AI for Architects Training: A Practical Learning Plan

An AI training plan for architects works best when it is built around tasks architects actually do: researching precedents, drafting program requirements, and reviewing design proposals. Starting with those tasks gives you something concrete to evaluate. Tools like fenbs let you keep the prompts that worked and the lessons from each exercise across sessions, so practice builds on itself rather than starting over.

Training does not mean reading documentation. It means trying a real task, judging the output against your professional knowledge, and adjusting. An architect who has reviewed ten AI-drafted site analysis summaries knows where the gaps usually appear. That pattern recognition comes from practice, not theory.

What a useful training plan covers

Architecture work spans research, communication, coordination, and design judgment. AI can assist with all four, but the assistance looks different in each area. A good training plan touches each area separately so you know where AI adds real value and where it requires more oversight.

• Research: Precedent studies, building code summaries, zoning lookups, material properties, and climate data.
• Brief writing: Program requirements, spatial relationships, adjacency diagrams described in text, and client communication drafts.
• Design review: Annotating design rationale, drafting project descriptions, and summarizing design evolution for client presentations.
• Coordination: Meeting notes, RFI draft summaries, and scope-of-work language for consultants.

Work through each area with exercises drawn from a real project you know well. That familiarity means you can judge accuracy immediately rather than having to research whether the AI answer is correct.

Phase 1: Research and precedent review

Start by asking an AI assistant to summarize a building type you have already studied: a courthouse, a community health center, a transit hub. Compare the summary against what you know. Notice where it is accurate, where it oversimplifies, and where it is wrong. This exercise builds the habit of reviewing AI research outputs rather than accepting them.

Move on to code lookups. Ask the assistant to summarize occupancy requirements for a building type you are working on. Compare the answer against the actual code text. AI assistants can miss jurisdiction-specific amendments, interpret requirements too broadly, or confuse edition years. Catching those gaps early is the point of this phase.

Write down the prompts that gave useful starting points and note what follow-up questions were needed. A prompt that works once in research will probably work again. Save it with a note about what it was good for and where it needed verification.

Phase 2: Brief writing and program development

Take a real program from a past project and ask an AI to draft a written program statement from the room counts and adjacency requirements. Review the draft for accuracy, completeness, and tone. The goal is not to use the output directly but to understand how much briefing the AI needs before its draft is useful.

Practice giving the AI a client description in plain language and asking it to suggest a preliminary program. Read the suggestion critically. Does it reflect the client's priorities? Does it miss constraints you know matter, like budget or site area? This exercise builds the skill of briefing an AI precisely, which matters more than the quality of any one output.

Phase 3: Design review and documentation

Ask an AI assistant to draft a design rationale paragraph for a project approach you have already developed. Review it for accuracy and clarity. Design rationale needs to reflect specific decisions made for specific reasons, not generic statements about architecture. The review teaches you what information an AI needs to write something useful.

Practice drafting project descriptions for public presentations. Give the AI the key facts: program, site constraints, structural approach, material palette. Ask for a 200-word project narrative. Edit the result. You will quickly find the level of input the AI needs to produce something you would actually use, and the editing itself builds faster review habits.

How to review what AI produces

Reviewing AI outputs is a professional skill as much as generating prompts is. The review habit is what turns a training exercise into safe professional practice. Ask these questions each time you review an output:

• Is every factual claim something I can verify? Code citations, material properties, and precedent claims all need checking.
• Does this reflect the actual constraints of this project, or did the AI generalize from a different context?
• Would a colleague reading this notice anything missing or misleading?
• Did the AI rely on information I gave it, or did it fill gaps with assumptions I cannot trace?

Flag outputs that passed your review as good examples. Collect them alongside the prompts that produced them. Over time, the collection shows you which tasks AI handles well and which require more input or more oversight. That knowledge is more valuable than any individual output.

Common mistakes when learning AI for architecture

The most common mistake is accepting AI research without checking primary sources. Building codes, zoning bylaws, and product specifications change. An AI answer that cites a code may reflect a prior edition or a different jurisdiction. Always trace a code claim back to the source document before using it professionally.

A second mistake is training on unfamiliar project types. If you cannot judge the accuracy of the output, you cannot learn from the exercise. Start with projects you know well, then extend to less familiar territory once you have calibrated how the AI performs on your core work.

A third mistake is skipping the recording step. A training exercise that leaves no trace contributes nothing to the next session. Write down what worked, what did not, and why. That record is the product of the training, not any single AI output.

How to build your AI training plan with fenbs

A training plan is a small project. It has phases, exercises, prompts worth keeping, and lessons worth recording. Start with one phase and one real project as the test case.

1. Create a free account and set up a project called something like "AI Training Plan." Add the three phases as tasks in To Do.
2. Before each exercise, read the Lessons learned you have already recorded. If this is your first session, write the constraints you expect to matter: project type, jurisdiction, scale.
3. Run the exercise and add your findings as a comment on the task. Move the task to In Progress while you are working through the phase.
4. After the exercise, record a Lesson with what you tried, what the AI got right, and what needed correction. Save the prompt in Skills and prompts with a note about when it is useful.
5. Before you stop, update Where we left off for the project so the next session starts from the right place.

After three or four sessions, review the lessons together. Patterns will emerge: AI tends to oversimplify this type of code lookup, or AI needs explicit site constraints to avoid generic suggestions. Those patterns are the real output of the training plan. Share them with colleagues working through the same process.

For a broader picture of how AI fits the design professions, see AI for architects and designers and AI for landscape architects. The AI by role topic covers other professions using the same approach.