How our Artificial Intelligence projects run
Every engagement follows the same six-step sequence. The structure exists so you always know where the project stands, what it will cost next and when you can expect to see results.
The six steps
Below is the path from first conversation to a live, monitored system. Timelines are typical ranges; your project may be shorter or longer depending on data readiness.
Discovery call
A 30-minute video call where you describe the business problem and we ask pointed questions. What decision does this model need to support? What data do you already collect? Who will consume the output? We take notes, you talk. No slide decks, no sales pitch. By the end we both know whether AI is the right tool or whether a simpler approach would serve you better.
Data audit
You share a sample of the data you have. We review its volume, completeness, label quality and any privacy constraints. The audit report is a short document, usually three to five pages, that lists what is usable, what needs cleaning and what gaps exist. If the data is not sufficient to build a reliable model, we say so here and suggest what you would need to collect before starting.
Scoping and proposal
Based on the audit we write a detailed scope: the model type, the target metric (for example, mean absolute error below 4% on held-out test data), the delivery format, the infrastructure requirements and the timeline. The proposal includes a fixed price for the prototype phase. You sign off or ask for changes before any engineering begins.
Prototype build
Our engineers clean the data, engineer features, train candidate models and evaluate them against the agreed metric. We use version-controlled experiment tracking so every run is reproducible. You receive a weekly progress email with current metric values and a brief note on what we tried and what we plan next. At the end of this phase you get a working prototype you can test with real inputs through a simple web interface or API.
Review and iteration
You and your team test the prototype. We collect feedback: where does it fail? Which edge cases matter most? We then run a focused round of improvements, re-evaluate and present the updated results. If the prototype does not meet the agreed metric after this round, you pay nothing for the prototype phase.
Production deployment and monitoring
We package the model into a containerised service, deploy it to your cloud environment (or ours, if you prefer managed hosting) and wire it into your existing systems. Monitoring dashboards track prediction volume, latency and data drift. We provide a 90-day support window after launch, included in the deployment fee, during which we fix any bugs and retrain the model once if incoming data patterns shift.
What you get vs. what you pay
Transparency matters. Here is a plain breakdown of deliverables and fee structures at each stage.
| Phase | Deliverable | Fee model |
|---|---|---|
| Discovery call | Written summary of the problem and initial feasibility assessment | Free |
| Data audit | Audit report (PDF) with data quality scores and gap analysis | Fixed fee, quoted before start |
| Scoping | Detailed proposal with metric targets, timeline and price | Included in audit fee |
| Prototype | Working model, API or web interface, experiment logs | Fixed fee; refundable if metric not met |
| Iteration | Improved model, updated evaluation report | Included in prototype fee |
| Deployment | Containerised service, monitoring dashboard, 90-day support | Fixed fee, quoted at scoping |
All prices are quoted in GBP excluding VAT. We do not bill by the hour for any standard phase.
Ready to start?
Book a free discovery call and find out within 30 minutes whether AI can solve your specific problem.
Get in touch