Frequently asked questions
Short, straight answers about how we test AI products and web apps, what we need from you, how pricing works and how to reach us. If your question isn't answered here, send it through the quote form and we'll answer it by email.
What do you test?
We test AI products and web apps. On the AI side that means chatbots, RAG assistants and agents: accuracy, data leaks, prompt injection and how reliably they behave. On the web side we build Playwright tests around your riskiest user journeys, then hand them over so your team can run and extend them.
How does an AI chatbot audit work?
We agree what a correct answer looks like with you, then build a golden set of questions from your own documents. We run them many times, check facts, leaks and boundaries with code first, and review every failure by hand. You get a ranked report with the exact conversation behind each finding.
Can you test our chatbot without access to our code?
Usually, yes. Most checks run through the same interface your customers use, or through an API, so we don't need your source code. Access to prompts, retrieval settings and logs helps us explain why something fails, and we agree exactly what we need before we start.
How do you test RAG assistants?
We test the search and the answer separately. First we check whether the right passages are found for each approved question. Then we check that every claim in the answer is supported by those passages. We also test what happens when the answer isn't in your documents, and who can retrieve what.
What is prompt injection testing?
Prompt injection is when crafted text, from a user or hidden in a document or web page, makes an AI ignore its instructions. Testing means trying those attacks against your app in a safe environment, recording what works and how often, and giving you a repeatable set of attacks to rerun after every fix.
Do you build Playwright tests in our repo and CI?
Yes. We write the tests in your repository, in readable TypeScript or JavaScript, and set them up to run in your own CI under your accounts. Traces and video are saved when a test fails. You own the code from day one, so your team can keep going without us.
Can you migrate us from Selenium or Cypress?
Yes, in stages. We audit your current suite, group tests by risk, and move one batch at a time. The old and new tests run side by side, and a new test only replaces the original once it fails when we deliberately break the app. Nothing is switched off on trust.
How is pricing decided?
Every project is quoted on its scope, so there's no price list. The main drivers are how many AI features or user journeys we cover, how much material the AI answers from, how much test code already exists, how your pipeline is set up, and how much reporting you need. Start with the quote form.
Do you sign NDAs and keep findings private?
Yes. We're happy to sign your NDA, or a mutual one, before you share anything sensitive. Findings and test data stay between us and the people you name. We don't show your work, your company name or your results anywhere without your written permission.
How do I contact you?
Use the quote form at /contact and we'll reply by email. You can also email support@tbstechservices.com. Privacy and data requests go to privacy@tbstechservices.com. Tell us what you're shipping and what worries you, and we'll come back with questions and a proposed scope.
Ready to find out what is breaking?
Tell us what you’re shipping. We reply by email, and quote each project on its scope.