Services How we work About Articles Start a project
Start a project

Define

Discovery
Phase

When AI writes the code, the specification is the product. We work out what the system has to do, what your data really looks like and which systems it has to talk to — before anyone builds the wrong thing quickly.

Start a conversation How it works ↓

What you get

Everything you need
to build — with anyone.

Workflow map

How the work actually happens today, drawn with the people who do it — including the exceptions and workarounds nobody wrote down.

Data audit

What your data really contains: volumes, gaps, duplicates, free-text fields, and what has to be cleaned before anything is built on top of it.

Integration map

Every system the solution has to read from or write to, with what each one's API really allows — tested, not taken from the vendor's brochure.

Architecture & specification

A specification precise enough for engineers and coding agents to build from, with the architecture, the edge cases and the decisions behind them.

Acceptance criteria

How everyone will know it works: measurable criteria and, for AI features, an evaluation set built from your real cases.

Build plan & estimate

Phases, risks, a realistic timeline and a firm price for the build — or an honest recommendation not to build, if that is the answer.

Process

Two to four weeks,
four steps.

I

Scope call

One conversation to agree what discovery covers, who we need to talk to, and the fixed price.

II

Interviews & access

We sit with the people who do the work and get read access to the systems and data involved.

III

Audit & prototype

We test the integrations, measure the data, and prototype the riskiest part to prove it can be done.

IV

Specification & plan

You get the documents, a walkthrough with your team, and a build plan you can take anywhere.

Discovery workshop

Why the order flipped.

Five years ago, building was the expensive part and specification was a formality. Now a coding agent can produce a working feature in an afternoon — so a vague brief no longer costs you a slow build, it costs you a fast build of the wrong thing.

The hard questions have not moved: where the data lives, which system is the source of truth, what happens when the API is down, who has to change how they work. Discovery is where we answer them, while answers are still cheap.

Not ready for a call? The free AI automation assessment takes two minutes and tells you which task is worth automating first.

Common questions

What clients usually ask first.

How long does a discovery phase take?
Usually two to four weeks. A single workflow with two or three integrations sits at the short end; anything that crosses departments or touches several legacy systems sits at the long end. We agree the length before we start, not halfway through.
How is discovery priced?
As a fixed price, agreed up front from a first call about scope. You know the cost before we start, and the result is yours whether or not we build what it describes.
Do we have to build with you afterwards?
No. The specification, architecture and build plan are written so any competent team can deliver from them, including your own. Most clients do continue with us, but the deliverable does not depend on it.
We already have a specification. Do we still need this?
Possibly not. Send it over and we will tell you honestly. A specification that names the data sources, the integrations, the edge cases and how success is measured is ready to build from. One that lists features is a wish list, and that is where discovery earns its cost.
Why not just build a prototype with AI and see?
Do that too — a prototype is a cheap way to find out what you want. It is a poor way to find out what your data looks like, which systems will refuse to cooperate, and who has to change how they work. Discovery answers those questions before they become rework.

Know the problem,
not the solution?

Start a conversation