Workflows & data · Machine learning
Machine learning where a simple rule is not enough
Machine learning finds patterns in your past data to predict or sort new cases: which leads are likely to buy, how much stock you will need, which customers may leave, or which category a ticket belongs to. We check whether your data can support it, build and test the model, and put it to work inside your existing workflows.
What is included
- A check of your data: what it can predict and how reliably
- A simple baseline first, so you can see what the model really adds
- Model training and testing on data it has not seen
- Deployment as an API or inside n8n workflows
- Monitoring for drift, with a plan for retraining
Machine learning or an LLM?
For predictions from numbers and history, such as forecasts and scores, classic machine learning is usually more accurate and far cheaper. For reading and writing text, LLMs are better. Many projects use both.
How much data do we need?
Enough past examples of the thing you want to predict, with the outcome recorded. We look at your data first and tell you honestly if it is not enough yet, and what to start collecting.
Questions about machine learning
Will we understand why the model decided something?+
We prefer models that can show which factors mattered most, and we report that with the results.
What happens when our business changes?+
We monitor accuracy over time and retrain when the patterns change.