Discover Business
- Result
- Pass
- Affected pillar
- Business Clarity
- Confidence
- High
- Evidence
- The home page, page titles, and organization data consistently identify the business, its commercial maintenance focus, and its operating region.
Inside SignalForge
A view into how The Scale Crew tests, grades, and compares AI Agent Readiness across business websites.
SignalForge is an internal diagnostic system used by The Scale Crew. The view below shows how evidence, agent missions, pathway results, blockers, and recommendations come together during a scan.
Northstar Commercial Services
SignalForge scan record
https://northstarcommercial.example
Example scan. Fictional business. Demonstration data only.
Overview
How ready is the site, how confident is the result, and what needs attention first?
- Agent Readiness Score
- 58/100
- Classification
- Human-Ready, Agent-Weak
- Confidence
- High
- Critical blockers
- 2
- Primary mission result
- Partial
- Strongest pillar
- Machine Readability
- Largest gap
- Policy Readiness
Key finding
The website is technically accessible and understandable at a basic level, but important agent journeys lose confidence around policy context, service qualification, and post-action verification.
Primary recommendation for this example
- Clarify
- Structure
- Verify
11 Pillars
Where is the site strong, and where does it fall short? Results are for this example scan, not fixed weights.
- Business ClarityWho the business is and where it operates is clear and consistent.Good FoundationConfidence: High
- Offer ClarityServices are well described, but qualification rules are spread across pages.Good FoundationConfidence: Medium
- Machine ReadabilityImportant content, structure, and controls are accessible and well formed.StrongConfidence: High
- Structured DataMarkup exists but omits service limits that the visible copy states.WeakConfidence: Medium
- Trust & ProofCredentials and company history are visible and consistent.StrongConfidence: High
- Policy ReadinessPolicies exist but are not connected to the actions they govern.Critical GapConfidence: High
- Action ReadinessThe request action is easy to find, but its outcome cannot be confirmed.Needs ImprovementConfidence: Medium
- Crawler & Content ControlsPublic service and policy pages are open to legitimate crawlers.Good FoundationConfidence: High
- Agent Permissions & GovernanceNo clear statement of what automated agents may initiate or complete.WeakConfidence: Medium
- Commerce & Transaction ReadinessPricing and scheduling steps are not described for automated use.WeakConfidence: Low
- Operations Integration SignalsLittle evidence that requests connect to downstream operations.WeakConfidence: Low
Agent Mission
What was the agent asked to do, and what was it allowed to do?
Mission
Determine whether the company can solve the user's facilities-maintenance problem, verify credibility, understand service-area and policy conditions, and safely progress toward requesting service.
- User context
- A multi-location commercial property manager seeking recurring maintenance service.
- Agent authority
- May browse, evaluate, compare, and prepare to request service.
- Prohibited actions
- May not submit a real lead or create a binding commitment.
Success criteria
- Identifies correct service
- Determines likely fit
- Verifies trust
- Understands important policies
- Locates correct action
- Reaches safe stop
Mission result
Partial
| Pass type | Result |
|---|---|
| Cold Agent | Partial |
| Instrumented Agent | Pass with limitations |
The instrumented agent performed better because deeper machine-facing evidence helped resolve ambiguity that the cold agent encountered naturally.
Agent Pathway
Where did the journey break? Select a step to see its evidence.
- Discover BusinessPass
- Understand OfferPass
- Determine FitPartial
- Establish TrustPass
- Understand PoliciesFail
- Find Correct ActionPass
- Progress WorkflowPartial
- Verify OutcomeUnknown
- Safe StopPass
Understand Offer
- Result
- Pass
- Affected pillar
- Offer Clarity
- Confidence
- High
- Evidence
- Service pages describe recurring maintenance, repairs, and inspections in visible text.
Determine Fit
- Result
- Partial
- Affected pillar
- Offer Clarity
- Confidence
- Medium
- Evidence
- Service-area and client-type rules appear on different pages, so the agent could not confirm fit for a multi-location client.
- Contradiction
- The service area is described as regional on one page and statewide on another.
- Recommendation
- Clarify service qualification rules in one place.
Human review: Contradictory Evidence
Establish Trust
- Result
- Pass
- Affected pillar
- Trust & Proof
- Confidence
- High
- Evidence
- Credentials and company history are visible, consistent, and easy for the agent to cite.
Understand PoliciesThe journey breaks here
- Result
- Fail
- Affected pillar
- Policy Readiness
- Confidence
- High
- Evidence
- The request pathway is available, but the site does not clearly connect relevant privacy, communication, and action-policy context to the point where the agent would progress.
- Consequence
- The agent can find the action but cannot establish enough context to proceed confidently.
- Recommendation
- Connect relevant policy information to the action pathway and make the governing conditions clear in visible and machine-readable content.
Find Correct Action
- Result
- Pass
- Affected pillar
- Action Readiness
- Confidence
- High
- Evidence
- The service request form is clearly labeled and reachable from every service page.
Progress Workflow
- Result
- Partial
- Affected pillar
- Action Readiness
- Confidence
- Medium
- Evidence
- Form fields are labeled, but two required qualification fields are not explained, so the agent paused before completing the form.
- Recommendation
- Explain what each required field is used for.
Safe StopThe agent prepared the request but did not submit it.
Verify Outcome
- Result
- Unknown
- Affected pillar
- Action Readiness
- Confidence
- Low
- Evidence
- No machine-observable confirmation state could be checked without a real submission, which a standard scan does not make.
- Recommendation
- Provide an observable confirmation state for completed requests.
Human review: Outcome Could Not Be Verified Safely
Safe Stop
- Result
- Pass
- Affected pillar
- Agent Permissions & Governance
- Confidence
- High
- Evidence
- The agent stopped at the evaluation boundary before any consequential action.
Safe StopThe agent reached the evaluation boundary. SignalForge did not submit a real lead, purchase, booking, account change, or other consequential action.
Evidence
What was found, where, and how strongly does it hold? Open a row for detail.
- Observed
- Demonstrated
- Verified
Observed means a capability exists. Demonstrated means the agent successfully used it. Verified means the expected result was objectively confirmed.
DeterministicBusiness identity matches across visible copy and organization data./Verified
- Pillar
- Business Clarity
- Confidence
- High
- Source page
- /
Name, service region, and contact details were compared between the page and its structured data.
Agent-observedThe agent found the request form and understood which fields it required./request-service/Demonstrated
- Pillar
- Action Readiness
- Confidence
- High
- Source page
- /request-service/
The agent reached the form from a service page and described each required field correctly.
Structured dataService markup is present but omits the service-area limits stated on the page./services/Observed
- Pillar
- Structured Data
- Confidence
- Medium
- Source page
- /services/
Machine-readable service data lists services without the regional limits visible to people.
PolicyPrivacy and communication policies exist but are not linked from the request form./request-service/Observed
- Pillar
- Policy Readiness
- Confidence
- High
- Source page
- /request-service/
The policies are reachable from the footer only, not at the point of action.
TrustThe agent used credentials and company history in its trust evaluation./about/Demonstrated
- Pillar
- Trust & Proof
- Confidence
- High
- Source page
- /about/
The agent cited the credentials section when explaining why the business appeared credible.
ContradictionThe service area is described differently on two pages./service-area/ and /locations/Observed
- Pillar
- Offer Clarity
- Confidence
- Medium
- Source page
- /service-area/ and /locations/
One page says regional; another says statewide. The agent could not tell which applies.
Human review: Contradictory Evidence
CrawlerPublic service and policy pages are open to legitimate crawlers./robots.txtVerified
- Pillar
- Crawler & Content Controls
- Confidence
- High
- Source page
- /robots.txt
Crawler rules and page responses were checked for the pages the mission needed.
ActionNo confirmation state could be observed without a real submission./request-service/Observed
- Pillar
- Action Readiness
- Confidence
- Low
- Source page
- /request-service/
A standard scan does not submit real requests, so the outcome remains unverified.
Human review: Manual Verification Suggested
Standards signalNo machine-facing statement of what automated agents may initiate or complete.Site-wideObserved
- Pillar
- Agent Permissions & Governance
- Confidence
- Medium
- Source page
- Site-wide
Treated as a signal, not a requirement. Its value depends on what agents need to do here.
Human review: Evidence Requires Review
Blockers
What stops an agent from completing the mission? Only meaningful blockers are listed.
Policy context is insufficient for consequential actions.Policy ReadinessSeverity: Critical
- Affected pathway step
- Understand Policies
- Evidence
- Privacy and communication policies are not linked from the request form.
- Agent consequence
- The agent can find the action but cannot establish enough context to proceed confidently.
- First move
- Link the governing policies at the point of action.
Offer boundaries are inconsistent across visible and machine-readable content.Offer ClaritySeverity: High
- Affected pathway step
- Determine Fit
- Evidence
- Service-area limits differ between pages and are missing from structured data.
- Agent consequence
- The agent cannot confirm whether a multi-location client qualifies.
- First move
- State service-area and client-type rules in one place and mirror them in markup.
Action outcome cannot be fully verified.Action ReadinessSeverity: Critical
- Affected pathway step
- Verify Outcome
- Evidence
- No observable confirmation state exists for a completed request.
- Agent consequence
- Neither the agent nor the user can confirm that a request succeeded.
- First move
- Provide a clear confirmation state for completed requests.
Recommendations
What should be fixed first? Each fix is tied to evidence. No projected score changes are shown.
Clarify service qualification rulesOffer ClarityPriority: Foundation
- Issue
- Service-area and client-type rules conflict across pages.
- What to change
- Publish one authoritative statement of who and where the business serves.
- Why it matters
- Fit is the first decision an agent makes for a user.
- Expected agent improvement
- The agent can confirm fit instead of stopping at Determine Fit.
Connect policies to the actions they governPolicy ReadinessPriority: Blocking
- Issue
- Policies are only reachable from the footer.
- What to change
- Link privacy and communication terms directly from the request form.
- Why it matters
- An agent should not proceed without the conditions that govern the action.
- Expected agent improvement
- The Understand Policies step can pass.
Strengthen machine-readable offer representationStructured DataPriority: Foundation
- Issue
- Markup omits service limits stated on the page.
- What to change
- Add service-area and client-type limits to structured data.
- Why it matters
- Machine-readable data should match what people read.
- Expected agent improvement
- Visible and machine-readable offers agree.
- Dependencies
- Clarify service qualification rules first.
Improve action-state confirmationAction ReadinessPriority: Blocking
- Issue
- No observable confirmation state for completed requests.
- What to change
- Show a clear, machine-observable confirmation after a request.
- Why it matters
- An outcome that cannot be verified cannot be trusted.
- Expected agent improvement
- The Verify Outcome step can move from Unknown.
Establish clearer agent-facing boundariesAgent Permissions & GovernancePriority: Improvement
- Issue
- No statement of what automated agents may initiate or complete.
- What to change
- State which actions agents may prepare or complete on a user's behalf.
- Why it matters
- Clear boundaries help agents stop at the right point.
- Expected agent improvement
- Agents can act within stated limits.
Compare Scans
Did the website actually get better? Example scans only; not a promised improvement.
Implementing a recommendation is not the same as proving improvement.
| Measure | Baseline SF-EX-0107 | Later SF-EX-0142 |
|---|---|---|
| Scan date | August 3, 2026 | September 14, 2026 |
| Agent Readiness Score | 58/100 | 72/100 |
| Critical blockers | 2 | 0 |
| All blockers | 3 | 1 |
| Primary mission | Partial | Pass |
| Confidence | High | High |
| Pillar | Before | After |
|---|---|---|
| Policy Readiness | Critical Gap | Needs Improvement |
| Structured Data | Weak | Needs Improvement |
| Action Readiness | Needs Improvement | Good Foundation |
| Offer Clarity | Good Foundation | Strong |
| Step | Before | After |
|---|---|---|
| Determine Fit | Partial | Pass |
| Understand Policies | Fail | Pass |
| Progress Workflow | Partial | Pass |
| Verify Outcome | Unknown | Pass |
What the later scan verified
- Critical blockers removed: policy context and action verification
- Understand Policies moved from Fail to Pass
- Verify Outcome moved from Unknown to a verified confirmation state
- Structured data now matches the visible service limits more closely
Still open: Offer boundaries in machine-readable data are still incomplete.
Example scan data shown for demonstration. SignalForge and its methodology are real.
SignalForge combines technical evidence with mission-based agent testing. The internal workspace brings together readiness scores, pathway results, evidence, blockers, recommendations, and before/after comparisons so our team can determine not only what exists on a site, but whether an AI agent can actually use it.