One manual process, automated and running in production.
Three weeks to put a single real workflow into production on self-hosted n8n, with language models doing the judgement work and a human approving anything consequential. Running on your infrastructure, so your data never leaves your tenancy.
Singapore is two and a half hours ahead of our engineering team, so calls land inside both working days. An engineer replies within one business day.
This is worth your time if
- A team spends hours a week moving data between systems that do not talk
- You want automation that reasons over messy input, not just if-this-then-that rules
- Data residency matters, so a per-task SaaS automation tool is not acceptable
- You have tried a no-code tool and hit the ceiling where real logic was needed
It is not for you if
The process is not yet written down anywhere. Automating an undefined process produces an undefined result, and we would rather map it with you first than sell you a pilot that cannot land.
What gets built
One workflow end to end, on your own n8n instance. Every step is a node you can open, read and change after we hand it over.
-
Trigger Inbox, form or webhook
-
Extract Model reads the content
-
Approval A person confirms
-
Act Writes to your systems
Four deliverables, all written down
Everything is handed over as something you own and can act on without us. That is deliberate: if it only works with us in the room, it is not a deliverable.
One workflow live in production
Not a demo or a sandbox. A single end-to-end process running on your infrastructure, handling real volume, with the failure paths built.
Self-hosted n8n, deployed and documented
Containerised with a pinned version, TLS, backups of the workflow database, and a runbook. No per-execution pricing and no third party holding your data.
Human-in-the-loop where it matters
Anything consequential pauses for an approval rather than acting on a model output. Which steps those are is a decision we make with you, explicitly.
Handover your team can operate
A written runbook, the workflow exported as code, and a session with whoever will own it. The test is whether they can change it without us.
How the 3 weeks run
Stated day by day so you can see exactly what you are committing to, and what we need from you.
-
DAYS 1 - 4
Pick the process and map it
We choose one workflow together on the basis of hours saved and how clearly it is defined, then write down the current steps, inputs and exceptions.
-
DAYS 5 - 9
Deploy the platform
Self-hosted n8n with a pinned version, TLS, credential storage and backups, on your infrastructure or a region you nominate.
-
DAYS 10 - 17
Build and harden
The workflow, the model prompts, retries, failure alerts, and the approval gates. Run against real data in parallel with the manual process.
-
DAYS 18 - 21
Cut over and hand over
Switch the process across, watch it for several days, then hand over the runbook and train the owner.
What this engagement does not include
Automating a second workflow, model fine-tuning, and any licence your own model provider charges for inference. One workflow done properly tells you far more about whether this works for you than three done badly. If you want us to carry on, the fee for this engagement is credited in full against that work when it starts within 60 days.
Judge us before you pay us
We are a new name to you, so here is evidence you can check yourself rather than claims you would have to take on trust.
We self-host this class of system already
Our own CRM is self-hosted on Docker with a pinned version, nginx and TLS, and nightly database backups shipped to object storage and verified by restore. The operational practice in this pilot is the one we run.
Read how we think about the tooling
Part of 143 technical articles under our own bylines, including work on AWS automation and conversational interfaces.
Verify the company exists
Chadura Tech Private Limited, incorporated 8 May 2018. Our CIN and GSTINs are published and can be checked against the Indian government registers.
Company record and teamTell us about your setup
A couple of answers is enough to start. We will tell you on the first call whether this is worth doing, including when the answer is no.