Key takeaways
Budget about $135,000 a year for a QA automation engineer, and about $150,000 for a senior one.
That's the median listed salary in 101 US job postings QA Wolf reviewed in November 2026. SDET titles were listed higher than QA automation titles at the same level.
If you build test automation in-house, budget for infrastructure and AI, not just salaries.
For 200 web tests and 50 iOS tests run on every pull request, browsers, real devices and AI tools come to about $15,000 to $20,000 a month at list prices.
Before you sign with a test automation vendor, find out what one unit of billing covers.
Ask whether infrastructure, compute time and AI usage are in the price, and whether third-party costs are passed through at cost or marked up.
Plan for test maintenance from day one, whoever writes the tests.
In a QA Wolf experiment, half of an unmaintained end-to-end test suite stopped working within six weeks as the apps changed.
Compare hiring and outsourcing on the work that comes back every release.
Investigating failures, fixing broken tests and running the infrastructure recur every time you ship, and they decide which option costs less over a year.
As of 2026, US job postings list QA automation engineers at about $135,000 a year, and senior ones at about $150,000. If you hire in-house, the tests also need infrastructure, AI and hours of upkeep. If you outsource test automation, you pay a vendor's bill instead, and what it comes to depends on what the vendor charges for and what the contract leaves out.
How much does it cost to hire a QA automation engineer?
About $135,000 a year in salary. On top of that, you pay for the infrastructure and AI the tests use, and for the hours it takes to keep them working. The frameworks themselves are free, since Playwright and Appium are open source.
Senior roles and SDETs are listed highest
Median listed salaries in 101 full-time US job postings that QA Wolf reviewed in November 2026, from Built In, Dice and company career pages:
| Role | Mid-level | Senior |
|---|---|---|
| QA automation engineer | About $115,000 | About $137,000 |
| SDET (software development engineer in test) | About $137,000 | About $176,000 |
| All 101 postings | About $120,000 | About $150,000 |
Motion Recruitment's 2026 Tech Salary Guide is close, with base salaries of $99,375 to $128,750 for a mid-level QA automation engineer and $129,263 to $155,854 for a senior one. Recruiting, ramp-up and management add to the cost.
You also pay for infrastructure and AI usage
Here's what that comes to for a typical suite. Say you have 200 web tests and 50 iOS tests, each taking about 5 minutes, and the suite runs on every pull request. To finish a run in about 10 minutes, you need about 100 browsers and 25 real iPhones running in parallel, which matches QA Wolf's analysis of parallel runs.
| What you pay for | Rough cost a month |
|---|---|
| Browser grid: 100 browsers at once, at $99 to $149 each | $9,900 to $14,900 |
| Real iOS devices: 25 at once, at about $199 each | About $5,000 |
| AI coding agents: two automation engineers, at about $150 to $250 each | $300 to $500 |
| Total | About $15,000 to $20,000 |
That's about $180,000 to $245,000 a year before anyone's salary. The prices are list prices from the major device clouds (billed annually) and Anthropic's average Claude Code spend per developer in enterprise deployments, as of November 2026. CI minutes to drive each run add a little more. AI tools are already part of the job, since four in ten of the postings above mention them.
Your engineers also spend time building and keeping up everything else the suite needs:
- test environments stable enough to test against
- test accounts and data that reset between runs
- email and SMS inboxes for sign-up, login and two-factor flows
- a tool to manage the tests and their results
Maintenance and failure investigation take most of the time
Writing a test happens once. Investigating failures, fixing broken tests and reporting bugs come back every release. To see how many engineers that upkeep takes, fill in this worksheet with your own numbers. The example uses the same suite as above, and the times are what QA Wolf measured for its own QA engineers across 3.5 million tests, before AI agents.
| Step | How to work it out | Example |
|---|---|---|
| Test runs a month | Tests in the suite × suite runs a month (about 10 a workday here) | 250 × 200 = 50,000 |
| Failures a month | Test runs × the share that fail (1% here) | 50,000 × 1% = 500 |
| Time investigating | Failures × 17 minutes | 500 × 17 min ≈ 140 hours |
| Time fixing broken tests | Failures caused by product changes (half here) × 45 minutes | 250 × 45 min ≈ 190 hours |
| Time reporting bugs | Real bugs (1 in 10 failures here) × 30 minutes | 50 × 30 min = 25 hours |
| Engineers for upkeep | Total hours ÷ 160 working hours per engineer a month | 355 ÷ 160 ≈ 2 engineers |
That's about two engineers, or about $270,000 a year in salary at the 2026 median, before anyone writes a new test. AI tools speed each step up now, but the steps still come back every release.
Skipping the upkeep doesn't save the cost. In a QA Wolf experiment, half of an unmaintained test suite stopped working within six weeks as the apps changed. Flaky tests that nobody fixes teach the team to ignore the suite. When developers own the end-to-end tests, QA Wolf estimates that they spend 20% to 40% of their time on QA work instead of features.
How do test automation vendors charge?
Vendors charge by the hour, the test, the run, compute time or AI tokens, and each contract decides what that unit includes. Compare quotes on everything each one covers. For choosing a partner, from who investigates failures to who owns the tests, see the QA outsourcing guide.
What to watch for depends on how the vendor bills
- By the hour: the bill grows with upkeep, so a suite that breaks often costs more every month. Freelance QA engineers on Upwork charge $20 to $60 an hour, and 34 US contract postings for QA automation engineers and SDETs listed a median of about $62 an hour (November 2026).
- By the test: you pay for every test in the suite, whether it runs once a month or on every pull request. Check what counts as one test, and whether fixing a broken one costs extra.
- By the run: the bill tracks how often you test, so price the runs you'll actually do, including one for every pull request. Check what a run includes, such as investigating failures, retries and fixes.
- By compute time: slow tests and retries cost more. Check whether running tests in parallel changes the rate.
- By AI tokens or credits: the bill grows with how much AI work each run sets off, which is hard to predict month to month. Check for a cap.
The price may leave out infrastructure and AI usage
Some contracts include the runners, browsers, devices, parallel runs and AI usage. Others leave you to pay for them, which puts the costs above back on your bill. Watch for limits on how many tests can run at once, and for fees above those limits. Vendors that rely on other companies' services, such as device clouds or AI models, either pass those costs through at cost or mark them up. Either way, they're part of what outsourcing costs you.
Your team's time still counts
Your engineers still review results, answer the vendor's questions and decide what to test. If the vendor writes tests and hands them over, your team also takes on the upkeep of tests it didn't write, and those tests decay fast.
Ask these questions before you sign
- What does one unit of billing cover? An hour, a test or a run can include writing, running, investigating and fixing, or only some of those.
- Is the infrastructure in the contract? Ask about browsers, real devices, CI minutes and how many tests can run at once.
- Do you charge for compute time or AI usage? Find out how it's metered and whether there's a cap.
- Are third-party costs passed through at cost or marked up? Device clouds, AI models and test management tools all have bills of their own.
- Is maintenance in the price? Fixing tests that break when the product changes is the biggest recurring cost, so find out whether it's billed separately.
- What happens to the bill if tests run on every pull request? A price that looks low for a weekly run can grow fast once every change starts a run.
Is it cheaper to hire or outsource test automation?
It depends on who carries the work that comes back every release: investigating failures, fixing broken tests, and paying for the infrastructure and AI each run uses. A low salary or hourly rate can still cost more once you count that work.
| Hire a QA automation engineer | Outsource test automation | |
|---|---|---|
| What do you pay for? | Salary, about $135,000 a year per engineer (the median of US job postings, November 2026). | Whatever the vendor bills for: hours, tests, runs, compute time or AI tokens. |
| What gets left out of the estimate? | Runners, grids, device clouds, CI minutes and AI usage, plus recruiting and management. | Infrastructure outside the contract, marked-up third-party costs and your team's review time. |
| What happens as you test more often? | Infrastructure and upkeep grow with every run, and headcount grows with the suite. | It depends on the billing unit, so ask what testing every pull request would cost. |
| Who keeps the tests working? | Your engineers, on every release. | The vendor, if maintenance is in the contract. Otherwise, your team. |
Whoever does the work, aim to test every practical user flow: each distinct thing a user comes to do, plus the variations that change the outcome. Start with the critical flows, and run the suite on every pull request.
How does QA Wolf lower the cost of automated testing?
QA Wolf's AI agents do the work that makes test automation expensive, whether you hire for it or outsource it. They map your app, write the end-to-end tests and run them on QA Wolf's infrastructure. They investigate every failure, keep the tests working, and ask you when they need an answer.
Teams without automation engineers don't need one
Anyone on your team can start now and ask the agent to map your app. The agent opens a browser, explores each feature and maps your end-to-end workflows as plain-English outlines. Then a product manager or a manual tester can describe each test in plain language, and the agent writes it. Nobody has to write test code, and there's no one to hire and nothing to outsource.
Teams with automation engineers need fewer of them
The agents write, run, investigate and fix the tests, so the same suite needs fewer engineers. The engineers you have spend less of their time on it, because the agents ask them only when they need an answer.
- Tests in code you own. The agents write standard Playwright and Appium tests that your engineers can read, change and take anywhere.
- Every failure investigated. The agents work out whether a failure is a real bug or a test that broke because the app changed. They fix broken tests, with unlimited repairs, and report real bugs to you. When your tools don't have the answer, they ask a person.
QA Wolf's agents do this work better than general-purpose coding agents. On 1,567 evals across web, iOS and Android apps, QA Wolf passed 93%, against 78% for Claude Code and 74% for Codex.
Salesloft is a perfect example. With 2,000 automated tests and 300+ running in parallel on every pull request, Salesloft didn't need to hire seven SDETs, and it saves more than $750K a year in QA engineering.
Infrastructure and AI usage are included in each test run
Every test runs in parallel on QA Wolf's infrastructure, so there are no browsers, devices or test runners to manage. QA Wolf's pricing is per test run. It includes the AI and compute each run uses, from investigating failures to repairing tests, so there are no AI seats or tokens to budget for.
If you'd rather QA Wolf take on the whole job, its full service provides QA automation services.
Start now and ask the agent to map your app and automate your critical flows, or talk to sales.
Frequently asked questions
What is an SDET?
An SDET (software development engineer in test) writes code to test software: the automated tests, the framework they run in and the tooling that runs them. For end-to-end tests, QA Wolf's AI agents do that work. They write, run and maintain the tests in Playwright and Appium code you own.
How much do QA engineers earn?
In 101 full-time US job postings with listed pay that QA Wolf reviewed in November 2026, QA automation engineers and SDETs were listed at a median of about $135,000 a year. Mid-level roles came in at about $120,000 and senior roles at about $150,000. The US Bureau of Labor Statistics median for all software quality assurance analysts and testers, manual testers included, was lower, at $104,300 in May 2025.
How many QA engineers do you need?
Before AI agents took on test work, QA Wolf's rule of thumb was at least one QA engineer for every six product developers. For a closer estimate, count the upkeep: test runs a month, times the share that fail, times the time to investigate and fix each failure, divided by about 160 working hours per engineer. A 250-test suite that runs about 10 times a workday needs about two engineers for upkeep alone. With QA Wolf's AI agents writing, running and maintaining the end-to-end tests, you need fewer QA engineers, and a team with none can still build the tests.
How do you hire a QA automation engineer?
Give candidates a real task from your product, such as an automated Playwright test for sign-up. Then ask how they'd keep it passing as the product changes. Budget about $135,000 a year in salary, the median of US job postings in November 2026, plus the infrastructure and AI tools the tests need. For short-term work, freelance marketplaces such as Upwork list QA engineers by the hour.
How much does QA testing cost?
It depends on who builds and maintains the automated tests. US job postings list an in-house QA automation engineer at about $135,000 a year (November 2026), plus infrastructure and AI usage. A vendor's bill depends on how it charges and what its contract includes. QA Wolf charges $1 per web or Linux Electron test run and $2 per mobile or desktop run, with AI and retries included, as listed on its pricing page.
Is AI going to replace QA engineers?
AI agents now do much of the repetitive QA work, such as writing, running and maintaining end-to-end tests and investigating failures. People still decide what matters, judge new features and answer the agents' questions when intent is unclear. QA work shifts toward those decisions.
Is QA Wolf better than building an in-house QA automation team?
For most teams, yes. QA Wolf takes on the parts of an in-house team that cost the most: the hours spent investigating failures and fixing tests, and the infrastructure and AI the tests need. A team without automation engineers doesn't need to hire one, and a team with them needs fewer. QA Wolf's AI agents write, run and maintain the tests in parallel on QA Wolf's infrastructure, in standard Playwright and Appium code you own. On 1,567 evals across web, iOS and Android apps, QA Wolf passed 93%, against 78% for Claude Code and 74% for Codex.