Frequently asked questions about Artificial Intelligence

Most engagements run between 8 and 16 weeks. The first two weeks are a discovery phase where we audit your data, define success metrics and agree on a fixed-price quote. After that, model development usually takes four to eight weeks depending on complexity. The final two weeks cover integration testing, documentation and a handover workshop for your team.

Smaller tasks like a single classification model on clean, labelled data can ship in as little as four weeks. Large multi-model programmes with custom data pipelines have taken up to six months.

Not necessarily. We can work inside your own cloud environment or on-premise servers, so the data never leaves your control. For early-stage prototyping we often use anonymised or synthetic datasets that mimic the statistical properties of the real thing without exposing personal records.

If the project requires us to access identifiable data, we sign a data-processing agreement that specifies retention limits, encryption standards and breach-notification timelines. All our engineers hold current security clearances appropriate for handling sensitive UK data.

Our strongest verticals are retail, logistics, healthcare, insurance and legal services. In retail we have built demand-forecasting models for a chain of 140 stores. In logistics we deployed a route-optimisation engine that reduced fuel spend by 12% over a year. Healthcare projects have focused on clinical-note summarisation and triage prioritisation.

We have also completed projects in manufacturing (defect detection on production lines) and energy (predictive maintenance for wind turbines). If your sector is not listed here, get in touch anyway. The underlying techniques transfer well; what matters most is having structured data and a clear business question.

We use fixed-price contracts for defined deliverables. After the discovery phase we present a detailed scope document listing every model, API endpoint and integration point, along with the total cost. You pay in three instalments: 30% at kick-off, 40% at the prototype review, and 30% at final delivery.

For ongoing model monitoring and retraining we offer monthly retainer plans starting at £1,200 per month. The retainer covers performance dashboards, drift detection alerts and one retraining cycle per quarter.

Every project contract includes a performance threshold agreed during discovery. If the delivered model falls short of that threshold on the holdout test set, we continue iterating at no extra cost for up to four additional weeks. In the rare case where the data simply cannot support the target, we provide a written analysis explaining the limiting factors and recommend next steps, such as collecting more training examples or adjusting the business question.

Yes. We deliver models as containerised REST APIs that slot into any system capable of making HTTP requests. We have integrated with Salesforce, SAP, Microsoft Dynamics, custom Django back-ends and legacy COBOL systems via middleware. During discovery we map your architecture and agree on the integration approach before quoting.

Every model we build goes through a fairness audit before deployment. We test for disparate impact across protected characteristics using statistical parity and equalised-odds metrics. If bias is detected, we apply mitigation techniques such as re-sampling, re-weighting or post-processing calibration. The audit results are included in the delivery documentation, and we recommend repeating the audit after each retraining cycle.