Architecture Portal

Architecture and AI

Using AI in architecture work, and designing systems that use it.

In development

Adrian Sutherland · Version 1.1 · · © 2005–2026

Artificial intelligence (AI) raises two different questions for architects: how can it help with architecture work, and how should a service that uses it be designed? This section covers both.

The design guides follow a fictional support assistant. They explain how the application finds the rules and facts it needs, checks a recommendation and controls any change to a business record. Research examples explain the reasons for the advice. The practice guide covers briefing an assistant and reviewing its contribution to research, design and delivery.

Design a system built from prompts and skills

Start here if you are combining instructions, reusable procedures, sources and tools to produce an architecture review, a software change or a support recommendation. Read the three guides in order for a complete example, or choose the decision you need to make.

Guide What you should take away
1. AI-system blueprint Component responsibilities; where instructions, rules, facts and state belong; a small advisory starting point and a comparison of ways to obtain context.
2. AI worked example One support request traced through selected rules and facts, the returned proposal, validation, and an optional approved action with recovery.
3. AI regression testing A change from prompt version 3 to 4, the tests selected for that change, a call budget, checks of the test selection and the release decision record.

The example defines a fictional support service. It has not been implemented, and its proposed model tests remain unrun.

Use AI in architecture work

AI-supported architecture explains how to brief an assistant, check its sources and review the work it produces. Use it when AI helps with research, design, delivery, governance or service support.

These shared task collections sit under All templates. They pair a human brief with an AI prompt example for architecture practice; they are not components or ready-made test suites for the AI-system blueprint:

Your task Briefs, prompts and worked examples
Review principles, responsibility or release readiness Governance
Check requirements, tests or a proposed change Delivery and quality assurance
Diagnose a problem, examine a runbook or assess recovery Service support

Connect the AI design to the wider service

The AI design is part of a wider service. The following links lead to the Portal’s general architecture guidance:

About this edition

This October 2026 guide connects the Portal's AI-supported architecture practice with its AI-system blueprint, example and testing strategy. Linked pages retain their own editions.

Scope, limitations and next checks

Status and accountability

Read this page with its boundaries visible

Status

In development

Last reviewed

Intended users

  • Architects using AI in their work and designers assembling domain AI systems

Non-goals

  • A prescribed product stack or a replacement for the wider architecture method

Limitations

  • Design guidance and adaptable examples. The fictional support service is unimplemented and its proposed model tests remain unrun.

Next evidence sought

  • Apply the guidance to a real task and review the outcomes, effort and limits.