If I have to touch infrastructure, I double my estimate.
In early 2025, that was the prevailing sentiment among developers at PayFit. It wasn't that we didn't have tools, we had plenty. We also had a platform team. Actually, we had five of them.
And that was exactly the problem.
The Organizational Pendulum
Between 2019 and 2022, PayFit swung between two extremes:
During our hyper growth phases, we’d have a centralized platform team building abstractions in a vacuum. By the time COVID hit, the pendulum swung the other way, we’d embed SREs directly into product teams to "bring DevOps culture", only to find they were being pulled into daily firefighting instead of building leverage.
We oscillated between monorepos and multi repos. Every swing left behind "ghost" migrations, half-finished initiatives from 2021 or 2022 that developers were still forced to maintain in 2024.
The "Scorecard" Era
Trust reached an all time low during what I call the "Scorecard Era" around 2023. We had SREs embedded in teams, but their objectives were often conflicting. While the product team was trying to ship a critical feature, the SRE was focused on checking boxes on a platform scorecard, metrics that mattered to the platform team but felt like a pain to the developers with no immediate business impact.
We had four or five different platform related teams, and if you asked them how to deploy a new service, you might get four or five different answers. Gregor Hohpe (who explores the "Platform as Product" mindset in his book Platform Strategy) has a great analogy for this: fruit basket vs. fruit salad. A "fruit basket" is a collection of whole fruits, powerful tools, but you have to peel, chop, and mix them yourself. A "fruit salad" is the integrated experience where the platform team has done the heavy lifting for you.
We weren't providing a fruit salad. We were throwing a heavy fruit basket at developers, and it was full of pips.
One Voice, One Team
The biggest change wasn't technical. It was organizational. By late 2024, we had finally stopped the fragmentation. We consolidated those 4-5 different teams into a single Internal Developer Platform (IDP) team.
This "one voice" speaking with a single roadmap and source of truth was the prerequisite for everything that followed. It gave us the opportunity to stop the "pendulum swings" and adopt a true "Platform as Product" mindset.
We also had a rare alignment in leadership: a shared conviction that a strong platform is a competitive advantage, and a collective willingness to prioritize long term stability over immediate feature velocity.
The Frontend Precedent
Trust isn't built with slides, it’s built with proof. Our frontend teams had already moved to an Nx monorepo and were seeing massive benefits in caching and developer ergonomics (a journey featured on the Nx blog and Nicolas Beaussart just gave a great talk over at the React 2026 conf in Paris).
They definitely opened the door and led the way for the rest of us. By the time we proposed the same model for backend and infrastructure, we weren't asking for a "leap of faith", we were simply following a path they had already cleared.
The Change Policy: Paved Road over Mandate
The "one voice" consolidation wasn't just about efficiency, it was about strategy. It allowed us to shift from a mandate driven culture to a "Paved Road" model.
We stopped trying to force teams to migrate. Instead, we focused on making the new Internal Developer Platform so much faster, more stable, and more intuitive that it became the path of least resistance. When the Paved Road is 10x better than the legacy "Dirt Road", you don't need a mandate, teams will naturally choose it. This reduced organizational friction and let us focus on building a product that developers actually wanted to use.
The Turning Point: The Interview Tour
With a unified team finally in place, we realized we couldn't just "tool" our way out of the remaining trust deficit with the teams. We stopped building and started listening.
Throughout H1 2025, we conducted an "Interview Tour". We identified key players across every domain & team and conducted 45 minutes deep dives. We didn't ask "what tools do you want?" We used a set of prepared, open ended questions to drive the discussion and collect essential qualitative feedback on "where it hurts."
The feedback was raw:
- "If I have to touch infrastructure, I double my estimate."
- "Migrations feel like a trap because they never end."
- "I don't know who to talk to when my pipeline breaks."
The Result
- Product first mindset: We treated our developers as customers, not captive users. We rebuilt the platform foundations in public, starting with CI/CD, and made the work visible as it happened instead of revealing it only once it was finished.
- Working with teams, not for teams: We kept talking to users continuously, shared wins publicly, gave kudos for every contribution, and backed the story with concrete data. The goal was not to build a platform for developers in isolation, but to build it with them.
- A better developer experience: We focused relentlessly on the basics developers actually feel every day, faster workflows, more reliable pipelines, and less platform friction. That also meant cleaning up our own setup first and dogfooding the same repository patterns, tooling, and workflows we wanted other teams to adopt. In other words, we cleaned our house before inviting guests.
- Starting with the right early adopters: We deliberately began with a handful of teams we already had strong relationships with. We gave them our full attention, treated their issues with the new system as our top priority, and fixed problems fast. That was the least we could do: they were trusting us with a brand new platform, and their feedback directly helped us improve it.
- The turning point: At first, we were the ones asking teams whether they wanted to migrate. Then, gradually, the dynamic flipped. More and more teams wanted in, to the point where demand started to outpace our capacity to support migrations. That was a great problem to have, and the clearest signal that trust was coming back.
By shifting from "Platform doing things TO us" to "Platform working WITH us", the narrative started to change. We didn't create momentum with mandates. We created it by listening, improving the house, and proving that the paved road genuinely delivered better DX, faster workflows, and more reliable outcomes.
But to deliver on the promise of a "simpler & faster workflow", we had to tackle one of the monsters under the bed: CI/CD complexity.
In Part 2, I’ll dive into how we rebuilt our CI pipeline with Nx to drive standardization, performance, and stability across the entire organization.
Top comments (0)