A software-architecture practice for the AI era

Systems your team can
build, run and own

Three decades of production-system experience, applied to the hard decisions AI now brings.

Getting AI to work in a demo is the easy part; the hard part is everything after.

Through workshops and principal-led consulting, we help teams close that gap – and build the capability to handle what comes next.

AI keeps getting more powerful. Accountability stays human.

Consulting and advisory

Have a problem to solve? We work alongside you on a live system, architecture, or decision. We’ll start with the smallest useful engagement.

Production-Ready AI

When a live system needs senior intervention.

A stalled PoC, a difficult architecture or security decision, or a system that needs to cross the gap between convincing demo and dependable production. We diagnose, architect, build where needed, and hand the capability back to your team.

Production-Ready AI →

Architecture Advisory

When the hard decisions keep coming.

Regular senior support for architects, engineering leads and technical leaders carrying calls they will have to live with. Bring the live decision; leave with the reasoning, design and decision record.

Architecture Advisory →

Workshops and briefings

When the capability needs to live inside your people. Practice-based: your team works on its own systems, code, and decisions – the person, the team, and the project.

The project

Agentic AI for Products & Systems

For technical leaders deciding whether an agent belongs at all.

Agent demo → production decision

See the workshop →

The team

AI in the SDLC

For teams adopting AI across delivery without giving away control.

Faster team → accountable team

See the workshop →

The person

Software Architecture: Principles, Patterns & Practice

For senior engineers taking on architectural responsibility.

Senior engineer → architect

See the workshop →

Explore workshops and briefings →

Examples of problem solving

Three examples from real engagements where the obvious answer was not the right one.

  • Don’t automate the human who still owns the decision. In legal document review, the architecture was designed around the lawyer’s review workflow rather than pretending autonomous review was the goal. Read the field note →
  • Don’t build the feature if the reusable capability is the real asset. A sales assistant became a shared knowledge engine supporting multiple products rather than another isolated chatbot. Read the field note →
  • Don’t vectorize the warehouse just because RAG looks like the easy answer. For an enterprise analytics assistant, we rejected a proposal to embed the entire warehouse and instead used schema-aware SQL generation, splitting exploratory questions from must-be-right ones. Read the field note →

Recent work includes production systems for Fortune 500 and FTSE-100 companies.

See the case studies →Read the field notes →

Our approach

Tools change quickly. The hard questions don’t. Across consulting, advisory and workshops, we use the same five-step discipline:

Frame → Options → Size → Decide → Record

The detail changes with the problem. The principle does not: match the depth of reasoning and verification to what is at stake. See details of our approach →

Facing a hard architectural decision?

Send us the problem. You'll get an honest assessment.

Request an assessment