This portfolio documents how I assess AI use cases and work through architecture, governance, security, cost and implementation decisions. The Architecture Decision Framework is available now. Further public evidence will be added as it is completed.

Approach

The use case comes first

Some problems need retrieval-augmented generation or an agentic workflow. Others need simpler automation, better information architecture or process redesign with no AI in it. The useful work is deciding which route fits, then designing the controls and operating model around it.

That means understanding the business problem, the systems involved, the data boundary, the consequences of a wrong answer and the people who remain responsible.

Available

Architecture decision framework

AI is one option among several. The decision framework maps use-case characteristics to outcomes including process redesign, conventional automation, prompt-based AI, RAG, agentic workflows and human-in-the-loop patterns.

It makes the trade-offs visible rather than hiding them behind a tool recommendation.

Read the framework →

What the framework helps decide

  • Process redesign or conventional automation
  • Prompt-only AI or retrieval-augmented generation
  • Predefined or agentic workflows
  • Human control and action authority

Planned outputs

Patterns, controls and economics

Planned

Architecture pattern library

Reusable patterns for RAG, agents, human review, AI gateways and the boundary between workflow automation and AI.

Planned

Governance toolkit

Practical controls for ownership, risk assessment, human review and the lifecycle of an AI-enabled service.

Planned

Cost and value model

A way to test whether a use case is worth building once model usage, integration, operations, assurance and change are included.

Planned

Case studies

The case studies will connect the full chain: business problem, architecture decision, risks, controls, cost, implementation approach and reflection. They will use public or synthetic information and show what was designed, built or tested, including the assumptions and what did not work.

About the work

Architecture grounded in delivery

I lead a technology team and run a consulting practice alongside it. My background is in Microsoft 365 architecture, workflow automation, systems integration and the operating decisions around AI. I am building this portfolio because enterprise AI architecture cuts across all of those areas.

Each output will include the reasoning behind it, not just the finished diagram or recommendation.

Consulting

Working through an AI use case?

Send me the business problem, the systems involved and what the AI would be expected to do. If there is a sensible piece of architecture work behind it, we can discuss where to start.

consulting@joshwickes.com