Three steps, each one small enough to stop after
The most expensive mistake in a knowledge-graph project is building the graph before checking that a graph is the right tool. So the work is split into steps, and each step ends with a written answer and the option to stop there.
KG Review
2-3 daysFor teams that are considering a knowledge graph, or already have one that is not delivering. Two versions:
- No graph yet: I look at the problem, the data and the questions people need answered, and compare a graph against the simpler options (vector search, SQL, better search). You get a recommendation: build a graph, don't, or what to do instead.
- Graph in place: I review the ontology, data and extraction quality, how the graph is tested, and what it costs to run. You get a written report with problems ranked by impact.
Deliverable: written report and a 1-hour walkthrough call.
From €1,500 · fixed price, agreed after a short call
Ask about a Review →Pilot
about 6 weeksFor when the review says a graph is worth building.
- Scoping (week 1): who will use it, which decisions it supports, which sources go in, and 5-10 control questions the graph must answer. These are the success criteria.
- Build (weeks 2-5): a working Neo4j graph from your sources, queries or a chat assistant on top, and a trail from every answer back to its source.
- Result (week 6): the control questions run against the graph, with the results in writing, and a recommendation: scale it, narrow it, or stop.
Deliverable: working graph, code and data model, test results, recommendation.
From €6,500 · fixed price, set after scoping
Infrastructure used during the pilot (for example Neo4j AuraDB or language-model API usage) is billed at cost.
Check whether your use case is ready →Implementation
scoped after the pilotFor taking a successful pilot to production: scheduled ingestion and updates, a test suite, integration with your systems, documentation and handover to your team.
Scope and price are set from what the pilot showed, not guessed up front.
Talk about taking a graph to production →Prices are in euro and exclude VAT. Business clients in other EU countries are invoiced under the reverse-charge mechanism.
WHAT I DON'T DO
- Promise results before scoping. The control questions define what "working" means.
- Large-team, multi-year programmes. I work alone and take on one project at a time.
- Keep you dependent on me: code, model and documentation are yours.
HOW IT STARTS
- Fill in the pilot readiness checklist (3 minutes) or send a short message.
- 30-minute call about your problem and data.
- You get a written proposal for a Review with a fixed price.
FAQ
Can you work with confidential data?
Yes. NDA before we see data; work can run in your environment.
Do you work remotely?
Yes, from Wrocław, Poland (CET), in English or Polish. One part often goes better in a room: the scoping workshop at the start of a pilot, where your domain experts and I write the control questions together. When that helps, I come to you for the day; travel for on-site days is billed at cost.
Do we need Neo4j already?
No. If a graph is the right tool, the pilot sets it up; Neo4j AuraDB or self-hosted.
What if the answer is "don't build a graph"?
Then the Review has saved you a pilot. That outcome is part of the service.
Not sure which step fits?
Tell me what you are trying to answer and what data you have. We start with a 30-minute call, and you get a written proposal after it.