Services and prices
Two kinds of work: testing what your AI chatbot says, and test automation for the rest of your product. Both at a fixed price, agreed before we start.
AI chatbot testing
Our main service. We test what your bot says to customers using the PROBE method, at fixed prices in US dollars.
Launch pricing. We’re a new studio, so our first three Snapshot clients pay $990 instead of $1,490, and our first three Audit clients pay $3,900 instead of $6,900. The Audit launch price comes with one request: a case study naming your company, which you approve before publication, and a reference.
PROBE Snapshot
A fast, evidence-based look at two areas of your bot. The easiest way to find out whether we’re useful to you.
What we do
- Test two areas of your choice. By default we pick the two where bots most often get companies into trouble: identity and AI disclosure, and promises and capability limits.
- Run adaptive conversations through your live chat, as your customers use it, with personas agreed with you.
- Check every factual claim the bot makes in those areas against your website and policies.
What you get
- A 4-page report: each finding with the bot’s exact words, the date, severity, the behavior you should expect instead, and a suggested fix.
- A 30-minute debrief call with your team.
- A recommendation on what, if anything, to test next.
What we need from you
- A link to the bot, or test access if it sits behind a login.
- A signed one-page testing authorization. We send it to you.
- One contact person, and a short form about your users and your concerns. It takes about 10 minutes.
PROBE Audit
All eight areas, a score you can track over time, and fixes ready to use.
What we do
- Test all eight areas plus one industry module: companion and emotional safety, health, finance, e-commerce and refunds, or products for older adults.
- Run adaptive sessions with several personas, weighted by the risks that matter for your users.
- Save the most revealing questions word for word as frozen probes, and score the bot with the PROBE Score.
What you get
- The persona list and the one-page risk map we tested against.
- A full report with every finding, its evidence, severity and expected behavior.
- A PROBE Score from 0 to 100, and a breakdown by area.
- Suggested changes to your system prompt and knowledge base, written so your team can apply them directly.
- 20 frozen probes, saved test conversations with pass and fail criteria. They’re yours to keep and rerun.
- One retest within 90 days, to confirm the fixes worked and to update the score.
- A debrief call with your team.
PROBE Guard
Coming soon
Automated regression for your bot, so fixes keep holding after every change.
We will turn your frozen probes into automated tests that run through the same chat interface your customers use, not just the model’s API. That catches problems an API test can’t see, such as an outdated knowledge base in production, a broken handover to a human, or a prompt that differs from the one in your repository. The tests will run in your build pipeline on every release, and a wrong answer will block the release.
Tell us if you’d like to be one of the first to use it.
Test automation and QA
Led by our test automation architect, who has spent 10 years building and stabilizing automated tests for web applications and APIs. Each engagement has a fixed price, quoted after a short call, because the scope depends on your application.
Test automation foundation
A working automated test suite for the key journeys of your product, built so your team can own it.
Who it’s for
Teams with little or no test automation, often about to hire their first QA engineer, who want a solid starting point instead of a hire’s first months of trial and error.
What we do
- Agree with you on the user journeys that matter most, such as sign-up, checkout or the core workflow of your product.
- Build a test framework with Playwright or Selenium, in the language your team already uses: TypeScript, Java, C# or Python.
- Write the first set of automated tests for those journeys, for the web interface and the APIs behind it.
- Run the tests in your build pipeline, so they check every change.
What you get
- A working framework in your repository, with tests for the agreed journeys.
- Documentation your team can follow without us.
- A handover session, so your developers or testers can extend the suite themselves.
Flaky test rescue
For test suites that fail at random, run slowly, or have stopped being trusted.
Who it’s for
Teams whose automated tests fail for no clear reason, block releases, or get ignored because nobody believes them anymore.
What we do
- Analyze which tests fail at random or run slowly, and why: timing and waits, test data, the test environment, or fragile ways of finding elements on the page.
- Fix or rewrite the worst offenders, starting with the ones that block releases most often.
- Set simple rules for writing stable tests, so the problem doesn’t come back.
What you get
- A test suite your team can trust again, and that runs faster.
- A short report on the causes we found and what we changed.
- Guidelines for writing new tests the same way.
Test strategy review
A clear view of how your team tests today and what to automate first.
Who it’s for
Teams that aren’t sure what to test, what to automate first, or whether their testing effort is paying off, for example before a major release or a new hire.
What we do
- Talk to the people who build, test and release your product.
- Review your current tests, tools and release process.
- Weigh what’s worth automating against what’s better tested by hand.
What you get
- A short report: where you are, the main risks, and what to automate first.
- A realistic plan for the next three months.
- A debrief call with your team.
Training and coaching
For testers or developers who are taking over test automation.
Who it’s for
Teams where someone is learning to write and maintain automated tests and wants to get it right from the start.
What we do
- Run hands-on workshops on your own code base, not on toy examples.
- Cover stable test design, test data, page objects and running tests in your build pipeline.
- Coach your people remotely while they write their own tests, with reviews and answers to their questions.
What you get
- People on your team who can write, run and maintain automated tests on their own.
- Workshop material based on your product, yours to keep.
How an engagement runs
- A 15-minute callWe ask about your product, your users and what worries you, and tell you whether we can help.
- A one-page proposal, the same dayScope, fixed price and dates. No surprises later.
- Authorization and accessFor chatbot testing, you sign a one-page testing authorization. For automation work, we agree on access to your code and test environments.
- The workWe test or build in the agreed window and keep you posted. For a Snapshot, you don’t need to do anything during this time.
- Delivery and debriefYou get the report or the code, we walk your team through it, and you decide what’s next.
Questions about pricing, data and working with us
Do you need access to our systems or code?
Not for a Snapshot. We test through the same chat your customers use, after you authorize it in writing. For an Audit, a staging environment helps if you have one, but it isn’t required.
What if you don’t find anything material?
Then you don’t pay for the Snapshot. If the report has no finding rated medium or higher on our severity scale, there’s no invoice.
Will testing disturb our customers or our support team?
Very little. We use test identities and invented details, and we agree on the time window in advance. If your bot hands conversations over to people, we tell your support team when we’re testing, so they can spot and close our test chats quickly.
How do you handle our data and the transcripts?
Test transcripts are shared only with you, used only for your engagement, and never published unless you approve a case study. We’re happy to sign an NDA before we start.
If you share real conversation logs with us, we sign a data processing agreement first. As a company based in the EU, we follow the GDPR.
Will your report make us compliant with the EU AI Act or state chatbot laws?
No report can promise that, and we don’t certify compliance. We show you how your bot behaves, with evidence. Your lawyer decides what that means legally, and our findings give them something concrete to work with.
Which chatbots do you test?
Customer-facing chat assistants in English: support and sales bots, help-center assistants and AI companions. It doesn’t matter whether the bot runs on an off-the-shelf support platform or your own LLM setup, because we test it the way your users meet it.
How do invoicing and payment work?
We invoice in US dollars from our company in the EU, and you pay by bank transfer. We can also invoice in pounds sterling or euros. A Snapshot is invoiced with the report. An Audit is invoiced 50% at the start and 50% on delivery. Invoices are payable within 14 days, and each side pays its own bank fees.
Can a launch price be combined with the Snapshot credit?
Yes. If you pay $990 for a Snapshot and book an Audit at the launch price within 90 days, the Audit costs $2,910. Launch prices apply to our first three clients for each service. The Audit launch price comes with a request: a case study naming your company, which you approve before publication, and a reference.
How do you price test automation work?
Every engagement gets a fixed price, agreed in writing before we start. Because the scope depends on your application, we quote after a 15-minute call and, if needed, a look at your current tests. We usually invoice 50% at the start and 50% on delivery.
Which technologies do you work with?
For test automation: Playwright and Selenium, with TypeScript, Java, C# or Python, plus Cucumber for teams that write tests as business scenarios, and Postman or SoapUI for REST and SOAP APIs. We work in your repository and your build pipeline, with Git.
Not sure which one you need?
Tell us about your bot or your tests, and we’ll recommend the smallest step that answers your question. Sometimes that’s a Snapshot, and sometimes it’s nothing at all.