Scaling AI Enablement: How Odevo Upskilled Engineering

On This Podcast

In this conversation, we examine how Odevo moved from ad hoc AI experimentation to a deliberate, structured training programme after sensing in early 2025 that agentic tools represented a genuine step change rather than incremental improvement. Tomasz describes issuing an RFP for engineer-led training, insisting the sessions be run by practising engineers rather than a generic training company, since developers can quickly spot when a facilitator lacks real credibility. We also discuss why Odevo deliberately paced the programme into modules spaced one to two weeks apart, giving engineers time to digest and apply each concept rather than compressing everything into an intense multi-day bootcamp. A pilot cohort of roughly 20 percent of engineers, run first, surfaced practical problems such as painful dev container setups before the wider rollout began.

In this conversation, we examine the measurable outcomes of Odevo's training, including a tenfold increase in pull requests among previously high-performing engineers and first-PR speed dropping to a day or less across most repositories. Tomasz explains that beyond the raw numbers, curiosity itself increased, with engineers spontaneously exploring topics like Ralph Wiggum loops and competing frontier models rather than waiting to be shown new tools. We also discuss a marked psychological shift toward boldness, with engineers now willing to attempt full application rewrites or prototype ideas in an evening rather than spending months on architecture discovery. Tomasz notes that tedious but important practices like unit testing and documentation have become far easier to embrace once an agent can simply be asked to do them without complaint.

In this conversation, we examine how non-developers across Odevo, from accountants to CFOs, have begun building real applications using Lovable, ranging from personal expense trackers to full career and performance review frameworks. Tomasz explains that rather than imposing heavy red tape, Odevo built lightweight guardrails, including a short training module on data sensitivity and a scanner that audits every vibe-coded repository pushed to GitHub for security risks. We also discuss the philosophy behind this approach, summed up as build it, you own it, which pushes citizen builders to exhaust their own options before escalating to the technology team. Tomasz argues that this openness, rather than gatekeeping, is what is genuinely paying off in engagement and innovation across the organisation.