FDE Műhely
HU Let's talk

Services

Three phases, in this order

Each is a prerequisite for the next. You can enter at any of them, but skipping one is a bad idea: without evals you don't know whether it works, without discovery you don't know whether you're building the right thing.

01

AI discovery

2–4 weeks · fixed price

We map how one department's processes actually run, and tell you what's worth automating, and what isn't.

We sit with the team and watch the work happen. Not by survey: by observation. We record the steps, the systems, the time spent, and most importantly the exceptions. What you get at the end is a document you can build from.

What you get

  • Operating map, step by step, with exception branches
  • Time and capacity data per process
  • ROI matrix: what to automate, what to leave, and why
  • Architecture for the proposed solution
  • Estimate of time and cost to deploy

Discovery stands on its own. It commits you to nothing, and we don't try to turn every discovery into a project. If the answer is that it isn't worth it, we write that down too.

02

Evals and measurement

2–3 weeks

Before anything goes live, we give you a number for whether it works.

With a non-deterministic system, "looks good" is not a result. We assemble a golden dataset from your own historical cases, define what a correct output means, and run it. What you get is a number, plus the failures categorised: where they come from, why, and which are fixable.

What you get

  • Golden dataset built from your own cases
  • Measurement report: passed / failed per run
  • Failure modes broken into categories
  • The threshold: what runs unattended, what goes to a human
  • A regression set for future model changes

This set stays yours. It's the reason switching to a new model six months from now becomes a measurement rather than an article of faith.

03

Deployment

4–8 weeks

We build on what already runs at your company. Starting in shadow mode, then step by step.

We don't migrate systems. We build on the APIs of your existing ERP, CRM and document store. The system first runs in shadow mode alongside your team: they see what it would propose, but they still decide. Once the measurements hold, it earns autonomy, one category at a time.

What you get

  • Integration with your existing systems
  • Shadow mode, then staged go-live
  • Full audit trail: every step reviewable
  • Alerting, SLA monitoring, cost monitoring
  • Documentation and training for your internal team

If you can't inspect what the system did, you won't trust it, and you'd be right. That's why logging isn't optional here.

Arrangements

How you can engage us

Embedded FDE

One senior engineer with your team two to five days a week, for months. They map the process, they build it, and they own it in production. The most common arrangement.

Good when you'll run several processes in sequence and want the knowledge to stay in-house.

Full team

An FDE, a data engineer and an eval owner on one programme. Several workflows in parallel, sharing one architecture.

Good when you're taking a whole department across, not a single process.

Discovery only

Fixed-price assessment, no obligation. At the end you decide whether to continue with us, with your own team, or not at all.

Good when you're still before the decision and need facts to make it.

Rates are on the pricing page, with real bands.

Which process should we start with?

Tell us which department burns the most manual hours. We'll come back with a concrete proposal for what discovery would look like at your company.

Get in touch