01 · Starter
- Scope
- A single flow connecting two tools, or repairing a broken flow
- Timeline
- 2–5 days
- Conditions
- Client grants access to the tools
HOW I WORK
How I work and how I quote
Send me one process or idea first. I reply with what I would build, what I would leave manual, and a written scope. A price follows once the scope is clear.
From first message to handover
- 01
Brief
You describe how the process runs today, the tools and data involved, and the result you want.
- 02
Scope
I reply with what I would automate first, what I would leave as it is, and what I need from you.
- 03
Quote
I send a written scope, timeline, conditions and price. Work starts after you confirm in writing.
- 04
Build
I build to the written spec and test on your sample data, with error handling and a person approving risky steps.
- 05
Handover
You get the work, setup notes, and a list of what was tested and what was not.
What to send for a quote
Answer what you can. Not knowing something is fine; I will ask.
- What happens today, step by step, and who does it
- Which tools and systems are involved, for example Zapier, Make, n8n, Google Sheets, Airtable, Notion or Telegram
- A few sample inputs or documents you can share (no passwords or keys)
- The result you want and how you will judge it
- Your deadline and budget range, if you have one
Packages
Each service comes in three packages with a defined scope, timeline and conditions. Every project is confirmed in writing before work starts.
02 · Standard
- Scope
- 2–3 flows with schedules, error alerts and handover documentation
- Timeline
- 5–10 days
- Conditions
- A test account is available
03 · Advanced
- Scope
- A multi-step flow system with CRM or internal integrations, documentation and a walkthrough
- Timeline
- 2–3 weeks
- Conditions
- Scope agreed in writing
01 · Starter
- Scope
- One LLM task (classify, summarize, extract) attached to an existing workflow
- Timeline
- 3–5 days
- Conditions
- Client provides their own API key
02 · Standard
- Scope
- A Q&A chatbot over one document set (RAG), tested on sample data
- Timeline
- 5–10 days
- Conditions
- Client has the right to use the data
03 · Advanced
- Scope
- Multiple document sources, answer-quality measurement, deployment to your infrastructure (Docker, a reverse proxy or a service you choose) checked at handover, and a walkthrough
- Timeline
- 2–4 weeks
- Conditions
- A named owner on the client side
01 · Starter
- Scope
- An FAQ bot on one platform (Telegram)
- Timeline
- 3–5 days
- Conditions
- Client supplies the Q&A content
02 · Standard
- Scope
- An agent using 1–2 permitted tools (look up, then draft a reply) with saved history
- Timeline
- 5–10 days
- Conditions
- Read-only access where possible
03 · Advanced
- Scope
- A multi-tool agent on Telegram and the web, with logs, permission limits, and an approval gate on each tool for irreversible actions
- Timeline
- 2–3 weeks
- Conditions
- The list of allowed actions is agreed. An admin page is custom work, with its own scope and timeline quoted separately
01 · Starter
- Scope
- A one-feature prototype to test an idea
- Timeline
- 3–7 days
- Conditions
- Requirements written down
02 · Standard
- Scope
- An MVP with an interface and one main AI flow (user accounts and a database are scoped separately)
- Timeline
- 2–4 weeks
- Conditions
- Scope fixed before work starts
03 · Advanced
- Scope
- A fuller internal tool or product, with deployment and a walkthrough
- Timeline
- 4–8 weeks
- Conditions
- Payments split by milestone
How I keep the work dependable
Errors are handled
Flows get an error branch and an alert, so a failure is seen instead of silently dropped.
A person approves risky steps
Anything that sends, deletes or spends goes through an approval step, and permissions stay as narrow as possible.
Tested on your data
I test on your sample data before handover and write down what was tested and what was not.
Where AI is not the answer
If a rule, a filter or a spreadsheet formula does the job, I say so instead of adding a model.
AI running costs
I can review what a flow spends on model calls and suggest cheaper settings. This is advice per order, not a ready-made product.