AI solutions
AI that works with your data and your rules
We build AI into the way you actually work. We start from a clear task, check the quality and protect your data, and a person approves the actions that matter.
What we build
Customer assistants
Answer frequent questions from verified sources and escalate to a person when unsure.
RAG over your documents
Search and answers across internal documents, catalogues and policies, with sources cited.
Document analysis
Data extraction from invoices, contracts and specifications, checked before it is saved.
Classification and routing
Automatic routing of inquiries, emails and requests by topic and urgency.
Content generation with editing
Drafts of product descriptions, replies and summaries that a person reviews.
Reporting assistants
Plain-language explanations and summaries of sales and operations data.
AI Automation Advisor
Describe a repetitive manual task and get a structured assessment — opportunities, integrations, risks and phases. No sign-up and no personal data.
Principles we apply in every AI project
A specific task and success criteria
We define what the solution must do and how quality will be measured before we start.
Minimal data
Only the necessary data is sent to the model, and personal data is removed wherever possible.
Human approval
Actions that send messages or change products, prices or orders require explicit approval.
Logging and traceability
Every proposed and approved action is recorded — who, when and why.
Provider independence
We build the solution so that the model and provider (such as Anthropic Claude) can be replaced.
Evaluation and monitoring
Test scenarios before deployment, and monitoring of quality and cost afterwards.
Approval policy for AI actions
AI agents can propose actions, but they do not carry them out on their own when the action has a real effect. This is how we classify actions:
| Action | Risk | Control |
|---|---|---|
| Reading public contentAllowed automatically | Low | Allowed automatically |
| Drafts and classificationAllowed; the result is reviewed | Low | Allowed; the result is reviewed |
| Sending messages to customersExplicit human approval | High | Explicit human approval |
| Changing productsExplicit human approval | High | Explicit human approval |
| Changing prices or ordersExplicit approval and audit log | Critical | Explicit approval and audit log |
| Access to sensitive dataExplicit approval and audit log | Critical | Explicit approval and audit log |
AI project lifecycle
- 01
Discovery
Choosing a process with clear value and acceptable risk.
- 02
Prototype
A quick check with real examples from your data.
- 03
Evaluation
A test set and criteria for quality, security and cost.
- 04
Deployment
Integration into your systems, access, logging and approvals.
- 05
Monitoring
Tracking quality, cost and user feedback.
AI services
We build agents, automations and custom applications, connect them to your systems and support them after launch.
- AI strategy & readinessA structured assessment of your store, data and processes that ends with a prioritised improvement plan.
- AI agent developmentAgents that carry out multi-step tasks through your systems — with clear limits, defined tools and human approval for actions that matter.
- AI workflow automationInquiry classification, CRM workflows, document processing and system integrations — with human approval where needed.
- Custom AI engineeringAI assistants, product description generation, internal knowledge bases and RAG solutions built into your workflows.
- AI integrations & data systemsREST API and webhook integrations, connecting Shopify and WooCommerce with CRM and ERP, cloud deployment, monitoring and protection.
- AI Managed Services & optimisationMonitoring, maintenance, quality evaluation, security and cost optimisation for AI agents and automations that are already in production.
After launch: AI Managed Services
An AI agent can start making mistakes even while the website and server look fine. After launch we keep an eye on quality, security and cost.
When AI is not the right answer
- When a simple automation or rule solves the task more reliably.
- When there is not enough good-quality data to work with.
- When a mistake would have serious consequences and cannot be checked by a person.
- When the expected benefit does not justify the cost of maintaining it.
Discuss this service with us
Describe your task and we will suggest a suitable scope and engagement model.