Private aviation runs on precise language for a reason: “serviceable” and “available” are different words, and confusing them costs more than a wasted trip. Software procurement could borrow that discipline, especially around NoOps — the proposition that a company can contract out and script so much of its operational burden that a separate operations group stops being needed. The proposition has merit at the margins and mythology at the centre, and the myths are durable because each one contains a grain of truth wrapped around an unlimited promise. We have watched versions of all six circulate through procurement meetings at businesses of every size, flight departments included, and each myth costs more the later it is caught. Here are the six we encounter most, stated the way a vendor or an overenthusiastic thread states them, then corrected the way an operations veteran would state them.
Six NoOps Myths, Cleared One by One
Myth 1: “NoOps means nobody touches it — ever”
The term carries two separate meanings that sales material quietly merges. The working version: operations is bought off and scripted down until a separate team is no longer required, with whatever remains absorbed into ordinary roles. The absolute version — fully autonomous IT with zero human involvement — is a different claim, borrowed from the marketing around AIOps, and it has never survived contact with a production system. When someone quotes you the term, ask which version they are selling, and ask what evidence would show the promise has been met. The gap between the two versions is where most of the industry’s arguments live, and where most disappointed purchases are parked.
Myth 2: “Serverless delivers it by default”
Serverless computing is a single road toward lighter operations, not the destination itself. The CNCF has published a whitepaper saying it in as many words: signing up for functions leaves both the machines and the specialists who understand them in place. A widely cited 2019 study from Berkeley granted that the model absorbs nearly all system administration while cautioning that costs behave unpredictably at scale and lock-in deepens. And Amazon’s own Prime Video team provided the footnoted caveat: their monitoring service, built on managed functions, wilted near five percent of its target load and was rewritten on ordinary containers, cutting spend by ninety percent. The tool must suit the task; no pricing model changes that. Neither does a rebrand: the same caution applies to every new managed category that arrives promising the removal of a profession.
Myth 3: “The provider’s SLA is our availability”
A service is only ever as available as the odds that every critical component is up at the same moment. Your booking platform rides on hosting, identity, payments, and delivery networks; it is up only when all of them are up, so the truthful ceiling is the product of their promises — three commitments in the 99.9 range multiply to something beginning 99.8. Quoting the strongest single figure as your own availability is the availability equivalent of quoting one engine’s thrust as the aircraft’s total. Whatever performance you need beyond the multiplied figure has to be designed in, by someone whose name is on the design. Buffering, caching, and a defined degraded mode are the usual instruments, and each one is an engineering artefact with a maintainer — which is to say, operations by another name.
Myth 4: “It’s managed, so it’s backed up”
Managed means the vendor tends the machinery — patching, replication, power. It does not mean your records survive a mistaken deletion, a malicious insider, or a compromised login; the major cloud providers say so explicitly in their shared-responsibility documents. The Snowflake breaches illustrated the sentence with painful clarity: attackers never pierced the platform. They signed into customer accounts that lacked multi-factor authentication, password rotation, and network allow-lists, and helped themselves to whatever those accounts could reach. The lesson is uncomfortable precisely because it is ordinary: none of the missing controls is exotic, and all three are cheap on the day they are installed and ruinous on the day they are missed. Backup strategy, credential hygiene, and access review remain resident on the customer’s side of every contract — the quietest line item and the most fateful. Sign nothing until you can point to the person who owns those three, because a contract that leaves them unassigned has simply re-packaged the old risk with a new invoice.
Myth 5: “We can always migrate later”
This one dies only when priced. Leaving a provider has a real cost in engineering weeks, data verification, and re-testing, and teams that never write the number down discover it at the worst moment. History offers a named example: Google shut its Cloud IoT Core product down during 2023, leaving every subscriber to run a migration on Google’s calendar instead of their own. Nobody involved was careless; the product line simply stopped making commercial sense to its owner, which is a reason no customer clause can veto. An unpriced exit is not flexibility; it is the bill arriving later with interest. The estimate changes behaviour twice — once during negotiation, when it favours suppliers with tidy export routes, and again when renewal comes round, when it hands your negotiator a number instead of a hope. From January 2027, European rules remove switching and egress charges — reducing the transfer fee while leaving the engineering effort, and the discipline of estimating it, entirely intact.
Myth 6: “The AI agent is our operations department now”
The 2026 releases — AWS’s DevOps Agent, Azure’s SRE Agent — genuinely shorten diagnosis, and nobody should dismiss them. But an agent with authority to change production is a party with write access, and Microsoft’s own deployment guidance keeps a human reviewer in the loop for production incidents. Someone owns each agent’s permissions; someone owns its mistakes; someone notices when it confidently resolves the wrong thing. “The agent handles ops” is therefore not an answer to the staffing question. It is a new, unfamiliar entry on the staffing chart that most organisations have not yet filled in — and the honest owners of that entry will spend less time praising the agent’s speed than reviewing what it changed while nobody watched.
The NoOps Position That Survives the Corrections
Strip the six claims away and a defensible position remains, smaller and more useful than the banner version. Lean platforms with bursty traffic really can run without an operations department, provided four things stay true:
- every surviving judgement — retention, access, configuration, permissions — carries a named owner;
- restores are rehearsed on a published schedule, with the date of the last drill written down;
- every critical exit has been priced in engineer-weeks as well as money;
- automation of every kind, agents included, reports to a person who can reverse it.
Provided every remaining judgement has a named owner, restores are rehearsed on a schedule, exits are priced before they are needed, and automation — including the AI kind — reports to a person. Practitioners grade that end state as the top of a maturity ladder, with the bottom rung reserved for companies that simply announced the end state and hoped. The distance between those two rungs is exactly the distance between the myths above and the corrections beside them. None of the corrections is controversial among people who operate systems for a living; the myths persist mainly because each one compresses a real convenience into an unlimited promise, and unlimited promises sell faster than audited ones.
For readers steering their own platforms, this NoOps backgrounder gathers the vocabulary, the measurements, and the cautionary episodes into one place, and is more sober than most of the primary sources it cites. The pricing discipline transfers directly to flying decisions as well: understanding what drives private jet charter pricing is the same exercise of separating the quoted number from the real one, and knowing how empty leg jet flights operate rewards the traveller who reads the terms before the deal — because, as with platforms, the cheapest itinerary and the safest one are correlated only when someone checked. Our broader guide to empty leg travel makes the same closing argument this piece does: the fine print is the product, the promises are the packaging, and the teams that read both arrive unhurried. Six myths cleared from the flight plan leave room for what actually works — measurement, ownership, and a rehearsed way back.
