Skip to content
New guide: how to scope an AI project before you build

Intelligence

Artificial Intelligence

We find the place where AI genuinely helps, then build the data, evaluation and interfaces that make it useful in production.

Typical timeline
6-10 weeks
Group
Intelligence
First step
Fixed-price discovery

Scope, price and acceptance criteria are agreed in writing before the build starts.

Why us

Why choose us for this work?

The problem comes first

We start with the decision or process, not the model. A query or a rule is often the right answer.

Grounded in your data

Retrieval, sources and citations are designed in, so answers can be checked against what you actually know.

Measured before launch

We agree an evaluation set of real examples and score the system against it, not against a feeling.

Your team owns it

The data, evaluation sets, rules and interfaces live in your repositories and accounts, not with us.

Built for production

Permissions, monitoring, retries and rollback are part of the first version, not an afterthought.

What you get

What you get from Artificial Intelligence

An AI system you can measure, explain and change.

  1. 01

    A scoped AI system

    One concrete use case with a written definition of what good looks like.

  2. 02

    A grounded interface

    Answers with sources, confidence and a path to a human when it matters.

  3. 03

    An evaluation harness

    A repeatable set of examples and measures you can run before every release.

  4. 04

    Monitoring and controls

    Usage, cost, overrides and failure modes visible from day one.

  5. 05

    Documentation and handover

    Decisions, runbooks and tests your team can take over.

The problem

What usually goes wrong

Most AI projects start with a model and look for a problem afterwards. The result is a convincing demo that never becomes a system anyone trusts. The hard parts are data quality, evaluation, permissions and the boring work around the model.

Who this is for. Teams with a real decision or process to improve, and data they are willing to put behind it.

Signs you need this

If any of this sounds familiar

  • You have data that people currently read and interpret by hand.
  • A prototype worked, but nobody can say how well it really performs.
  • The same question gets answered differently by different people.
  • You need to know what the system is confident about and what it is not.

How the work runs

How we deliver it

  1. 01

    Discover

    We look at the data and the decision, and agree what a useful outcome means.

  2. 02

    Evaluate

    We build a small evaluation set and test a baseline before committing to a build.

  3. 03

    Build in steps

    Weekly working versions, with the retrieval and interface improving each time.

  4. 04

    Test the edges

    Missing data, ambiguous questions and confident wrong answers all get tested.

  5. 05

    Launch and hand over

    We ship, watch the measures, then hand the system to your team.

Deliverables

What you receive

  • Use-case definition and success measures
  • Grounded AI application with source citations
  • Evaluation harness and regression tests
  • Monitoring, access control and cost controls
  • Architecture decisions and handover docs

Stack

Tools we expect to use

TypeScriptPythonPostgreSQLVector searchCloudflare

We choose the stack to fit the problem. If your team already runs something else, we plan around it.

Questions

Common questions

Which model will you use?

It depends on the data constraints and the budget. The parts that matter, retrieval, evaluation, permissions and interfaces, stay yours and can be changed independently of the model.

Where does our data go?

We design around your constraints. Where data must stay in your environment, we use local or private inference. Where an external API is acceptable, we document exactly what is sent.

How do you know it works?

We agree an evaluation set of real examples before launch and score against it. A system without one is a demo, not a product.

Next step

Talk through your project.

A short conversation. You leave with a clear view of what it would take, and whether this service is the right first step.

Prefer email? Write what you are working on and we will reply.

Send a note

By sending this, you agree that we may reply to your enquiry. See our privacy policy.