AI and automation that gives you hours back
Work your team repeats by hand costs hours every week and lets mistakes through until a client spots them. We automate it, with rules first and AI such as Anthropic's Claude where rules can't do the job, and a person stays in charge.
What we do
- AI assistantsAnswers from your own documents, with sources.
- Document automationFields extracted, documents drafted and checked.
- Smart routingEmails and requests sent to the right person.
- Workflow automationThe repetitive clicks, done by software.
- Claude and LLM integrationAI inside the systems you already use.
- Safe by designData protection and a person in charge.
How it works in practice
AI and automation, at Kazap, means taking a task your team repeats by hand and having software do it. Most of those tasks follow rules: a request about property goes to the property team; a form with a company number fetches that company's details. Rules are cheap, predictable and easy to check, so we use them first.
AI comes in where a rule cannot do the job: reading a free-text email to decide where it goes, pulling fields out of documents in different layouts, drafting a first reply for a person to check, or answering questions from your own documents. We build with Claude every day, including integrations through the Model Context Protocol (MCP).
The automation we have delivered carries real work: a legal-document platform that assembles each document from saved data; a portal that routes requests to lawyers in 45+ countries in place of email threads with 50+ recipients; a lease platform that cut a sign-off from up to two weeks to two hours.
How it runs
Every step ends with a decision that is yours.
- List the tasks
We sit with your team and time the repetitive work.
You choose one task - Prototype
The task run on a sample of your own data.
Accurate enough to pilot? - Pilot
Inside the real process, with a person checking the output.
Go or no-go on agreed criteria - Version one
In daily use, with the checks kept in place.
Reviewed after launch - Support
We host it, watch it and improve it.
Keep, change or expand
Questions people ask us
Not here? Ask us over a call or a coffee.
How can a small business use Claude or ChatGPT safely with company data?
Through business APIs and settings that do not train on your data, with personal data kept out of prompts where possible, access controls, and a record of what goes where, agreed before the pilot starts.
How do I integrate Claude or another AI model into my software?
Usually through the provider's API, with retrieval over your own documents, guardrails, and a person reviewing what matters. We prototype it on a sample of your data first.
What can AI automate in a small business, and what can't it?
It is good at reading, sorting, extracting and drafting. Decisions nobody checks, and anything a client or regulator could be harmed by unseen, stay with a person or with rules.
Do we need AI, or would plain automation do?
Often plain automation. We use rules wherever a task follows them, because they are cheaper, more predictable and easier to check, and add AI only where they cannot do the job.
How is AI work priced?
The prototype and pilot are a fixed price. Version one is quoted after the pilot, with build and running costs shown separately; support afterwards is a day rate.
Let's talk over a call or a coffee
Tell us about your idea or what's slowing you down. We'll listen, ask the awkward questions, and tell you honestly whether we can help.
- Marcelo reads every message himself
- A call, or a coffee in London, at a time that suits you
- A straight answer: a plan, a price, or "you don't need us"