Compare your options
An AI tool can suggest a likely cause, but only an engineer can answer for it.
There are four common ways to cover production at night: hire your own SREs, buy an AI SRE tool, use a cloud MSP (managed service provider), or use us. Here is how they differ, and when each one is the right choice.
The hiring, AI tool and MSP columns describe what is typical, not any one company. Ask any provider, including us, to put their answers in writing.
| Topic | Hire your own SREs | AI SRE tool | Cloud MSP | Tracefox |
|---|---|---|---|---|
| Who answers at 3am | Your engineer on the rota | Software investigates first; your own engineer is still paged | A first-line service desk, often shared, escalating to specialists | An in-house Tracefox SRE, every time |
| What you get on each alert | Whatever your engineer finds | A likely cause with evidence, for your engineer to check | A ticket, sorted by priority | A likely cause checked by an engineer, and a fix if a runbook you approved covers it |
| Response time in a contract | Your internal policy | Usually none for people, because the tool does not supply them | Set per priority, sometimes a target rather than a commitment | 15 min to acknowledge, 30 min to cause and severity, 24/7/365 |
| Changes to production | Your engineers | Depends on the tool; check what it can run on its own | Their engineers, usually with administrative access | Only runbooks you approved; new issues come to you with the evidence |
| Work on why alerts happen | If you staff for it | Suggestions; someone on your side does the work | Usually a separate project | One or two goals each quarter, included in the fee |
| Time to start | Months per hire. Google’s published example is about 25 weeks to a fully productive SRE | Days to weeks, once connected to your data | Weeks, depending on onboarding | Cover starts in week one, from the alerts you already have |
| Cost | At least eight salaries for single-site night cover [S$ figure, with source] | A licence by seat or usage, plus your own on-call | A custom quote, often a share of cloud spend | From S$1,500 a month for up to 5 production services |
When we are not the answer
Choose something else if any of these apply.
-
Hire your own SREs if
- Reliability is the product itself, and you can fund a team of eight or more over time.
- You already have three or more SREs and can staff a rota.
-
Buy an AI SRE tool if
- You already run a healthy 24/7 rota and want investigations to go faster.
- You want software your team operates, not a service that answers for you.
-
Use a cloud MSP if
- You want one supplier to fix everything, new problems included, without paging your engineers.
- You need helpdesk, patching, licences or on-premises support.
Already using an AI tool?
The tool finds a likely cause, but someone still has to wake up and act on it.
An AI SRE tool can hand your engineer a likely cause faster. It does not take the pager, sign a response time or work through your alert noise each quarter. That is the part we do.
[Confirm: can Tracefox use the findings from a customer’s existing AI tool, such as one built into their monitoring? If yes, say so here.]
Ask any AI tool vendor
- Who is paged when the tool has finished investigating?
- What can it change in production without a person deciding?
- How often was its first likely cause right last month, and can we see that figure every month?
- Who works on reducing the alerts it investigates?