How we prove it: what you can check before you pay

Do not take our word for it. Here is every way to check how we work and how good the result is, from a first listen to acceptance tests against criteria you set yourself.

Eight checks, from free to contractual

  1. Listen to the voices

    Real recordings of our voices in call scenarios, in studio quality and as they sound on a phone line. No sign-up needed.

    Listen →
  2. See the console

    Real screenshots of the control room with a demo client and invented data: agents, campaigns, calls, recordings and the audit log.

    See the platform →
  3. Check the measured numbers

    Our lab results are published with their conditions. Your numbers depend on hardware, language model and scenario, so we measure them with you.

    See the measurements →
  4. Take a live demo call

    In a call with our engineers you see the system working and ask what you want. You talk to the people who built it, not to a script.

    Book a call →
  5. Get a free sample agent

    Tell us your scenario and we build a short sample agent for it, so you judge the result on your own case.

    Request a sample agent →
  6. Test your own documents or calls

    For ScanSystems we offer a test on a small set of your documents. For the Web-based AI Callcenter, a test call with an AI agent.

    Book a call →
  7. Read the security documents under NDA

    We share the data-flow and security documentation for your deployment under NDA, so your security team can check it.

    See security →
  8. Accept against criteria you set

    In a project we agree measurable criteria first, for example containment, transfer rate, caller satisfaction and response time. We test against them before go-live.

    See the implementation steps →

What a fair acceptance test looks like

A demo shows what a system can do on a good day. An acceptance test shows what it does in your company. So before a project starts we write down together what success means, in numbers you choose, and how each number will be measured. Those criteria become part of the project, and the system goes live only after it has been tested against them.

For a voice agent, the usual measures are the share of calls resolved without a person, the transfer rate and the reasons for transfers, caller satisfaction, response time, and booked meetings or confirmed appointments. For document processing they are how many fields are extracted correctly on your own documents, and how much checking your staff still need to do. For custom software they are the agreed functions, tested on your data, by your people.

Your own team takes part in the test. They use the system with real scenarios, report what does not work, and we fix it. Because the criteria were agreed in advance, nobody has to argue later about whether the result is good enough, and you can see for yourself whether the numbers on this site hold in your case.

What we have measured ourselves

  • 50,000+ simultaneous calls on one enterprise GPU server in our lab.
  • 136 ms median from the reply text being ready to the first sound of the voice.
  • 1.4 s median from the end of the caller's sentence to the first sound of the answer, in test calls with a cloud language model.

Lab measurements from 2026-10, not promises for your project. The conditions are on the measurements page.

What about references?

Our systems run in production at organisations in Bulgaria, and the names are under NDA. That is why you will not find customer logos or testimonials on this site.

When a project is serious, we talk about references in person, under NDA and with the customer's agreement. We do not publish ratings or statistics we cannot source.

Already running in production at organisations in Bulgaria — names under NDA.

Frequently asked questions

Can we test AI Fortis before signing a contract?

Yes. You can listen to the voices, see the console, take a demo call, get a free sample agent and read the security documents under NDA before you commit.

How do we know the system meets our requirements?

We agree measurable criteria before the project starts and test against them before go-live.

Can you share customer names?

Not publicly, because they are under NDA. References are discussed in person, under NDA and with the customer's agreement.

Are the numbers on your site guaranteed?

No. They are lab measurements with published conditions. We measure your own numbers in the pilot.

Pick the check that matters most to you

Tell us what you need to see. We will show it on your case in a short call.