Frequently asked questions about Artificial Intelligence

Most projects run 8 to 14 weeks from the first discovery call to a deployed, monitored model. The breakdown looks roughly like this: two weeks for the discovery sprint, six to ten weeks for iterative model development, and one to two weeks for integration testing and deployment. Smaller tasks, like building a single classification endpoint on clean data, can be done in four weeks. Larger programmes with multiple models and complex data pipelines have taken up to six months, though that is unusual.

Our discovery sprints start at £4,500. Full model-development projects range from £15,000 to £80,000. The price depends on how many data sources need connecting, how much data cleaning is required, and how the model will be deployed. We always provide a fixed-price quote after the discovery sprint, so you know the total cost before committing to the build phase. There are no hourly-rate surprises.

It depends on the task. A binary classifier (for example, "will this customer churn or not?") can produce useful results with 2,000 to 5,000 labelled examples. Computer vision models for quality control often need 500 to 1,000 annotated images per defect class, though we can augment smaller sets with synthetic data. Custom NLP models benefit from tens of thousands of domain-specific documents. During the discovery sprint, we assess whether your data is sufficient and, if it is not, recommend practical ways to collect more or to use transfer learning from pre-trained models.

In the 40-plus projects we have delivered, not one client has reduced headcount as a direct result. What typically happens is that repetitive sub-tasks get automated, and the people who used to do those tasks shift to higher-value work. A warehouse team that spent four hours a day on manual stock counts, for instance, now uses a computer vision system and spends that time on supplier negotiations and layout improvements. The model handles the counting; the team handles the thinking.

Most of our experience is in logistics, retail and healthcare. We have also completed projects for financial services firms, a construction materials supplier and a regional energy provider. The common thread is mid-sized organisations with enough data to train useful models but not enough in-house data science capacity to build and maintain them. If your industry is new to us, the discovery sprint is where we assess feasibility honestly.

Your data stays within the infrastructure you choose. For cloud-based projects, we use your own cloud tenancy (AWS, Azure or GCP) so that data never leaves your account. For on-premise deployments, we work on your servers. During development, if we need to pull sample data to our own machines for rapid prototyping, we sign a data processing agreement first and delete all copies when the project ends. We are registered with the ICO and comply with UK GDPR.

Models degrade over time as the real world changes. We offer 12-month support contracts that include monthly performance reviews, automated data-drift alerts and model retraining when accuracy drops below an agreed threshold. About 80% of our clients take this option. If you prefer to handle maintenance in-house, we provide full documentation, training sessions for your team and the complete source code.

Yes. We package models as REST APIs, containerised microservices or embedded Python modules, depending on what fits your architecture. We have integrated with ERP systems like SAP and NetSuite, CRM platforms including Salesforce and HubSpot, and various custom internal tools built in Java, .NET and Node.js. During the discovery sprint, one of our engineers maps your tech stack so we can plan the integration path early.