Most engagements fall between £8,000 and £45,000. A simple automation, such as classifying incoming emails or extracting data from PDFs, typically sits at the lower end. Predictive models that need custom feature engineering and integration with your ERP lean towards the higher end. We always provide a fixed quote after the discovery call, so there are no surprises mid-project. If your budget is below £8,000 we can sometimes suggest a smaller proof-of-concept to test feasibility before committing to a full build.
Four to ten weeks from signed agreement to live deployment. The biggest variable is data readiness. If your data is clean, labelled, and accessible through an API or database export, we can move fast. When data needs significant cleaning or when multiple source systems must be merged, add two to three weeks. We send a written timeline with milestones before work begins, and we flag delays the moment they appear rather than waiting until the deadline.
Yes. Around 40% of our clients have fewer than 50 employees. Smaller firms often benefit the most because a single automation can free up a significant fraction of staff time. We have worked with a ten-person accountancy practice in Cardiff that automated their bank-statement reconciliation, saving roughly twelve hours a week. The key requirement is not company size but whether you have enough data to make a model useful. We will tell you honestly during the discovery call if the numbers are too thin.
It depends on the project. For a demand-forecasting model we typically need at least 18 months of historical transaction records: dates, quantities, product identifiers, and any relevant context like promotions or seasonality flags. For a document-classification tool we need a representative sample of at least 500 documents across the categories you want to distinguish. We accept CSV, Excel, JSON, database dumps, or API access. During the data audit phase we run quality checks and let you know if anything is missing or inconsistent.
All project data is stored on UK-based cloud infrastructure. We use encrypted storage at rest and in transit. Access is limited to the engineers assigned to your project, and we delete your data within 30 days of project completion unless you ask us to retain it for ongoing support. If you have specific compliance requirements, such as ISO 27001 or NHS Data Security and Protection Toolkit, let us know early and we will map our processes to your framework.
Almost always. We have integrated models with Salesforce, Microsoft Dynamics, SAP Business One, Xero, and various bespoke systems. If your platform has an API or supports webhook triggers, we can connect to it. For systems without an API we sometimes build a lightweight middleware layer. We discuss integration requirements during the discovery call so there are no surprises later.
Every project includes 90 days of post-launch support at no extra cost. During that period we fix bugs, answer questions, and monitor model performance. After 90 days you can either manage the system yourself, using the documentation we provide, or sign a monthly support agreement starting at £600 per month. Support agreements cover model retraining, performance dashboards, and priority response within four business hours.
Yes. Once the final invoice is paid, you own the trained model weights, the training pipeline code, and all documentation. We retain the right to reuse general-purpose techniques and libraries, but anything specific to your data and business logic belongs to you. This is spelled out in our standard project agreement, which you receive before any work starts.
We define success metrics together before the build phase begins, for example a minimum accuracy of 92% on held-out test data, or a reduction in processing time of at least 60%. If the model falls short of the agreed metric after two rounds of iteration, you pay only for the work completed to that point. We have had to invoke this clause twice in four years, both times because the underlying data turned out to be too noisy. In those cases we provided a written report explaining why and what data improvements would make a future attempt viable.
A few practical steps that make the process smoother for everyone.
Before the first call, check whether you can export the data you think is relevant. Open a sample in a spreadsheet. Are there blank rows? Mixed date formats? Duplicate entries? Knowing the state of your data upfront saves time during the audit phase.
Resist the temptation to automate everything at once. Choose the process that costs the most time or money, prove the value there, then expand. A focused project delivers faster results and builds internal confidence in the technology.
The staff who currently handle the task manually know its quirks better than anyone. Include them in the discovery call. Their input helps us design a solution that fits real workflows rather than an idealised version of them.
Decide what "good enough" looks like before the build starts. Is it 90% accuracy? A 50% reduction in processing time? A specific cost saving per month? A concrete target keeps the project focused and makes the outcome measurable.
Drop us a message and we will get back to you within one working day. No sales pitch, just a straight conversation about whether AI fits your situation.
Contact us