Selected Work

Built,
not pitched.

Every engagement produces something real. These are working AI systems — not decks, not prototypes. Each one built to demonstrate what becomes possible when organisations stop waiting.

Explore
01 / 04
Content Intelligence

Self-Improving
Writer

An AI writing system that gets better with every attempt. Each output is analysed for gaps by an Overseer agent, which feeds corrections into the next iteration — turning a single prompt into a compounding loop of refinement.

See it work
AI OFFICELearning LoopAttempt 3 · 5 gaps
Input Text
Our quarterly review process involves gathering data from multiple departments, synthesising performance metrics, and generating a report for executive review...
Active Gaps
Lacks measurable success criteria
No timeline mentioned for review
Missing stakeholder sign-off steps
Generate →
✓ Attempt 3 — Accepted
Markdown Output
Overseer Analysis
Add success metrics per department
Specify review cadence
02 / 04
AI-Native Tooling

Product
Development Studio

An AI-native IDE that walks a product from raw idea to deployment-ready spec across ten structured stages. Each stage uses Claude to generate, validate and summarise — with full lifecycle visibility and token-level transparency.

See it work
artiface/ my-productPHASE 1
Lifecycle
01
Ideation
02
UI/UX
03
MVP Scope
04
Func. Spec
05
Build Spec
STAGE 01 / 10
Ideation
Capture the raw idea. What problem are you solving?
What data problem are you solving?
We have siloed analytics across 4 business units with no unified view...
Understood — a classic data mesh challenge. Let me scope the ideation stage...
Ask Claude about this stage...
send
Status
in progress
Progress
1 / 10
Est. cost
$0.024
03 / 04
Organisational Intelligence

AI
Office

A live intelligence system that watches an organisation think. Five specialist agents run in a continuous loop — detecting gaps, researching solutions, proposing improvements, building, and evaluating outcomes. The organisation becomes the model.

See it work
AI OFFICECONSOLELIVE
Agents Active
5
Gaps Resolved
24 ↑12%
Loop Cycles
147
Knowledge Items
312 ↑
Agent Log
[Observer] Detected pattern: invoice processing delay >48h in 3 depts
[Researcher] Analysing 47 docs for root cause...
[Designer] Proposing automated approval routing
[Builder] Workflow deployed to staging
04 / 04
AI Evaluation

AI Benchmark
Suite

A rigorous testing environment that runs AI agents against the GAIA benchmark — the international standard for measuring real-world task performance. Used to evaluate, compare and prove the capabilities of AI systems before deployment.

See it work
AI OFFICE — GAIA RUNNERLevel 1
73%
Overall Score · 22/30 tasks
Task Results
2024-task-001
✓
2024-task-002
✓
2024-task-003
✗
2024-task-004
✓
2024-task-005
…
4
Working systems
0
Slide decks
∞
Possible starting points

What would we build
for your organisation?

One session. Three deliverables. Something working by the end of day one.

Start the conversation