Search and answers from your own documents
An AI knowledge base that answers staff or customer questions using your manuals and procedures, with references to the source material.
From idea to proof of concept
From idea to proof of concept
It is easy to keep talking about possibilities. I turn your idea into the smallest useful working version, built with real data, so you can make the next decision on evidence.
Sound familiar?
Why ideas get stuck
- The idea has been sitting on the shelf for months, but nobody can say whether it will work.
- Suppliers want a fully detailed specification before they can begin.
- AI sounds promising, but you want to see it work before making a major investment.
- The discussion stays abstract while there is nothing to try.
A working version gives everyone something concrete to discuss.
The approach
What does a proof of concept involve?
A proof of concept is the smallest working version of a digital idea, built with real data to establish whether it is valuable, feasible and worth pursuing. The software actually performs the task you want to test, giving you evidence for your next decision.
The technology follows the idea. I use AI where it adds value: for example, a search and answer system based on retrieval-augmented generation (RAG), which answers questions using your own documents and cites its sources. Where systems need to work together, I connect them using platforms such as n8n, which can run in your own environment. Sometimes, sound web technology is all that is needed.
Every project starts with one question: what does it need to prove? The scope is usually small enough to make a quick decision, but substantial enough to learn something useful. See my four-step approach.
Examples
What a first working version might look like
An internal tool or dashboard
A small working tool for a task or overview that currently relies on spreadsheets and scattered emails.
Connecting systems
A trial that shows how data can move automatically from one system to another, without rekeying it.
A small web application or proof of concept
A first working version of a digital service, ready to share with users, customers or decision-makers.
That is how my own work began too. My applications — Clarido, ContentInzicht, Agima CMS and EMPF Preview — each started as a small working version and are now in production.
A starting point
Explore the possibilities first?
No specific idea yet, but a sense that more is possible? We can begin with a short AI discovery phase. In one or a few working sessions, I review your processes and information flows with you and your team: where work is repetitive, where time is lost, and where people struggle to find the information they need.
For each opportunity, I assess what AI or another technology could realistically achieve, what it would cost and what it could deliver. You receive a prioritised list and one or more ideas defined clearly enough to build and test. I also explain which ideas are best left alone, helping you avoid expensive experiments.
The outcome
What you receive
A working application
Software you can run and demonstrate yourself, whether to colleagues or customers.
A technical assessment
What a full implementation would involve: systems, data, integrations and risks.
Lessons learnt
An honest account of what the test established and what remains unproven.
A recommendation: develop, adapt or stop
Clear advice on the next step, including when it makes sense to stop.
Frequently asked questions
About proof of concept projects
How long does a project take, and what does it cost?
We discuss the options and timing before we begin. You receive a fixed proposal with clear agreements on scope, cost and the intended outcome.
Will the result be ready for production?
Not at this stage. A proof of concept establishes whether an idea works; making it ready for production is a separate decision. I can help with that next step, or you can use the technical findings to work with another developer.
What if the idea turns out not to work?
That is a useful result too. You find out before making a major investment, and the lessons learnt help identify more promising directions.
What if I don’t have a specific idea yet?
We can start with a short AI discovery phase. We review your processes and information flows, prioritise opportunities by value and feasibility, and build once there is something concrete to test.
Verifiable evidence
What this work does and does not establish
- Problem
- A digital idea remains abstract until someone tests it with real data and working software.
- Approach
- The question is narrowed to the smallest version capable of testing one decisive assumption. That version is then built and used.
- Status
- Available as a focused project with the scope, costs and intended evidence agreed in advance.
- Demonstrable result
- A working application, a technical assessment, documented lessons and a recommendation to develop, adapt or stop.
- Limitations
- A proof of concept is not automatically ready for production. Scalability, maintenance, security and integrations may require further work.
- Last reviewed
Contact
Which idea would you like to test?
Tell me a little about what you want to test. I’ll reply personally, and we’ll discuss the options and timing before we begin.
Prefer to contact me directly? Call +31 6 1400 2312 or email mwd@mahieu.nl.