A couple weeks ago, my colleague Andy V. and I delivered a webinar on some keys to adopting AI in software delivery: - Shared context that's updated with every release - Structured governance embedded in the context and actualized through skills and agents - Human review at key milestones throughout the lifecycle In the presentation I also demo an agentic process to refine development stories. I used GitHub Actions to trigger Claude Code, ambient context within the repository to inform requirements and acceptance criteria, and Slack to let Claude communicate with a PM proactively to resolve questions and approve stories for development. Check it out! https://lnkd.in/eHJq6W-k

We have officially launched The AI Enabled SDLC: Future Visions Opportunity Report 2026. 💭 For the last 18 months, we’ve been running the numbers on the AI-enabled Software Development Lifecycle (SDLC). We’ve tested AI augmentation across five distinct contexts - from isolated tasks to legacy enterprise systems. We didn’t write this whitepaper because we have all the answers. We wrote it because too many of the efficiency figures that we are hearing aren't unpinned by accurate measurement and more importantly, business context. If you want an honest conversation with industry leaders about where AI augmentation can really have a positive impact in your organisation, and where it probably won't, take a look at our findings below.

Great to see this go live! 18 months of data gathering from delivery teams, understanding the impact of AI in the Software Development Lifecycle. 💡 My key takeaway: Not all AI interventions are created equal, and the impact is determined more by your organisational context than by tool selection or technical capability.

The new Waydev is live. For the first time, engineering leaders can measure the full AI SDLC. See which AI tools your teams use and what you spend per vendor, per team, per repo. Follow AI-generated code from IDE to production and see where it ships and where it dies. Know your cost per PR, tokens consumed, and which agent wrote which line. Waydev Agent closes the loop by feeding insights back to your AI through MCP. One platform. From token to production. We're live on Product Hunt today. Link in comments.

ADLC is the new SDLC. But without runtime context, faster development just means faster failures. Here's what AI-driven development has changed: → Code is generated at scale → Deployment is continuous → Distributed systems are more complex The ground truth can’t be found in staging or QA anymore. It lives in production. The problem? Most AI tools still can't see what's actually happening at runtime. Lightrun gives AI agents and engineers live visibility to execution context, adding dynamic metrics, traces, and snapshots directly to running services. No code changes. No redeployments. No more guesswork. AI code generation is fast. Now let’s make it reliable.

Watch how Propel SDLC brings governance in this AI-first software development era. Today, AI tools are widely used in software development, but mostly without standards, traceability, or accountability. Hallucinations go unchecked. AI-generated outputs ship without oversight. Propel SDLC is KANINI’s AI-powered SDLC orchestration framework, which brings structure and governance to AI-assisted software development. Why Propel SDLC? ✅40% faster delivery ✅65% first-draft accuracy ✅Full SDLC phase coverage ✅Tool-agnostic & vendor-neutral ✅Greenfield and brownfield-ready Connect with us to learn more about Propel SDLC #PropelSDLC #AIGovernance #SoftwareDevelopment #SDLC #EnterpriseAI

We evaluated 50+ AI tools for enterprise grade application development. None of them worked for our teams. So we built our own. Introducing Cerebro, an AI Assisted SDLC.

A client asked me to evaluate BMAD-METHOD (an AI agents framework that structures your entire development process) for their SaaS MVP. Here is my verdict. And I think most people writing about it are missing that point entirely. BMAD gives your project a full agent team: Analyst, Product Manager, Architect, Developer, DevOps, QA, Scrum Master. Each one is a Markdown file with a role, a persona, and a handoff to the next. The workflow is built on Agile disciplines, human-shaped by design. The workflow mirrors the way a real team thinks and passes work forward. That part genuinely impressed me. But here is what I learned. Before a single line of code is written, you are engaging agents one by one, validating the PRD section by section, defining architecture in full detail. The upfront investment is significant. If you miss a constraint early, backtracking is painful. Token consumption is really high. And it is not autonomous - you are still in the driver's seat at every step. For an MVP, that is a problem. An MVP exists to test assumptions quickly. Requirements shift constantly. What you know on day one is not what you know on day ten. BMAD’s strength - its insistence on getting the plan right before building - becomes a liability when the plan itself is the thing you are still figuring out. This is where I think the conversation needs to mature. BMAD is a tool for a specific level of business maturity and readiness. You need a validated idea, stable requirements, and a team that can commit to the planning phase before it pays back. Used at the wrong stage, it slows you down and creates false confidence in a plan that will change anyway. Used at the right stage - a scoped, well-understood build with real complexity - it shines. The timing has to match the business. If you want to try it yourself, I’ve put together some free setup tutorials and resources worth checking out in the comments. https://lnkd.in/eitZqnFv

Most enterprises are adopting AI in the SDLC backwards. They start with the glamorous layer — "let's put Copilot on  every developer's IDE" — and then wonder why six months later  the pull request quality hasn't moved, security issues are up,  and the architecture is drifting faster than before. The order that actually works, from what I've seen driving  AI initiatives across large customers: 1. Governance layer first. Secret scanning, SAST, license  policy, prompt logging. Before any model touches production code. 2. Pipeline intelligence second. AI in the CI/CD loop — code  review, architecture drift detection, release readiness — where  findings are reviewed, not auto-merged. 3. Agentic assistance third. CrewAI-style multi-agent workflows  for well-scoped tasks (test generation, migration scripts,  documentation). 4. IDE assistance fourth. Only after the first three exist,  so the IDE suggestions land in a system that can catch the  mistakes AI still makes. The backwards version — IDE first, governance last — is how  teams end up with a productivity bump they can't measure and  a security posture they can't defend. I've been building a working reference for the first three layers  as an open-source PoC: Jenkins + CrewAI agents + secret scanning  + SAST + Grafana observability, all running in Docker.  https://lnkd.in/gR9mSibD For those running AI-in-SDLC initiatives — where did you actually  start, and would you do it in the same order again? #EnterpriseAI #SDLC #AIGovernance #SolutionArchitecture #OpenSource

AI journey #handsOn #2 ---------------------->>>>> I think I’ve reached the point where I understand this space well enough to start sharing a stronger perspective on what an SDLC Assistant could become. And honestly, that’s exciting. This felt like the right moment to share my current view of what an AI-supported SDLC (software development life cycle) can look like in practice, not just as a concept, but as something real teams could actually use. What’s next for the project: CORE: GitHub Projects workflow support, including review teams per implementation EXAMPLE: a dedicated tech stack branch to show how setup can be optimized for a specific project DEMO: a tutorial and walkthrough are probably next, because this needs to be seen in action Repo: https://lnkd.in/diA4cF8U I’d really like to hear how others see this evolving: If AI becomes a true delivery layer inside the SDLC, what should the core include first to make it genuinely useful? #AI #SDLC #SoftwareDevelopment #ProductManagement #GitHub #AIAgents #DeveloperTools