Services / AI & LLM Features
Add AI to your product in a way that's useful, safe and affordable.
Chat, search across your own documents, and automation of repetitive work. We build it into your existing product and plan for cost per request, data privacy and wrong answers from the start.
OpenAI · AWS Bedrock · RAG · Automation
$ Q: What's our refund policy for annual plans?● search 3 matching docs found● answer Full refund within 30 days…✓ sources shown billing-policy.pdf, p.2✓ data stays in eu-north-1· cost logged per request
Sound familiar?
Signs you need this
Customers ask questions your docs already answer
Your team copies data between tools by hand every day
You've tried a ChatGPT prototype but don't trust it with customer data
AI costs are unclear and you want a number before you build
The change
Before and after
| Before | After |
|---|---|
| Support answers the same questions daily | Users get sourced answers instantly |
| A demo nobody trusts with real data | EU-hosted, logged and access-controlled |
| Unknown AI costs | A cost per request agreed before building |
What you get
Deliverables
Use-case check
Which AI feature is worth building, with expected cost per month.
Document search and chat
Answers from your own content, with sources shown to users.
Workflow automation
Classify, extract, summarise and route data between your tools.
Data privacy setup
Models hosted in the EU where needed, no training on your data, access controls.
Guardrails
Limits, fallbacks and logging for when the model is wrong.
Cost tracking
Usage and spend per feature, with alerts.
How it runs
Four steps
- ScopeOne clear use case, success measure and cost target.
- PrototypeWorking version on your real data within days.
- HardenSecurity, limits, monitoring and tests.
- LaunchShipped to users, then measured.
What we need from you
- The documents or data the feature should use
- Examples of the questions or tasks to handle
- Someone to test answers in week one
Proof
Done before, in production
- LLM integration with OpenAI and AWS Bedrock on current work
- Production chatbots on Dialogflow and Microsoft Bot Framework for enterprise clients since 2016
From our lead engineer's roles at larger companies. See Work.
Fit and price
A good fit if
- You have a product and a specific problem AI could solve
- You have content or data to work with
Not a fit if
- Training your own models from scratch
- Data science or research projects
Price Fixed quote per feature after a 20-minute call. A first prototype usually takes 1–2 weeks.
FAQ
Is our data used to train the model?
No. We use API settings and providers that don't train on your data, hosted in the EU where you need it.
OpenAI or AWS Bedrock?
Bedrock fits teams already on AWS who want data kept in their account. OpenAI is often faster to start. We recommend one per case.
What does it cost to run?
We estimate cost per request before building and set alerts so there are no surprises.
Also useful
Tell us what you're building.
20 minutes, no pitch deck needed. You leave with a clear next step, even if it isn't us.