COMPLIANCE

Built around human oversight in surveying. AI assists, it doesn’t replace judgement.

Sitarva is designed around transparency, human oversight and clear accountability – principles that underpin responsible AI use in surveying practice.

COMPLIANCE CONTROLS

Transparency by design

Built around clear AI controls and human oversight.

Raw Voice Evidence Log

The raw voice transcription is appended to every final report as a verifiable record of the observations captured during the inspection.

Sitarva Assisted Processing

Surveyors can disable AI assistance at any time and continue using Sitarva with fully manual data entry.

AI Output & Review Controls

Sitarva applies defined controls around how AI-assisted content is organised, reviewed and presented, keeping the surveyor in control of the final output.

OUR APPROACH

AI that assists. Surveyors who decide.

Sitarva captures and organises what you observe on site. It does not assess building conditions, determine risk, or draw conclusions beyond what you record. Every report remains subject to human review before issue.
Sitarva is built to support the surveying process, from capture on site to draft report sections.
THE GUARDRAILS

10 product-level guardrails

These aren't policies written after the product was built. They're product-level constraints designed into how Sitarva processes survey information.

These constraints define how Sitarva processes survey information, keeping the system focused on recorded evidence.

1. Zero-Invention Guardrail

▲
```
PURPOSE
Prevents fabricated detail or implied inspection.
RULE
If information has not been explicitly recorded by the surveyor, Sitarva does not invent, assume or imply it. Where appropriate, the output remains "Not observed", "Unknown" or "Unverified".
Example
No observation about insulation → No statement about insulation or thermal performance
Restriction
Sitarva cannot claim that an element was inspected, assessed or confirmed when it was not recorded.
RISK MITIGATED
Fabricated facts, implied inspection and unsupported conclusions.
```

2. Observation ≠ Diagnosis Guardrail

▲
PURPOSE
Separates observed condition from diagnosis or interpretation.
RULE
Sitarva may describe what the surveyor observed, but must not introduce a diagnosis, failure mechanism or technical conclusion that was not explicitly stated.
Example
"Cracked ridge tiles observed" ✅ Descriptive observation only
Restriction
"Cracked due to fixing failure" ❌ unless explicitly stated by the surveyor
RISK MITIGATED
Unsupported diagnosis, expert overreach and unintended conclusions.

3. Cause Containment Guardrail

▲
PURPOSE
Prevents multiple speculative causes being combined into a single narrative.
RULE
Sitarva keeps condition statements descriptive and must not combine multiple possible causes unless they have been explicitly stated by the surveyor.
Example
"Consistent with age and localized movement."
Restriction
"Likely caused by age, weather, and poor workmanship." ❌ unless explicitly stated by the surveyor
RISK MITIGATED
Speculative causation, unsupported theories and unnecessary professional exposure.

4. Biological Growth Separation Guardrail

▲
PURPOSE
Prevents biological growth being treated as a defect, cause or evidence of failure without supporting observation.
RULE
Moss, lichen and similar biological growth must be described as observed and must not be linked to moisture ingress, failure or missing or damaged elements unless the surveyor has explicitly made that connection.
Example
"Moss observed on roof covering" ✅ Describe the growth separately from other defects
Restriction
No implied link to moisture ingress or failure without a recorded observation
RISK MITIGATED
Unsubstantiated moisture, defect or performance claims.

5. Section Lock Guardrail

▲
PURPOSE
Keeps observations within the scope of the active survey section.
RULE
Content generated for a section must remain relevant to that section and must not introduce unsupported observations or conclusions belonging to another area of the inspection.
Example
Roof Coverings section → describes roof coverings and observations recorded there
Restriction
Roof Coverings section → cannot imply that timbers, underlay or structure were inspected
RISK MITIGATED
Scope creep and implied inspection beyond the recorded evidence.

6. Concealed Components Guardrail

▲
PURPOSE
Prevents conclusions about concealed or inaccessible elements.
RULE
Sitarva must not imply that concealed or inaccessible components were inspected or assessed unless the surveyor explicitly recorded that they were observed.
Example
Roof underlay not observed → No statement about its condition
Restriction
Roof underlay, fixings or joists → No claim about condition unless observed
RISK MITIGATED
False assurance and implied inspection beyond stated limitations.

7. Repair Neutrality Guardrail

▲
PURPOSE
Keeps recommendations focused on the required action, rather than prescribing a specific repair method.
RULE
Recommendations may identify an appropriate action for the surveyor to consider, but must not invent or prescribe a specific repair method, material or specification unless explicitly provided by the surveyor.
Example
"Reinstate missing slates" ✅ Action identified without prescribing the method
Restriction
"Re-fix with stainless steel nails" ❌ unless specified by the surveyor
RISK MITIGATED
Unintended design, specification or workmanship responsibility.

8. Contradiction Detection Guardrail

▲
PURPOSE
Preserves conflicting observations rather than silently resolving them.
RULE
When recorded observations or statements appear to conflict, Sitarva flags the inconsistency for review rather than choosing which statement is correct.
Example
"Roof in good condition" + "Active leakage" → Flag for review
Restriction
Sitarva must not resolve the conflict by selecting one observation over the other
RISK MITIGATED
Silent contradictions, false certainty and guessed resolution.

9. Immediate Hazard Override Guardrail

▲
PURPOSE
Ensures explicitly recorded high-risk observations receive safety-focused treatment.
RULE
Defined high-risk observations trigger a safety-focused review and must not be softened or minimised by the system. The surveyor remains responsible for determining the appropriate action.
Example
"Gas smell" → Safety-focused review and appropriate warning
Restriction
The system must not soften a recorded high-risk observation with terms such as "minor", "localized" or "monitor" where they would contradict the recorded risk
RISK MITIGATED
Minimised hazards, unsafe conclusions and critical safety failures.

10. Human Authority Guardrail

▲
PURPOSE
Keeps professional judgement and final authority with the surveyor.
RULE
All Sitarva outputs are drafts for review. The surveyor retains final authority over the content, conclusions and decision to issue the report.
Example
Surveyor reviews, edits, and approves the report before issue
Restriction
Sitarva must not present its output as autonomous professional judgement
RISK MITIGATED
AI-as-expert misrepresentation and loss of human oversight.
GOVERNANCE & TECHNICAL STANDARDS

A bounded-logic environment. Not open-ended AI.

Unlike general-purpose AI systems, Sitarva processes survey information within defined boundaries, section limits and product-level guardrails. Its purpose is to capture and organise what the surveyor records, not to generate unrestricted opinion.

AI-assisted content remains subject to the surveyor’s review. The surveyor can edit, reject or amend the output before approving the final output.

Sitarva operates within a controlled processing environment, with specialised stages designed around defined surveying tasks rather than unrestricted general-purpose generation.

The system is designed to restate, structure and clarify information provided by the surveyor, rather than introducing unsupported observations or conclusions.

A validation layer checks for potentially inconsistent or non-standard inputs and can flag information requiring review before it is carried forward.

Multiple controls are designed to reduce overreach, speculation and misrepresentation, including contradiction detection, scope controls and human review requirements.

Sitarva works from the observations, evidence and information captured by the surveyor. It is not designed to fill gaps with information that was never recorded.

Observations are processed within the context of the report section they relate to, helping prevent information from one part of the inspection being incorrectly applied to another.

AI-assisted drafting works within predefined report formats and section structures rather than producing unrestricted narrative from a general-purpose prompt.

 

Outputs are constrained by the scope of the recorded observations and the relevant survey context, helping prevent the system from expanding beyond the information provided.

Sitarva does not independently assess building conditions, determine risk or reach professional conclusions. Those decisions remain with the surveyor.

 

AI-assisted content is presented for surveyor review before the report is issued. The surveyor can edit, reject or amend the output before approving the final report.

These controls define the boundaries within which Sitarva processes and organises survey information. They are designed to keep the system focused on the evidence captured at the point of inspection.
RESPONSIBLE OUTPUT

Built around responsible AI use in surveying

Sitarva is designed around principles of transparency, human oversight and accountability when AI is used to support surveying work. AI assists with the process; the surveyor remains in control of the output.

Transparency

AI-assisted content is identifiable within the Sitarva workflow, so the surveyor can see where AI has been used to build the report.

Human oversight

Surveyors review and edit AI-populated content before the report is generated. Nothing is issued without final human review.

Accountability

Responsibility remains with the surveyor. Sitarva assists with capture and organisation, it does not make independent surveying decisions.

Compliance position summary

Designed around clear AI controls.

Sitarva is designed to support professional surveyors in capturing, organising and reviewing their recorded observations. It does not make independent surveying decisions or remove the need for human review.

The surveyor remains in control of the final report, including its findings and conclusions.

Sitarva supports the inspection and reporting process through organised capture, transparent AI assistance and traceable records. The final report is reviewed by the surveyor before issue.
in practice

Built around a surveyor, not instead of them

You capture it

Your observations and evidence come from the inspection. Sitarva organises what you record and does not invent what wasn't captured.

Sitarva structures it

AI assistance organises your recorded information into the appropriate report format, within defined controls and section boundaries.

You decide

You review, edit and approve the report before it leaves your hands. Sitarva provides the framework and assistance; you make the final call.

SEE IT IN ACTION
READY TO SEE IT IN PRACTICE?

Capture the inspection. Preserve the context.

See how Sitarva turns spoken observations and on-site evidence into finished draft report sections.