I recall the moment it arrived at 11:47pm, the timestamp was right there in the commit log, precise and unapologetic. A complete architectural solution to a problem we had been circling in morning standups for three weeks. No preamble, no draft, no “just thinking out loud.” Ready for production review.
My first reaction, I will admit, was not admiration. It was something closer to mild institutional irritation. I had spent eighteen months building a programme governance model that ran on calendar discipline and visibility. And here was one of our best architects producing the clearest thinking of the entire quarter at a time when I was firmly, unambiguously, asleep.
That moment did not immediately feel like a lesson. It felt like a scheduling problem, which is precisely the kind of wrong answer that looks sensible at scale.
The Situation
I was deep into an enterprise data infrastructure overhaul – four regions, nineteen legacy systems, somewhere north of forty stakeholders depending on the week and the mood of the steering committee. The programme was running. Milestones were being met. By every external measure, I was in good shape.
However, something kept slipping. Not deliverables – those were landing. It was the quality of the decisions inside the deliverables. The morning syncs felt sharp. The design reviews were engaged. And then, consistently, inexplicably, the decisions made in those rooms would begin to unravel within days. Not because the people were wrong. Because the thinking was incomplete.
In late 2023, I pulled the contribution logs. I did this not to catch anyone out – the programme was not in crisis – but because I had a nagging suspicion that what I was measuring (attendance, responsiveness, meeting participation) had quietly drifted away from what I actually needed to track (thinking quality, architectural coherence, decision durability).
What I found was uncomfortable in the specific way that useful findings tend to be. Two of my strongest architects – people whose judgement I trusted more than most governance frameworks I have ever read – were doing their clearest, most structurally complete thinking between nine and midnight. Not occasionally. Consistently. The contribution logs made it embarrassingly obvious in retrospect: the commits that unblocked other people’s work, the design notes that reoriented entire workstreams, the quiet corrections that prevented expensive rework. Almost all of it after hours.
I had built a programme governance model around availability windows. The work was happening around them, not inside them. The thinking was occurring anyway – just invisibly, uncredited, at midnight, outside the architecture of the programme itself.
What I Got Wrong
The first thing I got wrong was conflating presence with production.
I found comfort in a full calendar. It signaled momentum. It signaled alignment. It signaled that I was, at a minimum, in the same room as the people who were supposed to be solving the problem. What it did not signal – and this is the thing governance models consistently fail to encode – is when the actual thinking happens. I had optimised for observable effort and was quietly surprised when that turned out not to be the same thing as best effort.
Chronotype is not a wellness concept. It is a cognitive variable. The research on this is not new – circadian rhythms, ultradian cycles, the neuroscience of alertness and decision-making under fatigue – but in enterprise transformation, it remains almost entirely absent from how I structure work. I talk about agility, about flow states, about psychological safety. And then I schedule the most cognitively demanding decisions at 9am on a Monday following a cross-regional steering committee.
The second thing I got wrong was the word I kept using.
When I began discussing how to respond to what I had found, the language in the room defaulted to flexibility and async policy. Both reasonable words. Both slightly beside the point. The word I kept avoiding – the one that actually described what my architects needed – was permission. Not a formal policy change. Not a new HR framework. The quiet organisational signal that your best thinking does not have to be witnessed to count.
That is a harder thing to give than a flexible working policy, because it requires something institutions are structurally reluctant to offer: trust in the output over trust in the process. It means accepting that I will not always see the work happening. I will see what the work produces.
The third thing I got wrong was treating this as an edge case.
Two architects doing their best work at night sounds like a curiosity. An anecdote. A nice story for a team retrospective. What the contribution logs actually showed was a systemic misalignment between when my programme demanded cognitive energy and when my people had it to give. That is not an individual quirk. That is an organisational design problem.
I restructured two workstreams around output quality, not presence. No midnight standups. No requirement to be online at any particular time. Clear deliverable expectations, clear quality standards, genuine latitude on when the work happened. Six weeks later: the same people, the same problems, measurably sharper decisions. Not marginally sharper. Structurally different.
What This Means for Your Organisation
Most high-performing organisations are already benefiting from their night-thinkers. They simply are not doing it intentionally, which means they are also not doing it efficiently, and they are almost certainly losing some of those people to organisations that make them feel less like they are working around the system. The competitive edge here is not in discovering that chronotype exists – your best people already know this about themselves. It is in building an operating model that makes their best thinking visible and valued, rather than something they contribute quietly, on their own time, hoping someone notices the timestamp.
The organisations that figure this out will not just retain better people. They will extract the thinking those people were already doing anyway – just uncredited, in the margins, at 11:47pm – and bring it properly into the work.
The Close
My programme did not need better people. It needed to stop scheduling around the hours when its best people happened to be least capable of thinking clearly.
I believe in a version of productivity that is about discipline and consistency and showing up. I have run large, complex programmes on the back of it. But there is another version – quieter, less visible, harder to govern – that is about recognising what conditions actually produce my team’s clearest thinking, and then having the institutional confidence to build around that, even when it means trusting work I cannot watch being done.
The 11:47pm insight shipped. It was right. The architecture held.
The governance model that almost prevented it from being taken seriously – that, I had to rebuild.
If you are thinking about the relationship between pace, progress, and the conditions that enable good work, this piece on staying in motion through uncertainty is worth the ten minutes.