Before You Deploy
An interactive field guide to building dependable AI
Build it. Break it. See what it takes.
Explore the decisions behind dependable AI. Design your system, see the trade-offs and test it under pressure — and leave with a practical design record and sharper questions for your team.
15 minutes. No account needed. Technical depth when you want it. Your progress is saved in this browser.
Make real choices
Data handling, evidence, model behaviour, automation, human judgement and operations.
See the consequences
Every capability you gain creates a risk, a cost or an obligation somebody has to carry.
Take away the design
Download the architecture, the open obligations and the questions for your team or vendor.
The first scenario
An AI-assisted technical document reviewer, tested against terms of reference, architecture principles, a recognised standard and previous work. The scenario, the client and the findings are synthetic.
Technical document review
AvailableAn AI-assisted reviewer tests a technical strategy against terms of reference, architecture principles, a recognised standard and previous work.
Knowledge assistant
PlannedAnswers questions from internal documents with contested and overlapping sources.
Case triage
PlannedPrioritises incoming cases for frontline teams under time pressure.
Decision support
PlannedSupports allocation decisions where the cost of a wrong recommendation is high.
Research synthesis
PlannedSynthesises mixed-quality evidence into a defensible position.
Monitoring and alerts
PlannedGenerates alerts and recommended actions from continuous signals.