Synthetic IndustryGet a price

Case study · Software team

A small product company had a backlog and no capacity. An AI software engineer now works through it every day.

Features waited behind fixes and upkeep. An AI software engineer takes each request to something the founders can try, and they decide what ships.

Who
A software-as-a-service company run by two founders, with one web product and no in-house developers.
The problem
Customers asked for features that waited behind fixes and upkeep, and the founders could not manage freelancers day to day.
What we did
An AI software engineer took on the product's backlog: features with a working preview, bug fixes with a test that proves them, reviews and upkeep.
The result
The founders prioritise, try each change in a preview, and accept it or ask for changes. A named founder merges and releases.
Time
Ongoing. A monthly service.
Get a price for something like thisThe service behind it

What was going on.

The company had a product customers liked and a list of things customers asked for. The founders could write a good request, but neither was a developer, and the freelancers they had used needed direction every day. Every feature waited behind a fix, and every fix waited behind upkeep.

They did not want to manage coding sessions. They wanted someone to take a request from them, come back with something they could try, and keep doing that, every working day.

What we found.

The agent works through the backlog the way a full-time engineer would, but it takes each item from a written brief to a result the founders can try. The brief says what should change, the edge cases that matter and the checks that will count as done. Large requests are split before anything is committed.

Every item comes back as a working preview and a plain-English walkthrough, with the checks that were run, the evidence, and anything that could not be verified. Bug fixes arrive with a test that fails before the fix and passes after it. A separate review looks for mistakes, and the founders accept the result or send it back.

The technical detail

Work happens in an isolated working environment, away from production, using representative made-up data with live email, payments and other outside effects switched off. Each result arrives as a proposed change (a pull request) with its checks run. Corrections needed to meet the agreed checks stay within the same item, and new requirements are re-scoped. A release happens only through a route the company has authorised.

What we did.

How the work ranSteps in order: 1 Write the brief: What should change and what counts as done; 2 Work on a copy: Away from production, real effects off; 3 Run and challenge: Tests, the running app, a separate review; 4 Show the result: A preview, a walkthrough, the evidence; 5 You decide: Accept it, or send it back with comments; 6 Release your way: Only through the route you authorise. You decide at step 1 and step 5.1 Write the briefWhat should change andwhat counts as done2 Work on a copyAway from production,real effects off3 Run and challengeTests, the running app,a separate review4 Show the resultA preview, awalkthrough, theevidence5 You decideAccept it, or send itback with comments6 Release your wayOnly through the routeyou authoriseYou decide here How the work ranSteps in order: 1 Write the brief: What should change and what counts as done; 2 Work on a copy: Away from production, real effects off; 3 Run and challenge: Tests, the running app, a separate review; 4 Show the result: A preview, a walkthrough, the evidence; 5 You decide: Accept it, or send it back with comments; 6 Release your way: Only through the route you authorise. You decide at step 1 and step 5.1 Write the briefWhat should change and what counts as done2 Work on a copyAway from production, real effects off3 Run and challengeTests, the running app, a separate review4 Show the resultA preview, a walkthrough, the evidence5 You decideAccept it, or send it back with comments6 Release your wayOnly through the route you authoriseYou decide here
The steps in order. A box with a coloured border is where the client decided.

How we checked it.

These had to pass before we asked for sign-off.

  • Each result has agreed checks, and every check produces evidence the founders keep.
  • Each bug fix comes with a test that fails before the fix and passes after it.
  • A separate review of each change found no unresolved mistake before it was shown.
  • Nothing is released to production by the agent. A named founder merges and releases.

What they have now.

The founders now send requests as short written briefs and receive previews to try. Most days at least one item arrives ready to accept or to send back with comments. Features no longer queue behind upkeep, because the same agent keeps the tests, the documentation and the dependencies current.

A named founder reviews and merges each change. The repository, the tests and the notes all live in accounts the company owns.

What we did not do.

  • No human supervisor is included unless the proposal names one. The decisions stay with the company: a named founder approves merges, releases and live changes.
  • The agent works in an agreed copy of the product, not on production, and holds no live credentials or customer data.
  • We promise no result about speed or cost, and a separate review is not a guarantee that every defect will be found.
  • The work covers one product. Several products, or a whole engineering department, is a different agreement.

The service behind this.

If your product's backlog is bigger than your capacity, a full-time AI engineer can hold the delivery while your people keep the decisions.

Full-time AI agent

AI software engineer

A full-time AI engineer for one existing web product: features, previews, fixes, PR review, CI, upgrades, tests and docs.

$10,000/month. Payment, billing, cancellation, renewal and treatment of incomplete work are agreed in the written proposal before the engagement starts. No charge, reservation or automatic subscription begins from a message.

Get a priceSee what is included

More case studies.

All case studies

Get a price for something like this

You are asking about: AI software engineer.

Add detail (optional)

A web address anyone can open, with no login in it. No code or logs.

We reply by email. Sending this emails your request to our team; nothing is kept in a website database. Please do not send passwords, keys or customer records. Privacy notice.

Sending this is not an order. We check we can help, then agree what is included, safe access and the price in writing before any work. Prefer email? hello@syntheticindustry.ai.