Artificial Intelligence pricing

Not necessarily. If you already have clean, well-documented data and a clear use case, we can skip straight to the build. During our initial call we will tell you honestly whether an audit would save time later. Many clients with messy or siloed data find the audit pays for itself by preventing wasted engineering hours.

The build-and-deploy price covers a single model solving one defined problem. If you need a second model, for example a demand forecaster plus a churn predictor, we scope each one separately and often apply a 15% discount on the second engagement because the data pipeline work is already done.

Our prices cover engineering time, not third-party cloud bills. We deploy to your own AWS, Azure or GCP account so you control the infrastructure directly. For a typical NLP classifier handling 10,000 requests per day, expect cloud costs between £80 and £150 per month. We size the infrastructure during the build phase so there are no surprises.

Yes, with 30 days written notice. You keep everything we have built: the model, the API code, the monitoring scripts and all documentation. We will also run a handover session with your internal team so they can maintain the system independently.

For the build-and-deploy package we split the fee into three instalments: 40% at kick-off, 30% at model-validation milestone, and 30% on deployment. The audit is billed in full before work starts because it is a short, discrete piece of work.

We define a minimum accuracy threshold in the statement of work. If the model falls short after two rounds of tuning, we refund the final instalment. This has happened once in 14 projects, and the root cause was insufficient labelled training data, which the audit would have flagged.