DOLMIR

Method

The problem first. Then the technology.

An AI project almost always fails the same way: the technology is chosen before the work is understood. The method is built to make that mistake hard to make — nine steps, from the first meeting to a system that improves over time.

  1. 01

    Observe

    A conversation of twenty-five minutes about the process that costs you most. Four questions, no sales pitch. At the end we both know whether it makes sense to continue.

    OutcomeAn honest assessment, even a negative one.

  2. 02

    Understand

    We analyse recent real cases — volume, times, channels, outcomes. This is the step that turns impressions into numbers.

    OutcomeThe baseline, on your data.

  3. 03

    Map

    We reconstruct the process as it is today: who does what, with which tools, where it stalls. Short interviews with the people who actually do it, not the people who describe it.

    OutcomeThe current process, documented.

  4. 04

    Design

    We define what gets automated, what stays human and which controls are needed. The scope is written down: what is included and, above all, what is not.

    OutcomeTechnical specification and a fixed-price proposal.

  5. 05

    Build

    Before building everything, we test the accuracy on your historical documents: if it does not reach the agreed threshold, we do not proceed. Then the system is built to the written scope.

    OutcomeAccuracy measured, not claimed.

  6. 06

    Connect

    Integration with existing systems where it is reliable, assisted entry where it is not. The management system stays where it is.

    OutcomeThe system inside your work.

  7. 07

    Verify

    The system runs on real cases, with a person approving every outcome. Every difference between expected and actual is measured and corrected before going further.

    OutcomeOutcomes verified, one by one.

  8. 08

    Launch

    Training for the people who will use it and thirty days of reinforced support. The indicators stay on, so you can check the result without having to take our word for it.

    OutcomeA system in production, measured.

  9. 09

    Improve

    Accuracy degrades when incoming documents or markets change. Periodic review and continuous tuning.

    OutcomeAccuracy maintained over time.

Engagement

It starts with something small.

The first step is a fixed-price process survey, agreed before we start and fully credited if you go ahead with the implementation. Its job is to produce the numbers to decide on — including, if that is where they point, the number that says not to proceed.

The cost of an implementation depends on the scope and is set at the end of the survey, never before.

Tell us the problem

Handover

What stays with you.

  • The code

    The repository is yours. Another developer must be able to pick up the work without us.

  • The documentation

    How it works, where to change it, what happens when something changes.

  • The measurement

    The indicators stay on, so you can check the result without having to take our word for it.