Skip to content

Possibilities,
expanded.

Explore further. Discover faster.

Every build starts from a written spec, with AI working against it.

Explore the engineering cycle

The engineering cycle

From a question to working software.

Each stage works from the spec. Engineers set the intent and constraints, AI explores, drafts and tests against them, and the work is reviewed before the next stage begins.

  1. 01

    Understand the right problem.

    Start with the users, the business and the outcome worth building for.

    A shared problem. A clear product direction.

    Human engineers

    Listen to stakeholders, understand constraints and decide what success looks like.

    AI can assist with

    Explore research, organise findings and surface questions for the team to investigate.

  2. 02

    Write the spec and prototype.

    Bring options into the open and write down what the software must do before committing to a direction.

    A written spec and a prototype the team can see, test and discuss.

    Human engineers

    Shape the experience, choose the architecture, weigh the tradeoffs and write the intent and constraints as a spec.

    AI can assist with

    Help explore alternatives, create early prototypes, examine edge cases and flag gaps in the spec.

  3. 03

    Implement in reviewable increments.

    Give engineers more room for the work that needs their judgement.

    Working software, delivered in reviewable increments.

    Human engineers

    Break down the work, guide implementation and review code against the spec and in its product context.

    AI can assist with

    Draft implementation, repetitive changes and documentation against the spec, and explore the code.

  4. 04

    Challenge what we build.

    Check behaviour, quality and the assumptions behind the implementation.

    Evidence that the product does what it should.

    Human engineers

    Define acceptance criteria, assess risks and verify the experience with real use cases.

    AI can assist with

    Generate test scenarios from the spec, support automation and help investigate failures.

  5. 05

    Release and learn.

    Turn delivery into a cycle of useful feedback and considered improvements.

    A product that keeps getting better.

    Human engineers

    Interpret feedback, choose what to improve next and update the spec as the product changes.

    AI can assist with

    Help analyse logs, organise feedback and keep technical knowledge easier to find.

Complexity,
engineered.

Clear boundaries. Deliberate connections. Software that works as a whole.

InterfacesWeb appMobile appServicesDataRecordsAnalyticsIdentityCatalogOrdersBillingAPI contractsData contractsDefined boundariesExplicit contractsDeliberate connections
InterfacesWeb appMobile appServicesDataRecordsAnalyticsIdentityCatalogOrdersBillingAPI contractsData contracts

Confidence,
built in.

Tests written from the spec. Thoughtful review.
Evidence before every release.

Code reviewMeaningful testsHuman judgementRelease decision
Code reviewMeaningful testsHuman judgementRelease decision

How we keep the work reliable

  • The spec sets the direction.

    Engineers write the intent, constraints and acceptance criteria. AI works against them, so its output can be checked against something written down.

  • Context before generation.

    Product goals, existing systems and team conventions guide the work. We give AI a defined task and check its output against that context.

  • Quality has to be demonstrated.

    Working code is a starting point. Tests, reviews and product feedback help us establish whether a change is useful, reliable and maintainable.

  • Use AI where it helps.

    We choose tools around the task and the information involved. The measure is a better development process and a better product.

Engineering never
stands still.

Explore our engineering capabilities and the thinking behind them.

Explore our insights