Artificial Intelligence FAQ

Most projects run between six and fourteen weeks from kickoff to production deployment. The first two weeks are always a discovery sprint: we audit your data, define success metrics and build a minimal proof of concept. If the proof of concept fails to meet the agreed threshold, you can walk away having spent only the discovery fee. Larger engagements with multiple models or complex integrations can stretch to five or six months, but we break those into two-week iterations so you see progress continuously.

It depends entirely on the problem. For a demand-forecasting model we typically ask for at least twelve months of historical transaction data, including timestamps, product identifiers and quantities. For an NLP classifier we need a representative sample of the documents you want to categorise, ideally a few thousand examples with labels. We can ingest CSV files, connect directly to SQL databases, or pull data through REST APIs. During the discovery sprint we help you identify gaps and suggest ways to fill them before model training begins.

Absolutely. Some of our most effective projects have been with teams of ten to fifty people. Smaller organisations often have clearer ownership of decisions, which means a model goes from prototype to daily use much faster. We have a lightweight engagement package specifically for businesses that need a single focused model rather than a full platform build.

The two-week discovery sprint is a fixed fee of £3,500. After that, projects are priced on a time-and-materials basis with a weekly rate that depends on team size. A typical single-model engagement with one engineer and one data scientist runs between £8,000 and £25,000 in total. We provide a detailed estimate at the end of the discovery sprint so there are no surprises. If you have a strict budget, tell us upfront and we will scope the work to fit.

Yes. Machine learning models degrade over time as the data they were trained on drifts away from current reality. We offer a monthly support plan that includes monitoring model accuracy, retraining on fresh data and adjusting feature pipelines. The plan starts at £800 per month for a single model. Clients who prefer to handle maintenance in-house receive a full handover package with documentation, retraining scripts and a runbook.

We deploy on AWS, Google Cloud and Microsoft Azure. If you already have infrastructure on one of those providers, we build on top of it rather than asking you to migrate. For clients with strict data-residency requirements we can also deploy models on-premises or on private cloud environments. Edge deployment on NVIDIA Jetson or similar hardware is available for computer vision workloads that need low-latency inference at the point of capture.

We sign a mutual NDA before any data changes hands. All data is encrypted in transit (TLS 1.3) and at rest (AES-256). Our engineers access client data only through audited, role-based accounts on the client's own cloud tenant whenever possible. We never use client data to train models for other clients. After project completion, all copies of your data on our systems are deleted within 30 days unless you request otherwise in writing.

Explainability is part of every project. We use SHAP values, LIME and attention-weight visualisation depending on the model type. Each delivery includes a report that shows which features drive predictions and where the model is least confident. For regulated industries such as finance or healthcare, we can produce documentation suitable for compliance review.