Book a free data assessmentFree assessment

USE CASES / AI

The eight things we get asked to build most often.

Most AI work in a company does not need a new model. It needs the data underneath to be clean, governed and documented, and then a fairly simple pattern applied on top of it. These are the eight patterns that come up again and again. Each one is a few weeks of work, not a research project.

Book a free data assessment All use cases
  • 01

    Pipeline failure summariser

    When a data pipeline fails at 3am, it reads the logs and writes a plain-English summary of what broke, what it affects, and what usually fixes it — before anyone opens their laptop.

    Who uses it:
    data engineers, on-call
    Typical build:
    3 to 4 weeks
  • 02

    Data catalog assistant

    Answers "which table do I use for this, and can I trust it?" across your warehouse, using the metadata and lineage the platform already collects.

    Who uses it:
    analysts, new joiners
    Typical build:
    4 to 6 weeks
  • 03

    Governed text-to-SQL assistant

    Lets a business user ask a question in normal English and get a correct answer, by writing a query against the approved metric layer only — never raw tables.

    Who uses it:
    business teams
    Typical build:
    5 to 8 weeks
  • 04

    Data quality issue explainer

    When a quality test fails, it explains in plain words which rule failed, which records caused it, and which reports are now affected.

    Who uses it:
    data teams, report owners
    Typical build:
    3 to 4 weeks
  • 05

    Cloud cost assistant

    Reads your cloud and warehouse billing data and answers "why did the bill go up last month", naming the specific job or query that caused it.

    Who uses it:
    engineering leads, finance
    Typical build:
    3 to 4 weeks
  • 06

    Data lineage explainer

    Turns technical lineage into a readable answer: where a number came from, what rules touched it, and what breaks if you change it.

    Who uses it:
    analysts, auditors
    Typical build:
    4 to 5 weeks
  • 07

    Internal document assistant (RAG)

    Answers questions across your own documents — policies, runbooks, architecture notes, past projects — always showing the source paragraph.

    Who uses it:
    whole organisation
    Typical build:
    4 to 8 weeks
  • 08

    AI-readiness assessment

    Not a build. A written review of whether your data estate can actually support the AI use cases you have in mind, and what has to change first.

    Who uses it:
    leadership
    Typical build:
    2 to 3 weeks

Before any of this, the boring part.

Every one of these needs the same foundation underneath it: data that is joined up, tested, documented, and controlled by real permissions. If that foundation is not there, the pilot will demo beautifully and then quietly die when it meets real data and real users. That is not a warning we add to sell more work. It is the single most common reason these projects fail, and we would rather tell you now than four months in.

If you are not sure where your data stands, start with the readiness assessment. It takes two to three weeks and you get a written answer either way.

What every one of these includes

  • A model that runs inside your own cloud account. Your data does not leave it.
  • Answers that show their sources. If it cannot show a source, it says it does not know.
  • Permission rules, so a person can only retrieve what they are already allowed to see.
  • A log of every question asked and every answer given.
  • A test set of real questions with agreed correct answers, so quality can be measured instead of guessed.
  • Documentation and a handover session, so your team can change it after we leave.