ostravský lab_
CSEN

01 / Services

Validating an AI idea

Most failed AI projects did not start with the wrong technology. They started with an unclear brief. Before choosing a model, we help you state the problem and how you will recognise a useful result.

Can AI help with this problem, and what needs testing before we invest?

Write to the Lab See the example protocol

01 / About the service

What we do for you.

Together we sharpen the brief: what you solve today, how you do it, what data exists and which decision needs evidence. That shows which assumptions the idea depends on and which of them can be tested quickly and cheaply.

The result is not a presentation about AI. It is a test plan describing what will be measured, on which data, against which measure of correctness, and which outcome leads to continuing or stopping.

When this fits

  • You have a specific repeated task done by people today and consider automating part of it.
  • You received an offer for an AI solution and want an independent view of what is realistic for your operation.
  • You want to know in advance what data and cooperation a test will need.
  • You need evidence for a budget decision or a next step.

What we need from you

  • A description of the current process and who performs the task today.
  • A few examples of the input and of a correct result, anonymised if needed.
  • Your expectations for quality, volume and speed.
  • The constraints that apply: data handling, systems, deadlines, budget.

02 / How it runs

How we work

Scope, price and deadline are confirmed before the work starts. The steps follow the size of the brief.

  1. 01

    First conversation

    We go through the task, its context and the decision that needs evidence, and agree what success would mean.

  2. 02

    Stating the assumptions

    We write down the assumptions the idea rests on and the constraints that apply to data and operations.

  3. 03

    Defining the measures

    We set what will be measured and how we recognise a result that is good enough, including the cases a person has to decide.

  4. 04

    Test plan and recommendation

    We order the tests by value and effort and list the open questions that still need answers.

03 / Output

Possible output, as agreed

The exact output is agreed with the scope. We always state the conditions under which the findings hold.

  • A short statement of the problem and the goal.
  • A list of assumptions and limitations.
  • A proposal of suitable technical options.
  • A test plan including the success criteria.
  • A recommended next step and the open questions.

What this does not include

  • It is not an implementation; development is agreed separately.
  • Without representative data the achievable quality cannot be confirmed, only a way to test it.
  • The recommendation holds for the described purpose and conditions; if those change, it has to be revisited.

04 / Practical questions

Common questions about this service.

Basic orientation for a first conversation. The details always depend on the specific brief.

Do we need the data ready?

Examples are enough for the first conversation. A representative sample is needed for the test itself, and we can plan its preparation together.

How long does it take?

It depends on the scope and on the material available. Scope, price and deadline are confirmed before the work starts.

What if the answer is that AI does not fit?

That is a valid result and usually cheaper than a failed project. The recommendation can also be a different approach or stopping the idea.

05 / Services

Other services of the Lab.

The services build on each other. You can start with any of them, depending on where you are.

Write to the Lab

A new version is available.

An update is ready. Load the latest content, or keep reading this page for now.