Artificial Intelligence
AI that works inside the process, not in a loose chat: assistants that know your documents, agents that carry out tasks with defined permissions and MCP servers that connect it all to your systems.
Who it is for
Signs that this service solves your problem.
- 01
Teams losing time looking for information in documents
- 02
Processes with repeated steps an agent can run and a person can review
- 03
Companies that want ChatGPT or Claude to access their data safely
What is included
What you get at the end — in writing in the proposal.
- Assistant with the company knowledge base
- Agent with defined tools and permissions
- MCP server for your systems
- Skills for the team to use in Claude and ChatGPT
- Usage logs and human review
- Cost control per use
How I deliver
Everything in writing with online previews; a call only if you prefer.
- 01
Pick the process
One process, one measurable result — in writing, before any code.
- 02
Sources and limits
What the agent can read, what it can do and when it needs approval.
- 03
Pilot
The agent runs with the team reviewing every output for a few weeks.
- 04
Operation
Logs, time-saved metrics and ongoing adjustments.
- Assistant with knowledge base
- 1 to 3 weeks
- Agent with tools
- 2 to 5 weeks
- MCP server
- 1 to 3 weeks
The exact timeline goes in the proposal, together with the price.
Questions
Reference models and tools — chosen for cost, quality and privacy.
01Will my data train the AI?
No. I use APIs with retention disabled when available, and the agent only accesses the sources you authorize.
02How much does it cost to run?
I estimate the per-use cost before the pilot and set up limit alerts.
OpenAI
Claude
Gemini
LangChain
Cursor
Python
n8n
Next step
Shall we get this built?
Send the brief by email, WhatsApp or the form. I reply with the missing questions and then with scope, timeline and price in writing — no mandatory meetings.
- 01Written brief
- 02Fixed proposal
- 03Deliveries with online previews