What agentic testing means for your team
Agentic testing is software testing done by AI agents that plan, run, and report tests, with a person checking the result. You give the agent a goal, and it works out what to test.
How agentic testing works
Older test tools start from a script. An agent starts from what the software is meant to do.
- 1
It reads the goal
It starts from a ticket or a user flow, so no script has to come first.
- 2
It plans the test
It works out what to check, and which accounts and data it needs.
- 3
It runs the test
It uses the product the way a person would, and copes when a screen changes.
- 4
It reports back
It says what passed, what failed, and what it did not test, with the proof.
Three ways to automate testing
The difference is how much of the work a person still does.
| Scripted automation | AI-augmented tools | Smoketest | |
|---|---|---|---|
| Who plans the tests | Scripted automationAn engineer | AI-augmented toolsAn engineer, with AI suggestions | SmoketestIt does, from the ticket |
| Who writes them | Scripted automationYour engineers | AI-augmented toolsYour team, faster with AI | SmoketestIt does |
| When a screen changes | Scripted automationSomeone fixes the test | AI-augmented toolsThe tool repairs many of them | SmoketestIt follows the goal |
| When it runs | Scripted automationIn your pipeline | AI-augmented toolsIn your pipeline or their cloud | SmoketestWhen a ticket reaches QA |
| Where results land | Scripted automationA pipeline log | AI-augmented toolsThe tool's dashboard | SmoketestResult, recording, and steps on the ticket |
| Who owns the release | Scripted automationYour team | AI-augmented toolsYour team | SmoketestA named person on your team |
When each one fits
Most teams end up using more than one. Here is how we would decide.
Scripted automation fits when
- Your engineers want tests next to their code
- A few flows need the same exact check
- Someone has time to look after the suite
AI-augmented tools fit when
- You already have people writing tests
- You want them to write and fix tests faster
- You want to stay close to the test code
Smoketest fits when
- You want each ticket tested as it reaches QA
- Nobody has time to write or look after tests
- You want the proof on the ticket
What it costs
Agents are priced by what they run or by the flows they cover, so they start well below a person's salary. A managed service adds the people around them.
Market ranges from BLS, May 2025, Bug0, and Vendr. Your quote will vary.
Passed · 4 of 4 requirements · 2m 26s · Attempt 2
- ✓Asks before it cancels
- ✓Says when the plan ends
- ✓Calls cancel_plan once
- ✓Billing shows the plan ending
Not tested: refund requests, because the ticket only covers cancelling.
Watch the recording · View attempt
Transcript and Billing both look right. This goes out today.
Here is what an agent leaves behind.
It did not write the code
Testing from the outside means your team gets an independent check.
It remembers your product
What it learns on one ticket shapes how it tests the next.
It shows its work
Every result comes with the recording, the steps, and what it skipped.
Questions about agentic testing
Looking at the wider market? Read Gartner's reviews
It is software testing done by AI agents that plan, run, and report tests. You give the agent a goal, and it works out what to check. A person reviews the result.
AI-augmented tools help a person work faster, with suggestions and tests that repair themselves. In agentic testing, the agent does the testing and a person checks the result.
No. Agentic testing uses agents to test your product. Testing AI agents means checking an agent you built. Smoketest does both, and each has its own page.
Gartner published its first Magic Quadrant for AI-Augmented Software Testing Tools in October 2025. Its review page now describes that market as transitioning to Agentic Software Quality Assurance Platforms.
Keep the ones you have. Scripted tests are good at repeating one exact check. An agent covers the new work on each ticket, where nobody has written a test yet.
It starts at $399 a month for 1,000 runs. Each run is up to 10 minutes, and a longer test counts as two runs or more, one for every 10 minutes. If your team needs more, we design a Scale plan with you, so you stop buying extra runs.
Every result links to the recording, the steps, and the reasoning, and says what it did not test, so you can check its work the way you would check a teammate's.
A named person on your team. Smoketest gives them the evidence: what passed, what failed, the recording, and what it did not test. It never signs off a release.
See what an agent does with your ticket.
Book a call and we will run one with you. You get the result and the recording the same day.