AI across the software lifecycle
Where AI helps in discovery, requirements, design, coding, testing, release, operations and support, what to keep human, how teams adopt it without chaos, and how to measure whether it is really helping.
🎯 By the end of this lesson you will be able to:
- Identify useful, low-risk AI uses in each phase of software delivery
- Decide which work must stay human-led and why
- Plan a team rollout with clear guidelines and honest measures of impact
In this lesson
AI is now part of every phase
So far you have learned how AI works and the building blocks behind AI products. This lesson turns to your daily work: how AI can help a whole team deliver software, from the first idea to supporting customers. The biggest gains rarely come from one clever tool. They come from many small improvements across the lifecycle, used with good judgement.
Useful AI by phase
| Phase | Where AI helps | Keep human |
|---|---|---|
| Discovery | Summarise interviews and feedback, group themes, draft personas | Deciding what problem matters and for whom |
| Requirements | Draft user stories and acceptance criteria, find gaps and ambiguities | Agreeing scope and priorities with stakeholders |
| Design | Compare options, draft diagrams and decision records, list risks | Architecture decisions and trade-offs |
| Coding | Suggest code, explain unfamiliar code, write unit tests, refactor | Code review, security and final responsibility |
| Testing | Generate test ideas and data, draft automation, analyse failures | Test strategy, exploratory testing, release judgement |
| Release and operations | Draft release notes, summarise incidents and logs, suggest runbook steps | Go/no-go decisions and incident command |
| Support | Draft replies, route tickets, find duplicates, update help articles | Sensitive cases, refunds, complaints and empathy |
| Management | Draft status reports, meeting notes, risk lists and plans | Commitments, people decisions and accountability |
One user story, many helpers
Story: "As a customer, I want to save my card so checkout is faster."
BA: AI drafts acceptance criteria and flags open questions (expired cards? multiple cards?). Dev: AI suggests code for tokenised card storage; the developer reviews it against security rules. QA: AI proposes test cases, including negative and security cases; the tester adds exploratory charters. PM: AI drafts the release note and stakeholder update from the closed tickets.
A word on coding assistants
Coding assistants are often the first AI tool a team adopts. They can speed up routine work a lot, but they also produce code that looks right and isn't: calls to functions that don't exist, outdated libraries, missing edge cases or security weaknesses. The rule that works: AI-written code gets the same review, tests and security checks as any other code, and the person who merges it owns it.
Rolling AI out across a team
- Agree the guardrails: approved tools, what data may be used (Lesson 6), and review rules for AI output.
- Start with a few high-value, low-risk uses: for example meeting notes, test idea generation and code explanation.
- Share what works: a shared prompt library and short demos in team meetings.
- Measure honestly against a baseline (see below), and keep what helps.
- Revisit regularly: tools change fast, so review practices every few months.
Measuring whether AI is helping
| Look at | Examples |
|---|---|
| Flow | Cycle time, time from story to release, time to first review |
| Quality | Escaped defects, rework, incidents, review comments per change |
| People | Developer and tester satisfaction, time spent on repetitive tasks |
| Customers | Support resolution time, satisfaction, repeat contacts |
What this means for your role
Developer
Use AI for boilerplate, tests, explanations and refactoring. Keep design thinking, security review and ownership firmly yours, and keep commits small so reviews stay meaningful.
Tester / QA
Use AI to widen coverage: more test ideas, data variations and faster failure analysis. Spend the time saved on exploratory testing, where human curiosity finds what scripts miss.
Business Analyst / PO
Let AI do the first draft of stories and summaries, then spend your time on the conversations that clarify value and resolve ambiguity.
Tech Lead / Architect
Set the team's standards for AI-assisted code and design, and watch review load and quality metrics as adoption grows.
Project / Delivery Manager
Use AI to cut reporting overhead, and use the baseline-and-compare approach to show leadership the real impact, not just anecdotes.
Scrum Master / Agile
Facilitate the rollout: guardrails, experiments, sharing sessions and retrospectives focused on how AI is changing the team's way of working.
Try it: map AI onto your team's lifecycle
- List your team's main activities from idea to support.
- For each, write one AI assist you could try this month and the human decision that stays with a person.
- Pick the two with the best value and lowest risk, and decide how you would measure whether they helped.
Key takeaways
- AI can assist in every phase of software delivery, from discovery to support.
- People stay responsible for decisions, reviews, security, commitments and sensitive cases.
- AI-written code needs the same review and testing as any other code, and the merger owns it.
- Roll out with guardrails and shared practices, and measure flow, quality, people and customer outcomes against a baseline.
Key terms
- SDLC
- Software development lifecycle: the phases from idea to running and supporting software.
- Coding assistant
- An AI tool that suggests, explains or edits code inside a developer's tools.
- Baseline
- Measurements taken before a change, used to judge its real effect.
Check your understanding
Answer all 5 questions. Score 3 or more to complete the lesson.
1. Which task should stay human-led even with AI help?
2. How should AI-written code be treated?
3. Your team says AI made them faster. What is the best evidence?
4. What is a sensible first step when rolling out AI to a team?
5. What hidden risk can appear when AI makes producing code and documents easier?