Let’s clear something up right away: Kubernetes is not DevOps. Neither is Jenkins, a two-day workshop, or a job title on a business card. If that stings a little, good — it stung most of us, and hospital software projects have been quietly paying for these misunderstandings for years. The good news is that every myth below has been tested against real delivery data, and each one cracks under the slightest pressure.
We work around eye care, where precision matters and so does candour about what a treatment can and cannot fix. It turns out the same honesty helps when the subject is DevOps. So here are the six myths we hear most often from clinical IT groups, along with what the evidence actually says. If you want the full definitions behind any term we use, the Ops Atlas entry on DevOps stands as the friendliest long reference we know.
First, what DevOps actually is
One paragraph of grounding, so the myths have something to be measured against. DevOps organises delivery around three “ways”: keeping work flowing in one direction in small pieces, tightening the feedback loop from production back to the builders, and making continuous learning part of everyday routine. The CALMS checklist fills in the rest — culture, automation, lean, measurement, sharing — and notice that four of those five are human arrangements rather than software. That is why the myths below all fail in the same spot: each one promises a technical shortcut around a human arrangement, and delivery outcomes simply refuse to cooperate with that bargain.
Myth one: DevOps is a product you can purchase
Vendors love this one, for obvious reasons. But DevOps is not a tool, a job title, a siloed team, or a box with a licence key. DevOps arranges software delivery so that builders and operators form one team around shared goals, shared tools and fast feedback — the arrangement that gets changes to users quickly, safely and on a repeatable cycle. You can buy automation; you cannot buy the sharing. Groups that try end up with an expensive pipeline bolted onto the same old queues, and a receipt nobody wants to explain to the finance committee.
Myth two: you must rebuild everything on Kubernetes first
Here is a relief: microservices and Kubernetes are not prerequisites. The research on delivery performance does not condition the good outcomes on any particular architecture. Plenty of high-performing teams run sensible monoliths deployed with discipline. What actually correlates with performance is how you work — small batches, fast feedback, safe reversibility — not what the infrastructure diagram looks like. Starting with a platform rebuild inverts the order of operations and burns two years before a single clinic feels a difference. Fix the working habits first; re-platform when there is a reason, not a fashion. A group that ships weekly from a boring stack will outrun a group that ships quarterly from an impressive one, and it will do so every single quarter.
Myth three: regulated industries simply cannot work this way
This myth has the most sincere advocates, usually people who have sat through an audit. And yet the evidence keeps finding high performers inside banking, aviation and healthcare. The trick is that regulated environments can adopt fully, because the controls can be automated: least-privilege access, OIDC-issued short-lived credentials instead of eternal passwords, expand-and-contract database migrations, and a quarterly audit that samples roughly ten random changes for evidence. That is a stronger control than a signature on a form, because it is exercised continuously rather than annually. Compliance and speed are not enemies; unaudited manual process is the thing both of them are actually fighting.
Myth four: the best teams max out every metric
The picture of an elite team winning on all five DORA measures at once — shortest lead time, highest frequency, fastest recovery, lowest failure rate, lowest rework — is tidy, and wrong. Strong teams show profiles, not perfect scorecards. Some deliberately accept a marginally worse failure rate in return for restoring service within minutes, which is the right trade when patients are booking appointments against your system. Treat the measures as a shape to be steered, and treat rework rate — the addition DORA made in 2024 — as the early warning that speed is being borrowed from quality. A team that reads its profile well knows which trade it is making and can defend it in front of a finance committee; a team chasing a clean sweep usually achieves only a tidy-looking spreadsheet.
Myth five: after an incident, add more approvals
Instinct says tighten the screws. The data says the opposite: it connects heavyweight change approval boards to worse delivery performance, not better. Every added gate pushes teams toward bigger, riskier batches — which cause exactly the incidents the gates were meant to prevent. What works better is smaller changes, peer review built into the pipeline, feature flags that let you switch a misbehaving release off in seconds, and recovery paths that have actually been rehearsed. One published case found a rollback took 190 minutes when improvised and about twenty when practised. Preparation, not paperwork, is what shortens bad nights — a point our earlier piece on why SRE and DevOps matter for enterprises that cannot afford outages makes at greater length.
Myth six: appoint a DevOps team and the problem is solved
The permanent “DevOps team” is the most seductive anti-pattern of the lot, because it looks like decisive action. In practice it recreates the wall the movement spent fifteen years dismantling: one group does the deploying, everyone else throws changes over it. The term describes a practice, not a department. What does work looks less glamorous — a pilot value stream taken from first step to last, with an enabling coach at the team’s shoulder. Adoption studies gave the coached-pilot route 4.25 of 5, with tool-buying programmes at 2.40 and memo-announced mandates at 2.50. Choose your myth-breaker accordingly.
Seeing past the DevOps myths for good
Notice the pattern across all six: each myth offers a shortcut that avoids changing how people work together, and each one fails for the same reason. Choosing an eye surgeon is a useful comparison — we wrote a guide on choosing eye hospitals for safe surgery and better outcomes, and the advice rhymes: judge evidence over advertising, ask what happens when things go wrong, and prefer the team that can explain its process plainly. Software delivery deserves the same scepticism. Ask what the numbers are, who owns a rollback, and how quickly feedback reaches the people who can act on it. The answers will tell you whether you are talking to a genuine capability or a myth in a polo shirt.
Where does a group typically sit when it still believes three of these myths? Early on the maturity ladder — siloed, perhaps drifting into automated islands. The later rungs, through integrated delivery and continuous delivery to a learning system, are reached by working differently, not by announcing intentions. The myths are simply the stories groups tell themselves to avoid that first honest step.
- Pick one myth from this list and name, out loud, where your group still believes it.
- Check your last incident review: did it end with a new gate or with a smaller, safer change?
- Ask for the five delivery measures by name. Confusion there is diagnostic.