ChatGPT Prompts for Architects: Briefs and Coordination

Useful ChatGPT prompts for architects organize client requirements, separate meeting decisions from proposals and prepare questions for professional review. Give the model approved information, a project phase and a clear output format. Keep code interpretations, technical judgments and approval with the responsible people.

The same procedures often repeat across projects. fenbs can hold reusable prompts and human-approved project rules so connected assistants read the current instructions. Use fictional examples when developing templates and keep sensitive project records in the systems your practice approves.

What should an architectural prompt specify?

Start with the project phase and the action you want to support. Organizing early client questions is different from reviewing coordination notes before an issue. State the building context, known constraints, intended audience and information the assistant is permitted to use.

Identify the source and revision of every supplied excerpt. A question about a superseded drawing should not be treated as a current design instruction. If information is absent or inconsistent, ask the model to flag it and identify who needs to resolve it.

OpenAI’s prompt guidance↗ (opens in a new tab) recommends clear instructions and refining them after review. Make that practical by requesting one bounded deliverable, such as a list of unanswered programming questions, rather than a complete project strategy.

• Purpose and phase: programming, concept discussion or meeting follow-up.
• Approved inputs: requirements, notes or excerpts with source and revision labels.
• Boundaries: no invented dimensions, commitments, consultant responses or approvals.
• Output: a question list, coordination table or draft summary for human review.
• Escalation: the person or discipline that must resolve each uncertainty.

How do you turn client goals into a question list?

Ask the assistant to distinguish requirements from preferences and assumptions. A client’s comment about a bright lobby may describe an aspiration rather than a measurable criterion. The useful next step is a question that helps the design team understand what the person means.

Organize this fictional client project summary for discussion.
Project type and phase: [fill in]
Client goals: [approved, nonconfidential summary]
Known constraints: [list with source labels]

Return: confirmed requirements; stated preferences; assumptions;
contradictions; questions for the client and design team.
For each question, explain which design decision it informs.
Do not invent room sizes, occupancy, budgets, schedules, code
requirements or approvals. Preserve the source of each stated
requirement. Label proposed success criteria as proposals for
a person to approve, never as agreed project requirements.

Imagine a fictional small library asking for both quiet study and flexible community events. The output should surface the possible conflict and ask about use patterns, rather than assume that a particular partition system resolves it. A person must confirm priorities before the team treats them as design direction.

Review the question list with the project lead and keep the approved answers with their reasons. The general work prompt templates provide other intake structures, but retain the distinction between a client request and a design team assumption.

Which prompt helps organize meeting notes accurately?

Give the model notes that preserve who said what. Without that distinction, a suggestion can become an apparent instruction and a discussion date can become an invented deadline. Request a draft that leaves unanswered ownership and approval questions visible.

Prepare a draft meeting summary for participant review.
Meeting purpose and date: [fill in]
Approved notes: [items labeled N1, N2... with speakers]

Return separate lists: decisions explicitly made by a person;
proposals still open; follow-up actions; unresolved questions.
For every item include its note ID. Use only owners, dates and
approvals actually stated. Otherwise write "To confirm."
Do not change project scope, infer client consent or turn an
action to investigate into permission to implement. Flag
conflicting statements for the project lead to resolve.

Have the participants review the summary under your practice’s normal procedure. Check that the description preserves conditions and exclusions. “Investigate shifting the entrance” must stay an investigation unless a person actually authorized a change.

The AIA Code of Ethics↗ (opens in a new tab) includes obligations concerning competence, confidentiality and client consent for material changes to project scope or objectives. It applies to AIA members; also consult the laws, regulations and professional requirements governing your work.

How can you prepare a focused design review?

Use ChatGPT to organize review questions around the decisions that remain open. Supply accurate descriptions or approved excerpts, and identify the disciplines whose input is still needed. The model should make the review easier to run, rather than appear to complete technical coordination.

Prepare questions for a human-led design review.
Project phase and review scope: [fill in]
Current design description: [approved summary]
Sources and revisions: [list]
Known decisions and constraints: [list]
Open consultant items: [list]

Return a table: issue; relevant source; question; discipline
or person needed; information missing; next review action.
Separate client intent, spatial coordination and technical
verification. Do not certify compliance, calculate capacity,
resolve technical conflicts or invent a consultant’s answer.
Flag any conclusion that the supplied material cannot support.

For the library example, the agenda could ask whether event setup interrupts the quiet study route and which operational information is missing. It should not assert that an exit arrangement, acoustic assembly or structural solution is adequate. Those questions need appropriate drawings, evidence and professional review.

Record the choices people actually make in fenbs’s Decisions and rules, including why and what still needs checking. That helps the next assistant respect a settled requirement without treating an unresolved consultant item as finished.

How should you handle code-related questions?

Ask for a research checklist, not a compliance verdict. The relevant question is which adopted provisions, amendments, project facts and interpretations need confirmation. A generic model-code answer cannot establish what governs a particular project.

The ICC’s code adoption resources↗ (opens in a new tab) explain adoption and amendments. Confirm the applicable edition, local changes and requirements with the responsible jurisdiction and project team. Keep verified references and the person who checked them visible in your approved project record.

Create a research checklist only, not a compliance assessment.
Jurisdiction and project description: [fill in]
Verified adopted edition and amendments: [supply or mark unknown]
Question: [narrow issue]
Approved source excerpts: [S1, S2... or none]

List information to confirm, relevant source locations supplied,
questions for the responsible professional, and questions for
the jurisdiction. Distinguish model-code text from verified
local requirements. If sources are absent, do not invent
section numbers, dimensions, exceptions or a compliance answer.

Independently open each source and check its applicability. If the jurisdiction or project facts are unknown, the useful output is a list of gaps. Do not let the apparent completeness of a table substitute for a technical check or a documented interpretation.

How do you maintain a reliable prompt library?

Give each procedure a purpose, input list, prohibited data, expected result and review owner. Test it with a contradictory source revision or an action missing an owner. A useful template preserves that uncertainty and routes it to a person. Keep that fictional test case with the template so a later revision can be checked against the same missing information before the team adopts it.

Save the generic lesson when a review catches a mistake. For example, require every meeting item to retain its source note because a previous summary blurred a suggestion and a decision. Keep the lesson general, without copying sensitive negotiations or consultant records.

The system prompt examples cover organizing standing instructions. For contract-oriented drafting, use the separate legal prompt templates with appropriate attorney review. Architectural coordination should not absorb those legal questions into an apparently complete meeting summary.

How to do this with fenbs

1. Create a free account and add a project for approved coordination procedures.
2. Save one generic prompt in Skills and prompts with fictional input and its human review checklist.
3. Record human-approved project rules and their scope in Decisions and rules, keeping sensitive documents in approved systems.
4. Follow the integration guide for your supported assistant and ask it to read the project rules and prompt before drafting.
5. Save each open follow-up and use Where we left off to identify the next question or professional review needed.

fenbs does not replace architectural drawings, building information modeling software or code review. Use it for permitted procedures, decisions and follow-ups around that work. The AI prompts hub has more ways to define repeatable assignments.