Andre Geimer DE Proof of Concept
An AI prototype with your real data.
I build a prototype for the strongest use case. It runs on your data and is tested by the people who will use it later. Then you know whether it pays off and decide on the next step.
Example
A prototype, not a slide deck.
Here a prototype reads supplier quotes and matches them against master data. Something like this can be built in a few weeks and tested with real documents.
Who it's for
- You have a specific use case and want to know whether it holds up.
- An earlier AI project fizzled out, and this time the result should be measurable.
- Management wants to see results before investing more.
What you get
- A working prototype, not slides
- Concept and architecture that also carry the next stage
- Measurements from tests with real data and real users
- A clear recommendation: scale up, adjust or stop
How it works
- 01Scoping
We set the goal, success criteria and data. What does the prototype have to show for us to go on?
- 02Build
I develop in short steps, with generative AI, RAG or automation, depending on the case. You see progress regularly.
- 03Test
The future users work with the prototype. We measure against the criteria set at the start.
- 04Decision
Result, effort for running it and my recommendation on how to proceed.
Typical use cases
A few examples of cases that work well as a proof of concept:
- Reading quotes and supplier documents and passing them to the ERP
- Capturing incoming invoices and pre-assigning accounts
- Making knowledge from manuals, tickets and file shares searchable by chat (RAG)
- Pre-sorting customer enquiries and preparing replies
Duration and cost
We agree the scope in the intro call. Hosting is in the EU by default, and we settle data protection and processing agreements before the first test.
Questions about the proof of concept
What if the prototype doesn't work?
Then that's a result too, and a cheap one. We set the success criteria up front. If they're missed, we stop instead of putting in more money.
Which AI models do you work with?
The one that fits the case, not a fixed vendor. Hosting is in the EU by default.
What happens to our data?
It stays in the EU. We settle data flows and processing agreements before real data is involved.
Who runs the solution later?
We work that out during the proof of concept. I can take on the next stage, enable your team or bring in people from my expert network for larger scopes.