Custom software and AI products

The problem

A spreadsheet has become a product, an internal process needs a proper tool, or an idea needs a first working version. Off-the-shelf software does not fit, and an AI feature needs to sit inside the product rather than next to it.

What it looks like

Generic examples of the kind of work. They are not client results.

• An internal tool replacing a spreadsheet

  Structured records, roles and an audit trail where there used to be a shared file.

• An API other systems can rely on

  A documented, tested interface for partners or for your own applications.

• A product with an AI feature built in

  For example a planner that works from a checked directory, designed so it cannot invent items outside it.

What you receive

• ✓A written spec and architecture before building starts
• ✓A working first version scoped to what can be tested
• ✓Automated tests and a deployment process
• ✓Source code and documentation handed over as agreed in the contract
• ✓A plan for who maintains it afterwards

The pilot and how we judge it

One user journey built end to end, so the team can use it before the rest is built.

Acceptance criteria, written before the pilot starts:

• ✓Acceptance criteria are agreed in writing before building.
• ✓The journey works on real data for the people who will use it.
• ✓Tests cover the behaviour the criteria describe.
• ✓Open problems and known limits are listed, not hidden.

Boundaries and honest limits

• We scope before we promise dates. A custom build is not a free trial.
• Our public project stories are listed on the Work page with their actual status. A project in development is not described as a finished customer success.
• Design, growth and maintenance work can draw on the wider MediaElegant team when a project needs it.