What Construction Professionals Must Know Before Uploading Project Documents to AI Tools
📋 What You Will Learn In This Article:
- Why the upload button is one of the most commercially risky actions in your daily workflow
- Which construction documents require careful consideration before sharing with AI tools
- Why turning off model training does not answer the full data question
- How to get useful AI assistance without sharing live project information
- What the RICS AI standard requires from surveying professionals
- A practical pre-upload checklist you can use right now
Before You Upload That BOQ, Contract or Claim File
AI can help with construction paperwork. Deciding what information you are allowed to share with it comes first.
Before you upload that BOQ, subcontract or claim file to an AI tool — do you know who else can see it once it leaves your screen?
This is not a question about whether the AI tool is useful. It almost certainly is. It is a question that comes before the upload button: are you authorised to send this particular information to this particular service?
A document can be technically easy to upload and still be professionally inappropriate to share. The file transfer takes seconds. The contractual and professional consequences of sharing the wrong information with the wrong service can take considerably longer to resolve.
This guide works through the practical decision process for QS professionals and construction commercial teams — what to check, what to avoid, and how to get genuine AI productivity benefits without creating unnecessary risk.
The Short Answer
Do not upload live project documents by default. First check the information involved, your authority to disclose it, your firm's approved tools and the relevant account settings. Where possible, use a fictional example or the minimum approved information needed for the task.
Start with Permission, Not the Upload Button
A sensible review begins with the task. If you only need a letter structure, there may be no reason to upload the contract. If you need a checklist for assessing a variation, a fictional description may be enough.
When the task genuinely requires project information, work through the following questions before sending anything.
Should This Document Go into an AI Tool?
Is the tool and workspace approved?
If not, stop and check with the responsible manager, IT or information-security team before proceeding.
Are you authorised to share this information?
Check employer policy, client requirements, confidentiality obligations, personal data and any relevant licence restrictions.
Can the task be done with less information?
Prefer a fictional scenario, a blank template or a limited, approved extract over the complete project file.
Have you checked the account and data handling?
Confirm training settings, retention, access, sharing and any connected third-party services for the specific workspace.
A BOQ Is Not Just a List of Quantities
Construction documents often contain more information than the visible page suggests. A priced BOQ may reveal a commercial position. A progress photograph may identify a location or individual. A claim file may expose the reasoning behind a dispute.
Use the examples below as prompts for a review, not as automatic permission to upload any particular document.
| Document | Information to look for | Lower-disclosure starting point |
|---|---|---|
| Priced BOQ or estimate | Tender rates, supplier prices, mark-ups, project identity and pricing assumptions. | Use fictional rates and a small sample when asking for formatting or calculation logic. |
| Subcontract or agreement | Party details, signatures, negotiated terms, contact information and confidential schedules. | Ask for a general review checklist first. Use real extracts only through an authorised process. |
| Claim submission | Commercial strategy, entitlement arguments, admissions, records and settlement positions. | Request a claim structure using a fictional event rather than uploading the live dispute file. |
| Payroll or labour records | Names, identifiers, salary details, bank information and other personal data. | Use a fabricated dataset for a spreadsheet demonstration. Seek specialist approval for actual records. |
| Drawings and BIM information | Proprietary design, restricted locations, client details and intellectual-property restrictions. | Use public or purpose-made sample material with suitable reuse permissions. |
| Meeting minutes or emails | Personal information, internal decisions, commercial discussions and attachments. | Create fictional discussion points for a minutes template. |
Removing the Project Name Is Not a Complete Check
Also inspect comments, tracked changes, hidden sheets, file properties, signatures, embedded images and contextual clues. Renaming parties "Contractor A" and "Client B" does not necessarily remove the sensitivity of the commercial information.
Not Used for Training Does Not Mean No Data Is Held
This distinction is easy to miss. Model training, storage, access and permission to disclose are separate questions.
Personal ChatGPT Accounts
OpenAI says that turning off "Improve the model for everyone" prevents new conversations from being used to train its models. Those conversations can still remain in chat history.
Business and Enterprise Workspaces
OpenAI says business data is not used for model training by default. The applicable workspace controls, contractual terms, retention arrangements and organisational policies still need to be checked.
Source: OpenAI — Data controls in ChatGPT and OpenAI — Enterprise privacy. Check the current documentation because settings and product terms can change.
OpenAI also states that Temporary Chats are not used to improve its models and may be retained for up to 30 days for safety purposes. Temporary Chat should therefore not be treated as a promise that no data is retained.
None of these settings gives an employee permission to disclose a client's information. A paid subscription is not, by itself, an approval to upload project records.
Ask two separate questions:
1. What does this service do with the data?
2. Are we allowed to send this particular data to it?
A QS Example: Get the Structure Without Sharing the Dispute
Suppose you need help organising a response to a variation proposal. The following example is fictional. It shows how to request a useful structure without sending the live commercial file.
❌ Unnecessary Disclosure
"Here is our signed subcontract, the subcontractor's priced proposal and our internal negotiation note. Draft a rejection using our confidential settlement position."
✅ Lower-Disclosure Request
"Using a fictional variation scenario, prepare a response structure covering scope, instruction, entitlement, quantities, rates, supporting records and next actions. Do not decide entitlement or invent contract clauses."
The second request does not answer the real contractual question. It provides an outline that a QS can complete and check within the approved project environment.
That limitation is useful. It keeps the drafting assistance separate from the assessment of actual entitlement.
A Copy-Ready Prompt for a Blank Review Template
Create a blank variation review template for a Quantity Surveyor. Include sections for event description, instruction reference, contractual basis, scope comparison, quantities, rates, time implications, supporting records, assumptions and reviewer approval. Use placeholders only. Do not invent project facts, clause numbers, prices or conclusions on entitlement.
What Does the RICS Standard Add?
RICS's "Responsible use of artificial intelligence in surveying practice" standard came into effect on 9 March 2026. Its stated areas include data governance, system governance, risk management, AI procurement, output reliability and communication with clients.
RICS explains that the standard applies to AI use that has a material impact on the delivery of surveying services. It is not a statement that every administrative use carries the same requirements.
For relevant RICS members and regulated firms, the full standard should be read and applied to the actual use case. Other readers can use it as professional context, but should not assume it automatically governs their organisation.
Source: RICS — Responsible use of AI in surveying practice.
Your Pre-Upload Checklist
Before sending project information to an AI service, check the following. These boxes are a reading aid only: they are not saved and do not create a formal approval record.
Completing this checklist does not certify legal or contractual compliance. If an answer is unclear, pause and ask the responsible manager, information-security team, data-protection adviser or legal adviser.
Review the Answer as Carefully as the Upload
An approved upload does not make the result correct. RICS identifies invented or false AI output, lack of contextual understanding and misinterpretation of instructions among the limitations and failure modes to consider.
For QS work, review the facts, document revision, clause references, dates, quantities, calculations and assumptions. An answer can be well written and still misunderstand the instruction or omit an important qualification.
Do not let a polished draft move directly into formal correspondence without the normal technical and commercial review.
What If You Have Already Uploaded a Sensitive File?
Do not assume that deleting the conversation settles the matter. Follow your organisation's incident procedure.
- Stop sharing further project information through that route.
- Report the issue promptly to the responsible manager or information-security team.
- Record the relevant facts: tool, account, date, file and any sharing or connected-service settings involved.
- Follow authorised instructions for containment, deletion requests, evidence preservation and any required notifications.
- Do not send copies of the sensitive file into additional tools while trying to investigate the problem.
Provider retention and deletion arrangements vary. The responsible team should assess the actual service terms and circumstances rather than rely on an assumption about how the tool works.
Use AI to Reduce Work, Not to Bypass Responsibility
Start with a blank template or fictional scenario when that will do the job. Where real project information is necessary, use the approved environment and obtain the required authority. The upload button is a technical feature, not a permission.
💬 Found This Useful?
If this article helped you, please consider:
- ⭐ Sharing on LinkedIn — every QS professional using AI tools needs to read this
- 💬 Leaving a comment below — does your firm have a policy for AI data handling yet?
Does your firm have a clear approved AI workflow for handling project documents? Share your experience in the comments!
Frequently Asked Questions
Can I upload a contract if I remove the names?
Not automatically. Negotiated terms, prices, project details and other contextual information may still be confidential. Check authority, restrictions and the approved handling process.
Does turning off model training make an upload acceptable?
No. It changes how new conversations are used for model training, subject to the provider's current terms. It does not establish your right to disclose the information or remove every storage, access and retention consideration.
Is a business AI account enough on its own?
No. Business workspaces can provide different data protections and controls, but the proposed use must still comply with the organisation's approvals and applicable obligations.
What is a practical way to start using AI for QS work without sharing live project data?
Begin with fictional scenarios, blank review templates, generic checklists and writing structures. Keep live commercial facts out of the task until an approved process is established. Our Free AI Prompt Generator provides ready-made prompts for common QS tasks that work without uploading sensitive project documents.
What does the RICS AI standard require?
RICS's "Responsible use of artificial intelligence in surveying practice" standard, effective from 9 March 2026, covers data governance, system governance, risk management, AI procurement, output reliability and client communication. It applies to AI use that has a material impact on the delivery of surveying services. RICS members should read and apply the full standard to their actual use case.
A Question for Your Team
Does everybody know which AI workspace is approved and what information may be shared? If not, that is a useful first discussion before the next document is uploaded.
Editorial note: This article provides general professional awareness, not legal advice or a complete compliance procedure. Product settings and professional requirements may change. Check the linked primary sources and your organisation's current policies.
Take Your QS Workflow to the Next Level
Choose the free tool that fits your current task:
Comments
Post a Comment