Check the DevOps terms before you sign
What stalls this decision is what you are left holding afterwards, not the price. Here are the numbers, the terms, and what you own from day one.
Tell us what you are running
Send us the shape of your setup and what keeps breaking. A named person reads it and replies.
What DevOps work actually costs to run
GitHub Actions, standard hosted 2-core Linux runner past the included allowance, read 2026-08-29.
US annual mean for Network and Computer Systems Administrators, BLS OEWS 2025, SOC 15-1244, median $99,130.
The compute is not where the money goes. BLS publishes no DevOps occupation, so Software Developers, SOC 15-1252, at a $148,100 mean is the upper bound on the same work. Neither figure is a comparison with the other, and neither tells you where we sit.
Who answers when it breaks, and how fast
| Provider | What the published policy commits to |
|---|---|
| What the published policy commits toThe Technical Support Services Guidelines, clause 3.5, Request Acknowledgement. Google may respond to a Request by acknowledging receipt of the Request. Customer acknowledges and understands that Google may be unable to provide answers to, or resolve all Requests. | |
| Microsoft | What the published policy commits toInitial Response Time is the period from when you submit your support request to when a Microsoft Support Engineer contacts you and starts working on your support request. The clock stops at engagement, not at resolution. |
| Cloudflare | What the published policy commits toSupport response times are published for Enterprise at P1 two hours and P2 four hours. Premium Enterprise is P1 one hour and P2 two hours. No support response time is published for Business. For pay-as-you-go and Free the page states that no SLAs are offered. Read 2026-08-30. |
| Heroku | What the published policy commits toNo numeric response-time commitment at any tier on the support channels page. Heroku states that Standard tickets get a reply within a few business days. Premium offers 24x7 direct access with no response figure attached. Read 2026-08-30. |
| Atyantik | What the published policy commits toBusiness hours are Monday to Friday, 10:00 to 19:00 IST. Inside those hours the response targets are P1 four working hours, P2 twenty-four working hours, and seventy-two working hours for everything else. Like Microsoft above, that clock measures acknowledgement by a named person, not resolution. No out-of-hours paging. After launch the arrangement is an Annual Maintenance Contract or a lean mode, both through the same documented change-note process. |
Why most teams stay where they are
Show data table
| Stage | Share of surveyed professionals |
|---|---|
| 2018 | 79% |
| 2019 | 79% |
| 2020 | 79% |
What we looked for, and what we could not find
Before writing any of this we went looking for numbers. We wanted the ones that would let you buy this on evidence rather than on trust. How often does this work get abandoned? What actually separates the teams that get it right? And what do the numbers say a good outcome looks like? Here is what came back, including the parts that came back empty. A provider who only shows you the searches that worked is showing you a sales deck.
What we build, and what you hold
01
Documentation
Documentation starts at the beginning of the project, not at the end of it. It covers the deployment and the CI/CD pipelines as they are built.
You get
Written deployment and pipeline documentation you hold from the first weeks, rather than a handover pack assembled on the way out.
02
Code and review
GitFlow branching, unit and integration tests, CI/CD gating every merge, and a senior software engineer reviewing the code before it lands.
You get
A merge history where every change passed the same gates, plus Technical Requirements Specification records of the architecture decisions and why each was taken.
03
Deployment
We deploy across AWS, Google Cloud, DigitalOcean, Cloudflare, Vercel and Netlify, so the target is your choice rather than ours.
You get
Working pipelines on the cloud accounts you chose, with the deployment steps written down next to them in plain language.
04
Running it
We build pipelines that target 98-99% uptime under normal operating conditions, and cloud-provider outages sit outside our scope. Everything within that scope is our responsibility.
You get
The uptime target, the scope boundary and what sits inside it, all three stated together in the engagement terms you sign.
05
The team
DevOps and QA are part of the delivery model. For full-team buildouts, developers are paired with at least one DevOps and one QA engineer.
You get
Named people on your engagement, with the DevOps and QA pairing written into the scope of a full-team buildout.
What happens if you leave
You cannot get your system back
Documentation is written from day one and the handoff is a process, not a hostage situation. You have that freedom from day one.
What we commit to
You can end the engagement at any point and leave with the documented system, the pipelines and the accounts. So far, no client has needed to use that exit, and we would rather say that plainly than let you assume it has been road-tested.
Nobody wrote anything down
NIST SP 800-53 Revision 5 control SA-5 asks for system documentation, and CM-2 for a baseline configuration. SP 800-218 SSDF PS.3 asks the same of software.
What we commit to
The SA-5 discussion names lack of support from developers and contractors as a reason documentation goes missing, and puts the cost of recreating it on the organisation that inherits the system. We do not certify against these frameworks; we write the documentation.
You will not be here in three years
Our longest active partnership has run more than a decade, and we have delivered 50+ enterprise engagements across 7 countries.
What we commit to
The proof that matters is not a logo wall. It is that the longest relationship has run more than a decade, and that when an engagement ends, what runs is yours, in your accounts, documented well enough for the next team to pick up.
Who this is not for
Before you spend another meeting on this, one thing is worth settling first. Do you need a contractual response time at two in the morning, or the system documented and yours?
Buy a managed provider
Choose this when
If you need someone paged out of hours, buy a managed provider that publishes 24/7 response, because we publish targets for business hours, Monday to Friday, only.
It costs you
Cloudflare publishes two hours for a P1 at Enterprise and one hour at Premium Enterprise. That is a real instrument, and we cannot match it by saying we will try.
Work with us
Choose this when
If what you need is the pipelines built, documented and running in your own accounts, with the terms written down first.
It costs you
The scope is agreed and written down before anything starts, and a change to it goes through a documented change note. The DevOps and QA pairing applies to full-team buildouts, not to one person dropped in.
DevOps questions we get asked
The seven that come up in nearly every first call, answered straight.
How much does it cost to set up CI, CD and infrastructure?
- The scope is agreed and written down before anything starts.
- A change to that scope goes through a documented change note, so nothing moves in the build on a re-scope you have not accepted.
- What each stage produces is named up front, so you can tell whether it landed.
- Ask whoever else is bidding for the same three things in writing, and compare those rather than the first number either of us says.
Do you offer 24/7 support with a guaranteed response time?
- Seventy-two working hours is eight working days, so say two working weeks if that is what you need to plan around.
- Those targets measure acknowledgement by a named person, not resolution, which is the same thing the Microsoft Initial Response Time measures.
- There is no out-of-hours paging.
- After launch you are on an Annual Maintenance Contract or a lean-mode arrangement, both through the same documented change-note process that governs delivery against agreed scope.
- It is worth knowing what the large providers commit to before you judge that. Google Technical Support Services Guidelines, clause 3.5, states that Google may respond to a request by acknowledging receipt. It also states that the customer acknowledges Google may be unable to provide answers to, or resolve all requests.
- Cloudflare publishes a two-hour P1 support response for Enterprise and publishes no support response time for Business.
- Heroku publishes no numeric response-time commitment at any tier.
- If you need a paged contractual response time, buy it from a provider that publishes one.
What happens to our pipelines when the engagement ends?
- We write documentation from the start of the project, covering the deployment and the CI/CD pipelines.
- The handoff is a process rather than a hostage situation. You have that freedom from day one rather than at the end of a notice period.
- So far, no client has needed to use that exit path.
- It matters anyway, because of what happens when documentation is missing: NIST SP 800-53 Rev. 5 control SA-5 names lack of support from developers and contractors as a cause of absent system documentation. It puts the cost of recreating it on the organisation that inherits the system. That bill lands on you, not on the provider who left.
What uptime do you commit to?
What outcome can you actually promise?
Do we have to give you production access?
- If you already have Git hosting, cloud accounts and storage, we work inside them. If you do not have them, we provision them, set them up, and hand the keys over to you.
- Code lives in your repositories and deployments live in your cloud accounts.
- If you offboard us, you keep everything, including the operational runbook.
- Who gets which level of access, and to what, is set per engagement during scoping. It is written down before anyone touches anything.
- One thing we will not do is claim more than we have: Atyantik does not certify against compliance frameworks.
How quickly can you get something running in production?
Ask us yours
Tell us what you are running
Send the shape of your setup and what is going wrong with it. You get a straight answer about whether we are the right people for it. It comes from a named person who has read what you wrote. This starts a conversation. It does not produce a quote, because nobody can price a system they have not looked at. Any number we gave you now would be one we would have to take back.
- No sales sequence
- A named person replies
- This form does not produce a quote