AI Engineering
How should your engineering team deliver when AI participates in the work? Establish responsibilities, evidence and release decisions that connect AI-assisted tasks to a reliable development lifecycle.
The work in context
Adding an AI coding tool changes a developer’s task before it changes a team’s delivery system. Requirements still need decisions, designs still need review, and somebody must accept responsibility for a release. The Engineering Operating Model defines how that work fits together. We examine the lifecycle, clarify who decides at each handoff, and establish the evidence needed to move forward. The aim is an engineering capability the client can own, rather than a collection of individual tool habits.
Engineering leaders may see more code produced without better throughput or consistent quality. Teams can differ in allowed AI use, output review and context access. These are operating questions alongside tooling questions. This work establishes agreed ways of working. Software Factory connects the model to shared technical foundations; Governance Structure resolves policy and approval decisions that the engineering process must respect.
A view of the work
Working together
Follow representative work from requirements through design, development, evaluation and release. Identify where decisions, context or feedback are missing. Use the team’s actual repositories, constraints and review practices to define the scope rather than adopting a generic lifecycle.
Define who owns requirements, technical decisions, evaluation evidence and release approval. Describe human accountability when AI prepares work. Specify escalation and exception routes so the operating model remains usable when a task does not follow the ordinary path.
Agree on how context, coding assistance, testing and evaluation participate in the lifecycle. Set review expectations and evidence requirements. Toolchain changes and technical enforcement are named separately when their implementation is part of the engagement.
Walk representative work through the proposed model. Check whether practitioners can use the responsibilities and gates, then record unresolved decisions. Define the transition, documentation and coaching required before the client can operate the agreed process independently.
Outputs and decisions
Illustrative example
Two teams use the same coding assistant but review changes differently. One retains test evidence with each change; another treats the generated code as equivalent to reviewed code. During an operating-model exercise, the team could define the evidence, review and release decisions needed for a bounded workload, then walk through a change and an exception. This example illustrates responsibility design. It does not claim that a client has adopted the model or achieved a particular engineering result.
Illustrative scenario, not a reported client result or a product capability claim.
Scope and responsibility
This page concerns engineering ways of working. Enterprise process and role redesign belongs in Strategy and Transformation; client policy authority belongs in Governance. Defining a model is distinct from integrating tools or rolling it out across teams. Agree on which design, pilot, implementation and transfer activities are contracted. The scope does not imply a completed software factory, application modernization or a guaranteed productivity increase. Those require separate delivery and acceptance criteria.
Practical questions
The work should begin with the process you have. Some practices may remain useful; others may need changes because AI affects context, review or evaluation. The recommendation should explain why each change is needed.
Only when agreed. The operating model defines the required behavior and responsibilities. Implementing integrations or enforcement has additional dependencies and acceptance evidence, which belong explicitly in the scope.
A representative walkthrough can test usability and expose gaps. A pilot can provide stronger delivery evidence when contracted. Longer-term throughput and quality claims require a baseline and subsequent measurement.
Next step
Bring a representative engineering workflow and its release decisions. We’ll identify the operating changes and implementation depth needed.
Discuss this work