AI development

Start with a task that happens often, has a clear right answer most of the time, and costs real effort today, such as classifying requests or extracting data from documents. Avoid starting with the hardest or most visible problem. We help you list candidates, score them on value, data readiness and risk, and pick one that can show results on its own. Our AI readiness assessment does this in a structured way.

The main drivers are the state of your data, the number of systems to integrate, how accurate the system must be, and how much human review the process needs. Running costs matter too: model usage, hosting and monitoring. We estimate after a short discovery, separate build cost from running cost, and show how running cost changes with volume so there are no surprises later.

A prototype on your data comes first, so you can judge results before committing to the full build. The time to production depends on data access, security reviews, the integrations involved and how much the evaluation shows needs fixing. Access to data and approvals is often the longest step. We plan these early and give you a timeline after discovery.

Your data stays in environments you approve. We prefer to run models in your own cloud account or through enterprise model services that do not train on your data. We agree what data the system may see, mask personal data where the task allows, and log access. Data used for evaluation is stored under the same rules as production data.

We agree an evaluation set with your team before building: real examples with the answers your experts expect. Every version of the system is scored against it, and reviewers check a sample of the harder cases. You decide the threshold for release. In production, we keep measuring and route uncertain cases to a person rather than guessing.

We work with commercial models such as OpenAI GPT models, Anthropic Claude and Google Gemini, with open models such as Meta Llama and Mistral, and with classic machine learning where it fits better. We choose per task on quality, cost, latency and where the model may run. We design so the model can be changed later. Naming a vendor does not mean a partnership.

Intellectual property in the work we build for you is yours from the first commit — no assignment step at the end. That includes the code, prompts, evaluation sets and fine-tuned models produced for you, and your data stays yours. Any XSIS tools that existed before stay ours and are licensed to you. Third-party models and services remain under their own licences, which we list for you so you know what each component allows.

AI systems need ongoing attention: data changes, models are updated and new cases appear. We can monitor quality and cost, review flagged cases, update prompts or models, and rerun the evaluation set before every change. This can be part of a support agreement or handed to your own team with the runbook and training.

Only if you agree. We tell you which tools we would use, how they handle code and data, and whether anything leaves your environment. If you prefer, we work without them. Code written with AI assistance goes through the same review and testing as any other code, and the engineer who submits it is responsible for it.