AI Data Handling Notice

This notice covers managed AI extraction workflows in Beacon Build, including historical library AI import and any related document-intake workflow that asks you to approve the current AI data handling notice.

The current notice version is 2026-05-17.1. Approval is recorded once per user, company, document type, and version. Once approved, the same current-version approval is recognized anywhere the AI import acknowledgement is required.

When you use managed AI extraction, the source file and relevant extraction instructions are sent to Beacon managed AI Agent so the uploaded document can be converted into draft structured fields for review.

Source files may include PDFs, spreadsheets, images, screenshots, bid summaries, cost reports, pro formas, or other project documents that you choose to upload for extraction.

Large text, CSV, and spreadsheet-derived sources may be split into managed intake chunks. Chunk processing is used to preserve source locators, reduce context loss, and reconcile candidate outputs behind the scenes.

You are responsible for confirming that you are authorized to upload the file and that sending it for extraction does not violate confidentiality, contractual, regulatory, or other legal obligations.

Beacon managed AI Agent is the product label for Beacon Build-managed AI extraction. The underlying providers may include OpenAI ChatGPT as the primary managed provider, Anthropic Claude as a managed fallback, and Google Gemini only where enabled.

OpenAI is treated as the primary managed provider for supported source types, including PDFs. Anthropic Claude may be used as a fallback when the primary provider is unavailable or cannot complete a managed extraction. Google Gemini is not part of the managed workflow unless it is separately enabled.

The extraction uses Beacon Build-managed provider credentials. You do not need your own provider subscription or API key to use this managed workflow while it is enabled for your workspace.

Under the current API policies of the providers we support, API inputs and outputs are not used for model training by default. The relevant provider API data-handling policy, privacy commitments, and terms still apply to the extraction request.

Beacon Build may retain the uploaded source file, processed file metadata, extraction summary, managed agent, provider and model metadata, source-location evidence, review status, and related audit records inside the company-isolated workspace.

For spreadsheets, the managed workflow may inspect sheet names, row and column patterns, formulas, named source areas, hidden or repeated source areas, and nearby labels to help identify messy or untitled fields. This inspection is used to build extraction context and source locators; it does not make draft values authoritative.

Beacon Build may retain managed intake sessions, chunk locators, and server-owned extraction candidates so a later acceptance step can verify that accepted memory came from retained source evidence.

This retained evidence exists so authorized users can review where extracted values came from before relying on or promoting the resulting record.

The structured fields saved from extraction remain draft information until a user reviews, corrects, and saves them into the relevant workspace area.

Customer project, estimate, budget, and cost data stays isolated to the customer company account.

Uploaded source evidence and extracted fields are not sold, disclosed to other customers, or used in another customer's workspace.

Customer pricing information, uploaded source evidence, and extracted fields remain scoped to the company workspace.

Service diagnostics may be used to improve reliability where they do not reveal a customer, company, project, source document, bid price, vendor pricing, or source line item.

Managed AI extraction creates a draft only. It can misread, omit, classify, or normalize information incorrectly.

You remain responsible for reviewing the source evidence, correcting extracted fields, and confirming the record before saving it or using it for benchmarks, estimates, bids, budgets, or other decisions.

Beacon Build does not treat an AI-extracted value as authoritative simply because it was produced by Beacon managed AI Agent.

This notice is workflow-specific and applies wherever the managed AI extraction feature is used. It does not replace the Privacy Policy, Terms of Service, or any mandatory privacy rights that apply in a user or company jurisdiction.

Beacon Build separates United States and Canada product assumptions where country matters, including currency, geography labels, market packs, escalation sources, benchmark cohorts, tax or fee context, and privacy/legal considerations.

Where uploaded documents contain personal information, the regional privacy sections in the Privacy Policy continue to apply. Company administrators remain responsible for confirming that the source file can be sent for the requested extraction workflow.

Company administrators may disable external AI processing for a workspace where that control is available. If disabled, managed AI extraction will not process uploaded documents for that company until the setting is changed by an authorized administrator.

If this notice materially changes, Beacon Build may require approval of a new notice version before managed AI extraction can continue.

Questions about this notice can be sent to privacy@beaconbuild.app.