Home/Software development process
How we build software

Our software development process. Scope, build, check, run.

Gigabit’s software development process has four stages. Scope takes one to two weeks and ends in a written fixed price and ship date. Build runs in two-week sprints with a working demo at the end of each. Quality gates check every release. Run keeps it healthy after launch, from $3,000 a month, or we hand it over.

The four stages

The four stages of our software development process

The same stages on every build, from a single workflow to a full platform, so you always know where the work stands and what comes next.

  1. 01 · Scope

    Agree what to build, and what it costs.

    An engineer maps the users, the workflows, the data model, the systems it connects to and what stays out of version one. The first call and the written quote are free. Where the work needs a deeper look first, paid scoping is there: the $2,500 Prototype Production Review of an AI-built app, or the $25,000 AI Transformation Sprint for a larger AI programme.

    How long
    1–2 weeks
    Ends with
    A written scope, a fixed price and a ship date.
    Gate to the next stage
    You approve the price. Nothing is billed before then.
  2. 02 · Build

    Build it in two-week sprints.

    Every sprint ends with a working demo, not a status report. The code lives in your repository from the first sprint. AI agents write the repetitive parts (tests, migrations, boilerplate and documentation); senior engineers design the system and review every change before it merges.

    How long
    2-week sprints
    Ends with
    A working demo every sprint, and the code in your repository.
    Gate to the next stage
    A change to the scope is priced in writing before it is built.
  3. 03 · Quality gates

    Nothing ships until it passes.

    Before every release: automated tests, observability (logs, tracing and alerts), guardrails and a rollback path that has been tried. AI features also get evals: test cases built from real examples that show the feature answers correctly before it goes live.

    How long
    Before every release
    Ends with
    A release that passed its tests, with monitoring in place.
    Gate to the next stage
    Release only when the tests and evals pass and the rollback path works.
  4. 04 · Run

    Keep it running, or take it over.

    After launch, Managed AI Operations keeps the system healthy: monitoring, eval regressions, fixes when a vendor changes a model or an API, a monthly report and a quarterly review. Or we hand over with the documentation, and your team or another one runs it.

    How long
    Monthly, $3,000–$20,000/mo
    Ends with
    A system that keeps working, a monthly report and a quarterly review.
    Gate to the next stage
    Each quarter the system is reviewed against the number it was built to move.
Who does what

How we build software: what AI agents do, and what engineers do

AI agents speed up the build. Senior engineers stay responsible for every line that ships.

WorkDone byChecked by
Scope, architecture and data modelSenior engineersYou, before you approve the price
Tests, migrations, boilerplate and documentationAI agentsA senior engineer, on every change
Product code and integrationsSenior engineers, with AI coding agentsCode review before every merge
The releaseThe quality gatesTests, evals and a working rollback path
Running it after launchManaged AI Operations, or your own teamA monthly report and a quarterly review

That split is how a small senior team can commit to a fixed price: our logistics SaaS MVP cost $48,000 and shipped in 11 weeks. What other builds cost and delivered is in our software development case studies, and every fixed price is on software development pricing. The same process runs every web app development and MVP we take on.

How a sprint runs

Two weeks, one demo, one shared board.

Each sprint is planned on a call and ends with a demo of working software. In between, the work is async-first, in the tools your team already uses. Engineering is in Dhaka (UTC+6). For US clients the work runs as an overnight hand-off: what you send before you log off is worked on that night, and planning and demos happen on scheduled calls. Europe, the Gulf and Australia share several working hours with Dhaka.

2-week sprintsA working demo every sprintOvernight hand-off for US teamsSlack · GitHub · Jira/Linear · Notion

The hours for each region are on where we work.

What you own

What do you own at the end?

The code

In your repository from the first sprint, under your review rules.

The cloud accounts and data

Yours from the first day, so you can move them to another team at any time.

The tests and evals

The checks that prove it works, so any team can change it safely.

The documentation

How it is built, how to run it and how to roll it back.

With those four, any competent team, yours included, can pick the system up. Without them, each new vendor starts from a blank page because the last one left nothing behind.

For AI programmes

Audit → Build → Operate: the same process, for AI

When you know you want AI but not where it pays, the four stages run as three phases with fixed prices. Every phase ends in software that runs, not a report.

PhaseOfferPriceTimeEnds with
AuditAI Transformation Sprint$25,0002 weeksA pilot on your own data, an ROI model and a costed build plan
Audit, one known workflowAI Quick Win Sprint$8,0003 weeksOne workflow built and deployed, with 30 days of light support
BuildGigabit AgentsFrom $8,000 · flat, per agentWeeks, sprint by sprintA production agent, its evals, the docs and the code, in your repository
Build, with a teamEmbedded AI Teams$3,000–$7,500 / engineer / moAssembled in 2 weeksSenior AI engineers in your team, first pull requests by day 5
OperateManaged AI Operations$3,000–$20,000/moMonthlyMonitoring, eval regressions, drift and vendor-change response, a quarterly review

The $25,000 Sprint fee credits toward a Gigabit Agents build. The engineers who scope the work are the ones who build and run it, working in your tools and close to the workflow: that is what forward deployed engineer means. If you are weighing it against hiring or buying a product, the build vs buy vs embed comparison sets out the trade-offs.

Principles

What we put first

Engineering before content

We were engineers first.

Pricing before pitching

The number is on the page.

Evals before deploy

We prove it works before you trust it with production.

Production before pretty

A working system beats a beautiful mockup.

Adoption before scope

One workflow in production beats ten in a plan.

Honesty before optimism

We tell you what won’t work, too.

FAQ

Our software development process, answered

What is Gigabit's software development process?

Gigabit's software development process has four stages: Scope (one to two weeks, ending in a written fixed price and ship date), Build (two-week sprints with a working demo each), Quality gates (tests, observability, a rollback path and evals before every release) and Run (Managed AI Operations from $3,000 a month, or a hand-over to your team).

How long does it take to build software with Gigabit?

Scoping takes one to two weeks. Most first versions then take 8 to 16 weeks of two-week sprints; Gigabit's published MVPs took 11 and 14 weeks. Smaller jobs are faster: one workflow ships in three weeks as a fixed-price sprint.

How can an agile process have a fixed price?

Scoping comes first and ends in a written scope, price and ship date, so the sprints build against an agreed target. If you change the scope during the build, Gigabit prices the change in writing before building it, and you decide whether it goes in.

How does Gigabit use AI to build software?

AI agents write the repetitive parts of a build: tests, migrations, boilerplate and documentation. Senior engineers design the system, write the product logic with AI coding tools and review every change before it merges. Nothing ships without automated tests, observability and a rollback path.

Who owns the code and the data?

You do. The code is in your repository from the first sprint, and the cloud accounts and the data are yours from the first day. The tests, evals and documentation are handed over with them, so any team can take the system on.

Do I have to keep Gigabit after launch?

No. After launch you can keep Managed AI Operations ($3,000–$20,000/mo) or take the system over with its documentation. Gigabit can also be brought in only to operate a system another team built.

What does forward-deployed mean?

Forward-deployed means the engineers who scope the work sit close to your workflow, in your tools and with access to your stack, and they are the same people who ship it and stay accountable for the result. It is the opposite of a strategy deck delivered from a distance or a team briefed through a ticket queue.

How is Audit → Build → Operate different from a consultancy's phases?

Every phase ends in working software you own. The audit (the $25,000 AI Transformation Sprint) ends in a deployed pilot rather than a findings deck, the build ends with code and evals in your repository, and operations are measured against the number the system was built to move.

Go deeper

From the field notes

Next step

Start with the scope.

Tell us what you want built. An engineer replies within one business day, and you get a fixed price and a ship date in writing before any work starts.