Frequently asked questions about Artificial Intelligence services

Most engagements fall between £15,000 and £80,000. A straightforward predictive model built on clean tabular data sits at the lower end. A multi-stage document-intelligence pipeline with custom training, integration into your ERP and a monitoring dashboard sits at the upper end.

We quote a fixed price after a paid discovery phase (usually £2,500 to £4,000) so there are no surprises. If the discovery phase reveals that AI is not the right approach, we tell you and you only pay for discovery.

A focused predictive-analytics project usually takes four to eight weeks from data handover to a production-ready model. Larger document-intelligence systems may take three to four months because they involve more integration work and iterative labelling rounds.

The single biggest variable is data readiness. If your data lives in well-structured databases with consistent naming, we move fast. If it is spread across dozens of spreadsheets with inconsistent formats, we need extra time for cleaning and reconciliation.

It depends on the problem. For tabular forecasting (e.g. predicting next month's sales by SKU), a few thousand rows of clean historical data is often enough. For natural language tasks like classifying support tickets by topic, we typically need at least 5,000 labelled examples to train a reliable classifier.

If you have less data than that, we can sometimes use transfer learning or pre-trained language models to close the gap, but we will be honest about the trade-offs in accuracy.

Most data is messy. We budget time for cleaning and transformation in every project plan. Common issues we deal with include missing values, duplicate records, inconsistent date formats and free-text fields that should have been dropdowns.

During the discovery phase we run a data-quality audit and give you a candid report. Sometimes fixing the data collection process first saves more money than any model could.

We include a 30-day support window after deployment at no extra cost. During that period we monitor model performance, fix bugs and adjust thresholds based on real-world feedback.

After that, you can either maintain the system yourself (we provide full documentation and a handover session) or sign a monthly support contract. Support contracts start at £1,200 per month and cover monitoring, retraining on new data and priority bug fixes.

Probably, yes. This is called model drift and it happens when the patterns in your data change. A demand-forecasting model trained on 2022 data may become less accurate by mid-2024 because customer behaviour, product mix or market conditions have shifted.

We build automated drift-detection alerts into every deployment. When accuracy drops below a threshold you define, the system flags it so you (or we, if you have a support contract) can retrain the model on fresh data.

You do. The contract assigns all IP created during the engagement to you. That includes model weights, training scripts, data pipelines, documentation and any custom libraries we build for your project. We retain no copies after the handover unless you explicitly ask us to for support purposes.