General questions
The discovery sprint is fixed-price: £3,500 to £5,500 depending on the number of data sources we need to review. After the sprint we quote the build phase as a fixed fee, typically between £15,000 and £40,000 for a single use case. Complex multi-model systems can go higher, but we scope those as separate projects so you can approve each one individually.
Ongoing maintenance (retraining, monitoring, drift alerts) is billed quarterly at £1,200 to £2,800. There is no long-term contract; you can cancel at the end of any quarter.
From first meeting to live deployment, most projects run 10 to 16 weeks. The discovery sprint takes two weeks. Data preparation adds two to four weeks. Model development takes three to six weeks. Integration and deployment take another two to three weeks. We share a detailed Gantt chart after the discovery sprint so you can plan around it.
It depends on the task. For predictive analytics, we generally need 12 to 24 months of historical data with at least a few thousand records. Demand forecasting for a product catalogue of 500 SKUs, for example, works well with two years of weekly sales figures.
Document automation can start with as few as 200 sample documents, provided they cover the range of layouts and edge cases you encounter. Computer vision projects using transfer learning need a few hundred labelled images per class. We assess data sufficiency during the discovery sprint and tell you honestly whether what you have is enough.
You pay the sprint fee and receive the full feasibility report. The report will explain why AI is not viable for the problem as currently framed, and it will suggest alternatives. Sometimes the answer is "collect six more months of data and revisit." Other times a simpler rules-based system would solve the problem at a fraction of the cost. We have recommended that path to three clients, and two of them came back later for an AI project once they had the data to support it.
Your data stays on your infrastructure or on a private cloud instance that only your team and ours can access. We do not use client data to train models for other clients. During the project we access data through a secure VPN or SSH tunnel. After handover we revoke our own access unless you have signed a maintenance contract, in which case we retain read-only access to the production database for monitoring purposes.
We have deployed models into environments running Python, Node.js, .NET and Java backends. On the data side, we work with PostgreSQL, MySQL, SQL Server, BigQuery, Snowflake and several common ERP systems including SAP and Microsoft Dynamics. If your stack is unusual, raise it in the initial conversation and we will confirm compatibility before you commit to anything.
You do. The statement of work assigns all intellectual property in the custom model, training pipeline and deployment code to you upon final payment. We retain the right to reuse general-purpose utilities and open-source components we brought into the project, but the trained model weights, your data, and any custom code are yours.
It probably will, and that is normal. Customer behaviour shifts, product lines change, market conditions move. Our monitoring dashboard tracks prediction accuracy against ground truth as it becomes available. When accuracy drops below a threshold we agree on during deployment, the system triggers an alert. If you are on a maintenance contract, we retrain the model using recent data. If not, you can retrain it yourself using the documentation and scripts we provide at handover.
Every project includes a half-day training session as part of the handover phase. We walk your team through the model architecture, the monitoring dashboard, the retraining process, and common troubleshooting steps. If you want more in-depth training on machine learning concepts for your data team, we offer a separate two-day workshop priced at £2,400 for up to eight participants.
Fill in the contact form on our home page or call us at +44 1666 675017. Describe the problem you want to solve in a few sentences. We will reply within one working day with initial thoughts. If it looks like a fit, we schedule a 30-minute video call to discuss scope, data availability and timeline. No obligation at that stage.
Still have questions?
We are happy to talk through your specific situation. Reach out by phone or email and we will get back to you within one working day.
Where to find us
Our office is in Scotland. Most client work happens remotely via video calls and secure data connections, but you are welcome to visit if you prefer face-to-face meetings.
United Kingdom, Scotland, Old Nikolaus, DY9 8RM, 4 Forrest Hill