AI-Native Engineering Programme
Led organisation-wide shift to Generative AI delivery. Role-specific upskilling paths, in-house AI coding assistant for legacy migration (AngularJS → React), governed via internal ADRs and responsible-AI playbooks.
The Challenge
As generative AI tools emerged in 2022–23, the challenge was not adopting them but governing them. Ad-hoc tool use without standards creates security risks, inconsistent quality, and a two-speed engineering culture where some engineers are 10× faster and others are left behind.
The goal was to make AI-native engineering a team-wide standard — not a superpower for a few individuals.
The Programme
- Role-specific upskilling paths: individual contributor, tech lead, and engineering manager tracks
- Internal ADRs governing approved tools, security boundaries, and usage patterns
- Responsible-AI playbooks covering IP risk, hallucination handling, and code review standards for AI-generated output
- In-house AI coding assistant for legacy migration: AngularJS → React, REST → GraphQL
- Monthly engineering forum to share patterns, failures, and experiments
The In-House Migration Tool
The most technically distinctive output was a RAG-powered coding assistant trained on the legacy codebase. Engineers could query it for context on legacy patterns, ask it to generate equivalent React/GraphQL code, and use it to accelerate a multi-year migration programme.
Built on Azure AI services with a retrieval layer over the existing codebase — not a generic LLM but a contextually aware assistant with project-specific knowledge.
Outcomes
- ~40% developer productivity uplift measured across pull request velocity and cycle time (DORA)
- 3 engineers promoted to Lead within 18 months — with AI engineering as a core skill in their promotion cases
- In-house migration assistant deployed and actively used by the team
- Zero security incidents from AI tool usage — attributable to the governance framework
- Programme adopted as a reference model across Publicis Sapient Nordics