Browse categories
Explore
Fiverr Pro
English
$
USD
I will build ai automation, ai agent, workflow automation in python
Michal
1 order in queue
About this gig
Your team does the same multi-step job every day: read something, pull the data out, put it somewhere, tell someone. I automate that job in Python.
RECENT WORK
A commercial property settles utilities with its tenants on its own: the workflow takes supplier invoices out of a mailbox, reads the rates and periods, works out what each unit used, and settles it against the advances paid. The model only reads the documents; the arithmetic is deterministic code.
WHAT YOU GET
- Your process running end to end, in your own environment
- Connections to your systems: API, database, inbox, spreadsheet or calendar
- Retries on failure and a log of every run
- Source code and a README
WHY IN CODE, NOT NO-CODE
Reliability. When a step fails at 2am this retries it and carries on from where it stopped. A no-code scenario just stops, and re-running it repeats the steps that already worked - so the same order gets written twice.
TYPICAL JOBS
- Read incoming documents and file the data where it belongs
- Watch a queue or an inbox and act on what arrives
- Pull data from several systems, build a report, send it
Describe your process in one paragraph and I'll tell you which package fits.
Get to know Michal
Michal
AI Product Engineer, AI automation in Python, AI Chatbots, Exited founder
- FromPoland
- Member sinceAug 2026
- Avg. response time1 hour
Languages
Polish, English, Spanish
I build AI integrations, AI automation, AI chatbots and agents in code: Python, FastAPI, Next.js, Postgres. (Not no-code).
Recent work: a legal-tech platform where reliable answers were the requirement, so every reference is verified before it is shown. A voice receptionist that holds a conversation and books a confirmed appointment into the scheduling system. Agent pipelines that retry failed steps and resume after a crash.
I founded a marketplace, sold it to an industry leader, and ran product there.
Packages have written limits: sources, pages, languages.
FAQ
What can the automation actually do?
Anything your systems expose through an API, a database or a mailbox: read documents, move data between tools, act on what arrives in a queue, produce and send reports. If a system has no API, tell me before ordering.
Will it act without asking me?
Only where you want it to. Any step that spends money, contacts a customer or changes a record can require your approval first - it waits, you confirm, it continues. That is included from Standard.
Why not n8n, Make or Zapier?
They are faster to start and harder to fix. Once a workflow matters, you need retries, a resume point after a crash and a trace of every run. That is what code gives you, and it is why this costs more.
Do I own the code?
Yes. Source code and a README are included in every package and it runs in your environment. You can hand it to another developer at any point.
What if my process changes later?
Small changes fit in the custom development extras (2, 5 or 10 hours). Bigger changes are a new order and I will quote them before you commit.
What will it cost me to run?
If you host it yourself you pay the model provider and your hosting directly, with no markup from me. Most single workflows land between $5 and $40 per month. I set a spending cap so it cannot surprise you.
Where does it run, and can you host it for me?
By default it runs in your own environment - your server or cloud account - so you keep control and pay the providers directly. If you would rather not run it yourself, I can host it and look after it. Ask me and I will send a monthly offer through Fiverr once the build is delivered.
What does the monthly plan cover?
Hosting if you want it, monitoring, model and library updates, fixing runs that fail, and small changes to the workflow. It is a Fiverr subscription of three or six months, starting after delivery. Larger changes are quoted separately before I start.
Who can see my data?
Only you, unless you ask me to host it. If I do, the server sits in the region you pick - EU or US - and I process your data only to run and fix your workflow. I sign a data processing agreement. Nothing is used to train models. I delete my copies when we stop working together.
What is kept in the logs?
By default a run log records the steps, the timings and the outcome, plus a reference to the record it worked on - not the contents. If you want the full payload stored for debugging, that is a switch, and you decide how long it is kept.
