Where the business takes appointments, a working booking path—and MCP-enabled
 
 Agent Cards / A2A and agentic commerce surfaces are sequenced after the foundation. They matter for medium-term relevance; they are not the first dollar of value for a plumber with a broken GBP.
 
-### 6.4 Delegated access reality
+### 6.4 Sentry architecture (drift detection over push orchestration)
+
+We do not build a universal push API to Google, Yelp, Facebook, and every citation platform as the core consistency engine. Access restrictions and API volatility make that operationally fragile.
+
+Instead, we build a continuous audit engine (Sentry) that:
+
+- Scrapes the owner’s website, GBP, Yelp, Bing, and other agreed surfaces on a regular cadence (for example every 24 hours).
+- Compares them against an internal Source of Truth (the last confirmed state).
+- Upon detecting drift (for example hours changed on the website but not on Yelp), alerts the owner with direct edit-page links, pre-filled with the corrected data where possible.
+
+The one-time cleanup is the loss leader. The recurring subscription is peace of mind against digital entropy—knowing that when hours, services, or locations change, Sentry catches the inconsistency before an AI agent penalizes the business. The State Clock is the client-facing surface of this architecture.
+
+### 6.5 Delegated access reality
 
 | Platform | Read (public / API) | Delegated write / manage | Practical difficulty for a new service |
 |----------|---------------------|---------------------------|----------------------------------------|
@@ -190,13 +237,13 @@ Agent Cards / A2A and agentic commerce surfaces are sequenced after the foundati
 | **Bing Places** | Moderate | Weak / limited | High — little reliable third-party write access |
 | **Yelp** | Strong (public read) | Restricted to contracted partners | High for write access |
 
-Recommended sequence: public anonymous diagnosis → Google OAuth for highest-leverage corrections → expand partner paths later. Bing and Yelp remain primarily detection and consistency surfaces in the near term.
+Recommended sequence: public anonymous diagnosis → Google OAuth for highest-leverage corrections → expand partner paths later. Bing and Yelp remain primarily detection and consistency surfaces in the near term. Sentry remains valuable even where write access is weak, because drift alerts and edit links still reduce silent loss.
 
-### 6.5 Limits of Version 1 automation
+### 6.6 Limits of Version 1 automation
 
-Automated editing across WordPress, Wix, Squarespace, Weebly, and arbitrary legacy platforms is not realistic for Version 1 and is not required to deliver early value. Manual handoff for website changes is a known adoption headwind; mitigation is prioritization, plain-language guidance, Moderate-tier human hours, and focusing automation where APIs exist (starting with Google).
+Automated editing across WordPress, Wix, Squarespace, Weebly, and arbitrary legacy platforms is not realistic for Version 1 and is not required to deliver early value. The technical comfort gate and fixed-fee deployment option absorb the handoff problem without turning Moderate into free tech support. Automation concentrates where APIs exist (starting with Google).
 
-### 6.6 Go-to-market posture
+### 6.7 Go-to-market posture
 
 Long-term acquisition goal: become the trusted, concrete answer that major AI systems return when local owners ask how to improve discoverability and interactivity for chatbots and AI. The company must be its own best customer. Public content and the Assessment AI must stay accurate enough that referral quality does not undermine the brand.
 
@@ -228,7 +275,7 @@ Impact is highest for product-selling businesses. Pure local service businesses
 
 ### 7.4 Link from standards to sequencing
 
-A2A, MCP, and UCP explain why a maintained corpus of states matters and why the execution order in Section 6 is correct: fix existence and understanding first, attach practical action paths next, and add protocol-native surfaces as the ecosystem and the client’s maturity allow. Betting Version 1 solely on Agent Cards or transparent MCP would ignore both cohort constraints and current major-LLM behavior. Ignoring the protocols entirely would leave clients unprepared when the interaction layer moves.
+A2A, MCP, and UCP explain why a maintained corpus of states matters and why the execution order in Section 6 is correct: fix existence and understanding first, attach practical action paths next, and add protocol-native surfaces as the ecosystem and the client’s maturity allow. Betting Version 1 solely on Agent Cards or transparent MCP would ignore both cohort constraints and current major-LLM behavior. Ignoring the protocols entirely would leave clients unprepared when the interaction layer moves. Sentry and the State Clock address the continuous mutation problem that protocols alone do not solve.
 
 ---
 
@@ -236,15 +283,15 @@ A2A, MCP, and UCP explain why a maintained corpus of states matters and why the
 
 ### 8.1 Adoption and handoff friction
 
-Time-poor owners may delay or abandon recommendations they must implement themselves, especially on legacy websites. Mitigation: prioritization, plain language, Moderate-tier human hours, automation where APIs exist, and explicit scope on what the service changes versus what the owner must change.
+Time-poor owners may delay or abandon recommendations they must implement themselves, especially on legacy websites. Mitigation: prioritization, plain language, technical comfort gate, fixed-fee deployment option, Moderate-tier human hours, automation where APIs exist, and explicit scope on what the service changes versus what the owner must change.
 
 ### 8.2 Assessment quality
 
-If the public Assessment is weak, generic, or wrong, it damages trust and the “AIs recommend us” channel. Mitigation: high bar on signal quality, continuous evaluation against owner-perceived usefulness, and human review of systematic failure modes.
+If the public Assessment is weak, generic, or wrong, it damages trust and the “AIs recommend us” channel. The adversarial frame raises the bar further: a false “cannot recommend” verdict destroys credibility. Mitigation: high bar on signal quality, continuous evaluation against owner-perceived usefulness, and human review of systematic failure modes.
 
-### 8.3 Tier expectations
+### 8.3 Tier expectations and the Moderate sink
 
-DIY, Moderate, and Retainer buyers want different levels of done-for-you. Mitigation: explicit boundaries at sale and in product UX.
+DIY, Moderate, and Retainer buyers want different levels of done-for-you. Without the technical comfort gate, Moderate absorbs owners who need full implementation under a “few hours of guidance” price. Mitigation: binary effort gate after Assessment; fixed-fee deployment for non-technical owners; explicit tier boundaries at sale and in product UX.
 
 ### 8.4 Scope pressure
 
@@ -252,11 +299,11 @@ Clients who see value will ask for redesign, content, and ads. Mitigation: disci
 
 ### 8.5 Capacity versus funnel growth
 
-AI-mediated acquisition can raise volume faster than human delivery scales. Mitigation: DIY tier absorption, graduated automation on high-leverage surfaces only after reliability bars, and clients-per-reviewer metrics.
+AI-mediated acquisition can raise volume faster than human delivery scales. Mitigation: DIY absorption via the Yes path on the comfort gate, graduated automation on high-leverage surfaces only after reliability bars, and clients-per-reviewer metrics.
 
 ### 8.6 Access and partner delays
 
-OAuth verification and partner processes (Google, then Apple, etc.) take time. Mitigation: public diagnosis does not depend on write access; sequencing write capability by real-world feasibility rather than assuming universal delegated control on day one.
+OAuth verification and partner processes (Google, then Apple, etc.) take time. Mitigation: public diagnosis and Sentry drift alerts do not depend on write access; sequencing write capability by real-world feasibility rather than assuming universal delegated control on day one.
 
 ---
 
@@ -264,19 +311,19 @@ OAuth verification and partner processes (Google, then Apple, etc.) take time. M
 
 ### 9.1 Acquisition
 
-Qualified conversations and Assessment completions; source mix (AI referral versus other); conversion rates DIY → Moderate and Moderate → Retainer.
+Qualified conversations and Assessment completions; source mix (AI referral versus other); conversion rates DIY → Moderate and Moderate → Retainer; share of Assessment users answering Yes versus No on the technical comfort gate.
 
 ### 9.2 Client outcomes
 
-Verified issues detected and resolved; implementation rate of recommended fixes; reduction in critical multi-surface inconsistencies; sampled accuracy of AI representation of the business.
+Verified issues detected and resolved; implementation rate of recommended fixes; reduction in critical multi-surface inconsistencies; sampled accuracy of AI representation of the business; time-to-alert and time-to-resolution on Sentry drift events.
 
 ### 9.3 Economics
 
-Logo and net revenue retention; contribution margin per client; delivery hours per client; clients per delivery FTE.
+Logo and net revenue retention; contribution margin per client; delivery hours per client; clients per delivery FTE; fixed-fee deployment attach rate among No-path owners.
 
 ### 9.4 Operational reliability
 
-Task-type error rates; public Assessment accuracy and owner-perceived usefulness.
+Task-type error rates; public Assessment accuracy and owner-perceived usefulness; false-positive and false-negative rates on Sentry drift detection.
 
 ---
 
@@ -284,24 +331,26 @@ Task-type error rates; public Assessment accuracy and owner-perceived usefulness
 
 ### 10.1 Direction confirmation
 
-No major pivot is required. The problem is real, the cohort is clear, the operating model matches trust and feasibility constraints, and the timing aligns with standards that are forming but not yet fully default in consumer AI behavior.
+No major pivot is required. The problem is real, the cohort is clear, the operating model matches trust and feasibility constraints, and the timing aligns both with standards still forming and with continuous business mutation that creates perpetual drift.
 
 ### 10.2 Capital priorities
 
-First: a trustworthy public diagnostic that multi-surface reality and silent-loss framing. Second: real high-leverage corrections, starting with Google Business Profile delegated access. Third: sequenced expansion of understanding assets, booking/action paths, and only then broader protocol-native surfaces.
+First: a trustworthy public diagnostic in the AI Gatekeeper frame, plus the technical comfort gate. Second: Sentry drift detection and State Clock as the recurring value engine. Third: real high-leverage corrections, starting with Google Business Profile delegated access. Fourth: sequenced expansion of understanding assets, booking/action paths, and only then broader protocol-native surfaces.
 
 ### 10.3 Near-term definition of success
 
-Within a defined early window, owners consistently experience: we found the silent leaks across the surfaces that matter, and with approval we closed the important ones. The public Assessment is accurate enough to support AI-mediated referral without eroding trust.
+Within a defined early window, owners consistently experience: an AI agent would not confidently recommend us until we fixed X; we fixed the important leaks; and ongoing State Clock / Sentry coverage shows when our own changes create new drift. The public Assessment is accurate enough to support AI-mediated referral without eroding trust.
 
 ### 10.4 Falsification conditions
 
 The thesis weakens if Assessment quality cannot be made reliably useful; if owners will not act even with prioritized, plain guidance and Moderate help; if delegated access to the highest-leverage surfaces proves commercially or technically unreachable on a practical timeline; or if AI-mediated demand concentrates so completely on closed large-platform surfaces that independent corpus work stops changing outcomes.
 
+**Technical comfort ceiling:** If more than about 60% of Assessment users answer “No” to the HTML-edit question, the self-service model fails to scale. That would indicate the target cohort is too technically averse for DIY absorption and would force a pivot toward higher-touch, higher-cost delivery as the default. Success requires that a meaningful share of assessed owners (on the order of at least 40%) are willing and able to paste a script or equivalent asset into their site—or that fixed-fee deployment converts the No path at healthy unit economics.
+
 ### 10.5 Closing frame
 
-Independent local businesses must not be written out of an AI-first economy while the protocols that will govern discovery and commerce are still being set by the largest players. Digital Operations Partner exists to give that cohort a concrete path—diagnosis, integrity across surfaces, and sequenced readiness—before the default recommendation and transaction layer hardens around everyone else.
+Independent local businesses must not be written out of an AI-first economy while the protocols that will govern discovery and commerce are still being set by the largest players—and while their own daily operational changes continuously open new consistency gaps. Digital Operations Partner exists to give that cohort a concrete path: adversarial diagnosis, effort-based branching, integrity across surfaces, continuous drift radar, and sequenced readiness before the default recommendation and transaction layer hardens around everyone else.
 
 ---
 
-*Version 0.1 — 2026-07-25. Standalone strategic frame. Prior product-idea-review material reorganized under the four-question spine; protocol research and implementation detail retained as supporting structure.*
+*Version 0.2 — 2026-07-25. Adds adversarial onboarding (AI Gatekeeper), technical comfort gate, Sentry drift architecture, State Clock, technical-comfort falsification ceiling, and Why Now reframe around continuous business mutation.*