RICS AI StandardBuildStack and the RICS AI Standard
Information for RICS members and regulated firms conducting due diligence on BuildStack under the RICS Professional Standard: Responsible use of artificial intelligence in surveying practice (effective 9 March 2026).
Section 4.1 - System OverviewWhat BuildStack does and how it works
BuildStack is a SaaS platform designed for UK building surveyors, contract administrators, and defect professionals. It assists qualified practitioners with:
- Defect diagnosis: generating probable cause analysis from surveyor-supplied descriptions and photographs
- Repair scope generation: producing draft repair specifications written to NBS-style standards
- Schedule of works: compiling costed schedules of works from logged defects
- Document export: producing formatted Word documents for professional review and issue
BuildStack does not perform inspections, make autonomous decisions, or issue outputs directly to clients. All outputs are produced as drafts for review, editing, and approval by the responsible surveyor before any client communication.
Section 4.1 - AI System DisclosureThe AI system used by BuildStack
BuildStack uses the Anthropic Claude large language model, accessed via the Anthropic API, to generate defect analysis and repair scope outputs.
- AI system type: Large language model (LLM) - natural language processing
- Provider: Anthropic PBC
- Model: Anthropic Claude
- Deployment: Cloud-based API, no on-premise compute
- Training data: Anthropic's published model documentation describes the training approach. BuildStack does not train or fine-tune the underlying model.
- BuildStack-specific prompting: BuildStack applies domain-specific system prompts built on UK building pathology, NBS repair specifications, Building Regulations compliance, and professional liability requirements. These prompts structure and constrain the model's outputs for the surveying context.
Known limitations and failure modes:
- The model may produce plausible-sounding but incorrect diagnoses where defect descriptions are ambiguous or incomplete
- Outputs may not reflect specialist knowledge in highly specific pathology areas (e.g. listed building materials, complex structural defects)
- Photo analysis is dependent on image quality and angle - poor photos reduce diagnostic accuracy
- The model has no access to the physical property and cannot replicate a site inspection
These limitations are why BuildStack requires professional review of all outputs before issue.
Section 3.1 - Data GovernanceHow BuildStack handles your data
BuildStack is operated by Caradon AI Ltd (trading as BuildStack), registered in England and Wales (company number 16701624).
- Data storage: All project data, issue logs, and uploaded photographs are stored securely on Railway cloud infrastructure (US East region). Data is encrypted in transit (TLS) and at rest.
- Photo storage: Uploaded photographs are stored on Cloudflare R2 object storage with private access controls. Photos are not publicly accessible.
- Data sent to Anthropic: When a surveyor runs BuildStack Analysis, the defect description, location, and any uploaded photographs are sent to the Anthropic API for processing. Anthropic's API data handling policy governs the use of this data. By default, data submitted via the Anthropic API is not used to train Anthropic's models. BuildStack users should review Anthropic's API privacy policy before uploading client data containing personal information.
- Data retention: Project data is retained for the duration of the user's account. Users may delete projects and issues at any time. Account deletion requests can be made to support@getbuildstack.co.uk.
- UK GDPR: Caradon AI Ltd is the data controller for data held on the BuildStack platform. Our Privacy Policy is available at getbuildstack.co.uk/privacy.
Recommendation for RICS-regulated firms: Firms should consider whether client property data (addresses, photographs) constitutes private and confidential information under the RICS standard and take appropriate steps in their own data governance policies before uploading such data to BuildStack.
Section 4.1 - Environmental ImpactEnvironmental impact of BuildStack's AI system
BuildStack uses cloud-based AI inference via the Anthropic API. No on-premise compute infrastructure is operated by Caradon AI Ltd.
The environmental impact of AI inference is primarily associated with the energy consumption of Anthropic's data centres. Anthropic publishes information about its environmental commitments and approach to responsible AI development at anthropic.com.
BuildStack does not operate its own model training infrastructure. The environmental footprint of BuildStack's AI functionality is limited to API inference calls made when surveyors run analysis - typically a small number of requests per project.
Section 4.1 - LiabilityLiability position
Caradon AI Ltd's liability position is set out in the BuildStack Terms and Conditions available at getbuildstack.co.uk/terms.
In summary:
- BuildStack outputs are provided as draft assistance only and do not constitute professional advice
- The responsible surveyor retains full professional liability for all outputs issued to clients
- Caradon AI Ltd excludes liability for losses arising from reliance on unreviewed BuildStack outputs
- BuildStack carries no professional indemnity insurance covering the surveying services of its users - users must maintain their own professional indemnity cover
RICS-regulated firms should note that their professional indemnity insurance position regarding AI-assisted outputs should be confirmed with their insurer.
RICS Standard - Compliance SupportHow BuildStack is designed to support compliance
BuildStack has been designed with the RICS AI standard in mind. The following features support compliance:
Professional review requirement: All BuildStack outputs carry a prominent ‘DRAFT - NOT FOR ISSUE’ watermark in exported documents. This watermark must be manually removed by the surveyor before the document is issued to a client, creating a natural review checkpoint.
Disclaimer on every output: Every exported document includes a disclaimer confirming that outputs are draft assistance requiring professional review and sign-off.
Output transparency: BuildStack displays all generated content within the platform before export, allowing the surveyor to review, edit, and approve each element of the analysis and repair scope.
Sign-off before export: BuildStack requires the responsible surveyor to confirm their name and accept professional responsibility for the content before any document is downloaded. This creates the written record of responsibility required by section 4.2 of the RICS standard.
Opt-out: Surveyors may log issues and produce reports without using BuildStack Analysis at any time. The analysis fields are optional and documents can be exported using only manually entered content.
Data handling: BuildStack does not share project data with third parties other than the Anthropic API for analysis processing and Cloudflare R2 for photo storage.
Due Diligence EnquiriesDue diligence requests
RICS-regulated firms requiring written responses to due diligence requests under section 4.1 of the RICS standard may contact Caradon AI Ltd at:
We will respond to written due diligence requests within five working days.
Send due diligence request
Caradon AI Ltd (trading as BuildStack) · Company No. 16701624 · getbuildstack.co.uk · This page was last updated May 2026