Choose timezone
Your profile timezone:
Artificial Intelligence and Machine Learning Seminars
Effective software development methodologies into the fast-evolving AI tooling era
by
→
Europe/London
Maths: MB-203 (QMUL)
Maths: MB-203
QMUL
Description
Join Zoom Meeting
https://cern.zoom.us/j/64400513926?pwd=Wvz54azZ8sFMEYrD3VxCm70uaLaaBc.1
Meeting ID: 644 0051 3926
Passcode: 42
-------------------------------------------
AI tools are changing software development faster than any previous “revolutionary” shift, but the biggest risk isn’t choosing the wrong tool. It’s adopting them without the engineering principles that keep systems reliable, teams aligned, and businesses safe.
In this 1-hour session, we’ll look at the most common mistakes teams made with past disruptive technologies, and how the same patterns are repeating with AI. We’ll revisit core methodology principles (risk management, predictability, maintainability, alignment) and show where AI deliberately breaks them, and when that’s actually useful. Then we’ll get practical: how to adopt AI tools with clear guardrails around security, privacy, IP, auditability, and least privilege; how providers may use your data; and what capabilities are still too immature to bet production engineering on.
We’ll also cover the human side: how AI changes team communication, documentation, and decision-making, where hallucinations happen, how misinformation spreads across roles, and how to mitigate it without slowing down.
We’ll end with a short workshop using a real challenge from the audience —focusing not on “shipping faster,” but on shipping better: clearer requirements, fewer regressions, stronger tests, and safer rollouts.
Lecture overview:
- Biggest mistake done so far with other revolutionary tools:
- how to avoid them now with the AI tools
- Back to the methodology principles:
- Why do exist (risk management, alignment, predictability, maintainability, SAFe)
- Why we still need them in an AI-accelerated workflow
- Where AI doesn’t follow them - and why that can be useful (exploration, divergence, speed)
- How to increase methodology effectiveness with AI (better discovery, faster feedback loops, stronger reviews)
- AI tools increase productivity as well as they increase business risks
- Core principles for adopting/configuring AI tooling (security, privacy, IP, auditability, least privilege)
- How provider use your data at the models advantage
- What we must still wait before to start to use:
- Openclaw case
- Team communication with AI tools:
- What can be improved (shared context, faster onboarding, decision logs, better docs)
- NotebookLM and other similar concepts
- What are the issues? How different role in a company reach to AI generated information for decision making: things go really wrong, really fast
- How to mitigate them? /Fail fast principle/ software dev methodology and simply... communicate more
- What can be improved (shared context, faster onboarding, decision logs, better docs)
- Workshop on one of the audience's problems and/or challenges and let's see how we can use AI tools to…
- Not “solving it faster”…
- Solving it better: cleaner requirements, fewer regressions, better design trade-offs, stronger tests, safer rollout.
What this presentation is not:
- A current snapshot of tools and their strengths.These change too fast. Instead, we’ll focus on principles that help you adapt quickly and guide how tools evolve and how you use them effectively.
- Agent design.Agents are great for clerical automation, but today they’re often not effective or reliable enough for real-world software development workflows.
- Vibe coding: it can be a powerful way to explore ideas quickly, but paradoxically it’s not what you use to build and maintain production-grade products.