Services

What we build, and how.

Custom software and AI integration for businesses that have outgrown the spreadsheet, the inbox or the license. Scoped in phases, priced up front, owned by you.

  • Web apps
  • Internal tools
  • Data & workflow
  • Documents & reporting
  • Mobile, offline-first
  • AI integration

What we build

Systems shaped like your operation

Custom software from scratch

  • Web apps and internal tools: tracking, quoting, scheduling, approvals, portals.
  • Data and workflow: forms, imports, calculated fields, revisions, search.
  • Documents: one-click PDFs, dashboards, the exports people actually want.
  • Mobile: where work leaves the desk; keeps working offline.
How this usually goes

The process exists only in someone’s head, a whiteboard, or a chain of spreadsheets. We start by drawing it as it is, then build the smallest system that runs it end to end: structured forms with validation, imports from the tools you already use, revision history, and the one report everyone needs.

Each slice is usable on its own, so the team starts working in it well before the last feature lands.

Replacing an off-the-shelf package

  • Start from what your team does, not the feature list.
  • Keep what works: exports, integrations, reports; migrate the history.
  • Run in parallel until you are confident, then retire the license.
  • End with one job done well, for a hosting bill instead of a per-seat fee.
How this usually goes

We watch how the package is actually used. Usually that is a short list of screens, and a long list of workarounds. The replacement covers the short list properly, keeps the exports and integrations people depend on, and imports every record, including the history.

It runs beside the package until you trust it. Then the subscription ends, and the tool changes whenever your process does.

AI integration

  • Intake and triage: messages and forms become records, routed by urgency.
  • Document extraction: invoices, data sheets and contracts read in; a person confirms.
  • Assistants over your own data: answers from your records, source shown.
  • Drafting: first drafts of quotes and replies inside the tools people use.
Where AI is not the answer. A missing system is not fixed by a model on top of it. Deterministic rules are cheaper where they apply, and we say so.
How we run AI features safely

Every AI feature runs on the server, against only the data the feature needs, through providers whose terms do not permit training on your data. Nothing runs in a browser tab with your records in it, and you decide what a feature may do on its own, such as booking a routine job, and what waits for a person to confirm.

How an engagement runs

Five steps, none of them a surprise

  1. 01

    Discovery conversation

    About an hour. Show us the bottleneck as it is.

  2. 02

    Written scope, fixed price

    What phase one includes, leaves out, costs, and when.

  3. 03

    Build in weekly slices

    Usable software every week or two, demoed to you.

  4. 04

    Launch and handover

    Data migrated, people trained, everything in your name.

  5. 05

    Support, if you want it

    Monthly, or nothing. The system runs without us.

What each step looks like in practice

Discovery. We want to see the bottleneck: the spreadsheet, the inbox, the software you resent. We ask who touches it, how often, and what breaks.

Scope. You get a document that says what the first phase includes, what it deliberately leaves out, what it costs, and when you will see it. Phases are small enough to be a decision, not a leap.

Build. Something usable every week or two, demoed on a call. You steer while steering is cheap. Real users see it before the phase ends, not after.

Handover. Documentation is written for the person who maintains it after us. Source code, database and hosting accounts are in your name.

Support. A monthly arrangement for fixes, small changes and keeping the platform current, or nothing at all.

Cost

Why this costs less than it sounds

  • Weeks, not quarters

    Modern frameworks make a robust web app a matter of weeks.

  • Tens of dollars a month

    Hosting for a business-sized system, no licenses that scale with headcount.

  • A small team ships more

    AI-assisted development means two people deliver what used to need ten.

A fixed build cost, then hosting. We do not publish prices, because scopes differ, but every proposal states one.

The longer version

Large software providers price for the largest customer they hope to win, then sell everyone else the same thing per seat, per month, forever. Custom software used to cost more because everything was built by hand.

Three things changed that. Modern frameworks and hosting make a robust, secure web application a matter of weeks, not quarters. Hosting for a business-sized system costs tens of dollars a month. And AI-assisted development means a small team ships what used to need a large one. So the numbers now work the other way round.

Questions we get

FAQ

Who owns the code?

You do. Source code, the database, the hosting account and any third-party accounts are set up in your name from the start. If we part ways, nothing needs to be handed back.

We already use a package. Do we have to replace all of it?

No. Often the right first phase is a tool that sits beside the package and removes the worst of the manual work around it. Replacing the whole thing is a decision you can make later, with evidence.

How long does a first phase take?

Most first phases are measured in weeks. The scope document gives a date. Because we build in slices, you see working software well before that date, not only at the end.

What technology do you use?

Mainstream, well-supported stacks that any competent developer can pick up: modern web frameworks, managed databases, established cloud hosting, and mobile frameworks that ship to both app stores from one codebase. We do not invent a stack per client.

Is our data safe with AI features?

AI features run on the server, against only the data the feature needs, through providers whose terms do not permit training on your data. You decide what a feature may do on its own and what waits for a person to confirm.

What if we are not sure custom software is the answer?

Then the discovery conversation is the right place to find out. Sometimes the answer is a spreadsheet fix, a setting in the tool you already have, or nothing. We will tell you.

Next step

Tell us about the bottleneck.

A short conversation is usually enough to tell whether custom software is the right fix, and what it would take.