How we work

How we work: the process first, then the AI

opencream.ai works process-first. Before we choose any technology, we map how the work really runs and measure a baseline. Then we test one hypothesis, build the software, roll it out in a controlled way and measure again. After launch we keep running and improving what we built.

Most of what sets us apart follows from that order of work.

Last updated: 28 October 2026

Why process-first

Why we measure before we build

The habit comes from the lab and the plant floor. Matthias Foerster, our founder, has a background in analytical chemistry and more than 15 years in specialty chemicals and manufacturing. In that world you know a process's numbers before you change it, because otherwise you can't tell whether the change helped.

We apply the same rule to AI. Without a baseline, nobody can say whether a new system saved time or moved the work somewhere else. So we write the starting numbers down with you before the build and check against them afterwards.

The map also decides what we build. We choose the technology once we know where the hours go and which exceptions cause them.

The engagement

An engagement in seven steps

  1. 01

    Discovery call

    A free conversation about your bottleneck. The first question is whether AI is the answer at all. If something simpler fits, such as a standard tool or a change to the process, we say so.

  2. 02

    Map the process and measure a baseline

    We talk to the people who do the work and follow real cases through, exceptions included. Together we agree which numbers describe the process today, for example cases per week or time per case.

  3. 03

    One hypothesis

    We write down the one change we expect and how we will check it, such as more cases handled by the same team. A single hypothesis keeps the project small enough to judge.

  4. 04

    Build

    We build custom software into your workflow and test it on your real cases and languages before it goes live.

  5. 05

    Controlled rollout

    The system starts with part of the volume or one team, and a person checks its output. We widen it when the results hold.

  6. 06

    Measure after launch

    We compare the new numbers with the baseline and share them with you, including what didn't work.

  7. 07

    Keep running and improving

    We host and maintain the system. What we learn from how your team uses it goes into the next version.

We follow the same steps whether we build an automation or an agent.

What we won't do

What we won't do

Some requests we turn down because they don't fit how we work.

  • Strategy decks without a build

    We don't sell reports that end up on a shelf. When we recommend something, we build it and run it.

  • Hourly billing

    We scope each project and agree a setup fee and a monthly fee before we start. If the build takes longer, the price stays the same.

  • Generic chatbots

    A chatbot added to a website rarely removes an operational bottleneck. We build for a defined task in your process.

  • One fixed percentage

    We won't promise a set saving before we have measured your process. Results depend on the process.

  • Promises about servers we don't run

    Our systems are hosted at Hetzner in Germany. We pass on Hetzner's own terms and don't add availability figures of our own.

If one of these is what you need, we are the wrong partner, and we will say so early.

FAQ

Questions about how we work

What is different about opencream.ai?

We map the process and measure a baseline before we choose any technology, then we build the software and keep running it. We don't bill by the hour. When a build fits more than one company, it becomes a product of its own, as TraxAgent and Corial did.

Do you also offer AI consulting?

Only as part of a build. The discovery call is free and looks honestly at whether AI will help. After that we build what we recommend, or we tell you that you don't need us.

Why do you measure a baseline first?

Without a starting value, nobody can say whether the new system helped. We agree the numbers with you before the build and compare them after launch.

What does an engagement cost?

A one-time setup fee plus a monthly fee for hosting, maintenance and updates, both scoped to your process before work starts. We don't bill by the hour and don't publish a price list.

What if the result is smaller than expected?

We show you the numbers as they are. The controlled rollout is there so we see this early, while the system can still be adjusted.

Who is behind opencream.ai?

opencream.ai is run by Opencream SASU from Chantilly, France, with a base in Düsseldorf for clients in Germany. The team works in English, German, French and Spanish.

About opencream.ai

Want to see how this works on your process?

Start with the discovery call. We look at where the hours go and tell you honestly whether AI is the answer.

Pick a time that works for you. 15 minutes. No pitch.