Software Engineering with AI
This is the With-AI tree — a parallel lens onto the same six-phase lifecycle the Process tree maps. Where the Process tree describes the activity, this tree describes how AI changes the activity: what shifts when AI is in the loop, the tool-agnostic workflow for each phase, the specific tools the author has validated in real client delivery, and the anti-patterns the agency should resist.
Read the Process tree first if the activity itself is new to you. AI is a layer on top of practice — it accelerates work you already know how to do, and it punishes work you do not. The trees are deliberately separate so you can pick the lens you need without one drowning the other.
The battle-tested rule
Section titled “The battle-tested rule”Every tool listed on every page in this tree has been validated by the author in real client delivery. If a tool has not yet been validated, it does not appear — the page ships with the tool-agnostic workflow plus an explicit tool research in progress note. There are no speculative listings, no “considered” tools, no comparison matrices that age into uselessness within a quarter. The site’s job in a hype-saturated space is to be the practitioner filter.
A tool may stay on a page only as long as the validation is recent enough to mean something. Tool entries older than nine months without a refresh trigger a content review; entries that fail revalidation come down with a note explaining why.
How each page is structured
Section titled “How each page is structured”Every working page in this tree follows the same five-section template — the same scan-and-leave discipline the Process tree uses, recalibrated for AI:
- What changes when AI is in the loop — what shifts in the activity, and just as importantly, what stays human-led.
- Tool-agnostic workflow — the end-to-end workflow described independently of any specific tool. You could execute this against any sufficiently capable AI assistant.
- Battle-tested tools and how to use them — specific tools the author has validated, with one-paragraph workflows per tool.
- What is not yet ready — anti-patterns, immature use cases, hype to resist.
- What the industry does — how leading agencies are using AI in this phase, with the trade-off-aware framing the Process tree uses.
The fifth section preserves the Process tree’s signature trade-off framing. The fourth section is the AI-tree counterweight — naming what to resist as deliberately as the Process tree names what to adopt.
- Pre-Sales AI-assisted scoping notes, proposal drafting, SOW research, pricing calibration.
- Discovery Interview synthesis, workshop prep, prototype generation, deliverable drafting.
- Requirements & Design FR/NFR drafting, ADR red-teaming, UX generation, infrastructure planning.
- Delivery — Project Management Status reports, backlog grooming, sprint calibration, retro synthesis.
- Delivery — Development Agentic coding loops, AI-assisted review, dependency automation.
- Delivery — QA / Testing Test-case generation, accessibility scanning, security triage, regression maintenance.
- Deployment / Launch Runbook generation, cutover-rehearsal scripting, monitoring config, smoke tests.
- Maintenance & Retainer Incident triage, PIR drafting, retainer reports, hypercare prioritisation, closeout synthesis.
Cross-tree navigation
Section titled “Cross-tree navigation”Every working page in this tree includes a back-link to its Process-tree counterpart near the top. Cross-phase links between AI pages name AI-to-AI handoffs — Pre-Sales-with-AI outputs feeding Discovery-with-AI inputs, Discovery-with-AI synthesis feeding Requirements-with-AI drafting, and so on. The two trees move at different cadences: the Process tree updates when agency practice evolves (semi-annual); this tree updates when validated tool reality changes (often quarterly).