In development · Automated ECU file service In development

Tuning File Portal: automating a manual, high-stakes process

A tuning file service is a queue of skilled manual work where a mistake can damage an engine. It is the hardest kind of process to automate, and the most valuable.

In development Status
Q4 2026 Expected launch
1 minute Automated stage
14 minutes Human check
Summary

What this is

This is our next product, currently in development. A dealer uploads an ECU file, the system identifies what it is and finds a match, a series of verification gates confirm the result is safe and correct, and the tuned file is returned - with human approval at the points where the stakes justify it.

We're describing it here honestly as work in progress rather than a finished case study, because the interesting part is the problem rather than a result we don't yet have.

Our role
  • Process design and automation architecture
  • Verification gate design
  • Dealer portal and file handling
  • Ongoing development

Why this process is hard

Most automation removes tedium from a low-risk task. This one sits on top of work where a wrong output doesn't produce an awkward email - it potentially damages an engine, and the business carries the liability.

That changes the design completely. The question stops being how much can be automated and becomes which specific decisions can safely be made without a person, and what has to be true before any file is returned. A system that is right 99% of the time is not acceptable when the 1% is a destroyed engine.

Multi-gate verification

The approach is a series of gates rather than a single check, each one capable of stopping the process and routing to a human with the reason attached. Nothing proceeds on an assumption, and anything unrecognised stops rather than guessing.

Identify the file

What it is, what it came from, and whether it is intact and complete.

Match against known data

Finding the correct corresponding work, with confidence scored rather than assumed.

Sanity-check the result

Automated checks that the output is within expected bounds before a human ever sees it.

Human approval

A person confirms before anything is returned. This gate is not optional and will not be removed for throughput.

Full audit trail

Every step recorded, so any file returned can be reconstructed and accounted for afterwards.

What it is expected to change

The value is not in removing the expert. It is in removing everything around the expert - the intake, the identification, the matching, the record-keeping, the chasing - so that skilled time goes on the decision that actually requires skill.

It should also make the service consistent in a way a manual queue cannot be. Every request follows the same path, with the same checks, regardless of who is working and how busy the day is.

The target shape is an automated stage of about a minute and a human check of about fourteen. We're stating both numbers deliberately: the fourteen minutes is not a failure to automate further, it is the part we won't remove, because the cost of being wrong here is an engine.

An honest note

This is in development, expected in the fourth quarter of 2026. The timings above are design targets rather than measured results, and we've labelled them as such rather than presenting them as achievements.

If you're considering something similar in your own business, we're happy to talk about the design problem now rather than waiting until we have a launch to point at.

Built

What is in it

  • Dealer upload portal with validated intake
  • Automated file identification and matching
  • Multi-gate verification before any return
  • Mandatory human approval on output
  • Complete audit trail for every request
Stack
TypeScriptNodePostgresQueuesObject storage

Want something like this?

Tell us what your business runs on and where it hurts. We'll tell you what it would take to build something similar for you.