Skip to content

Methodology notes, not market news.

We write about what we do: how to model the economics of an initiative, where rollouts break, and what a process needs before it can be automated at all. No predictions about the future of AI and no retelling of other people's reports.

An open notebook with handwritten notes and highlighted lines
  • A pricing worksheet with rows for materials, labour and expenses, a hand and pen over it

    Modelling the economics of an AI initiative before you start it

    A baseline cannot be reconstructed after the fact. What to measure before work starts, three ways to model the effect, and the cost line that gets forgotten most often.

    Aplora5 min read
    Read
  • A board of sticky notes sorted into backlog, this week and in-progress columns

    Why an AI pilot does not become production

    The prototype works, the demo went well, and three months later nobody uses it. Five reasons this happens — and which of them can be closed before the start.

    Aplora4 min read
    Read
  • A rep in a headset writing notes in a notebook during a call

    What a sales book needs before it can be automated

    The difference between a description and a criterion, why “build rapport” cannot be checked, and what a wording looks like when the system and the sales lead reach the same conclusion.

    Aplora3 min read
    Read
  • A printed contract with a pen on a wooden desk

    Fact control in automatically generated documents

    A document with an invented term is more dangerous than one never written. Five mechanisms that close that risk, and the rule for deciding what may be generated at all.

    Aplora3 min read
    Read
  • An advisor in a headset at their desk, pointing at the monitor during a call

    How conversion from hot leads at LearnIT grew 3.3×: an anatomy of the sales pipeline

    An implementation walk-through: how LearnIT turned call chaos and lost hot leads into a managed sales pipeline. Conversion went from 9% to 30%, calls that went nowhere fell from 60% to 20%, and quotes and contracts started going out in 10 minutes instead of a day.

    Aplora14 min read
    Read
  • A printed table of amounts and percentages, a calculator and a laptop on a desk

    Zero manual entry: how a pizza chain turned finance into a managed pipeline

    Incoming PDF invoices from aggregators and suppliers stopped being retyped into spreadsheets. A walk-through of the end-to-end pipeline: message classification, field extraction, duplicate defence, a parking lane for disputes, and a second loop for outgoing invoices.

    Aplora13 min read
    Read
  • Two support advisors in headsets working at laptops in a shared room

    From sampling calls to analysing all of them: an anatomy of the shift

    The supervisor listened to ten conversations out of a thousand and argued with the quality manager about who slipped. A walk-through of a pipeline that reviews every call: transcription, a two-loop analysis, CRM enrichment, dashboards and personal recommendations for the advisor.

    Aplora13 min read
    Read
  • An open notebook with a handwritten outline, section tabs and a pair of glasses, a laptop beside it

    A textbook in three weeks instead of six months: anatomy of a content pipeline

    How to cut the time and cost of producing course textbooks in EdTech: prompt packages instead of one magic request, human-in-the-loop with explicit SLAs, and feedback drawn from lecture transcripts.

    Aplora8 min read
    Read
  • A close-up of funding application forms, their fields still blank

    Ten manual roles down to zero: an anatomy of refinancing automation

    Applications for subsidies and grants were handled by ten coordinators by hand. A walk-through of the pipeline: a single source of operator rules, a personal PDF guide, daily monitoring of application windows and triggered communication.

    Aplora11 min read
    Read

Work out the order of magnitude yourself

Five calculations on your own numbers — no form, no data sent: everything runs in your browser. It is an estimate of potential on your assumptions, not a measured effect.

All tools

The rules this is written by

The same rules we publish case results by. You can check them in the text: every claim with a number carries its type and its source.
  • Fact, estimate and opinion are labelled explicitly and never passed off as one another
  • External data carries a source and a date — without them the claim is not published
  • No percentage without a measurement window and a stated method
  • Each piece is written for one reader's task, not for a search query
  • The English version is not a translation but a text in its own right
  • We don't publish for volume: four substantial pieces beat forty thin ones

Methodology notes, by e-mail

The same material we publish here: how to model the economics of an initiative, where rollouts break, and what to verify before work starts. Once a month at most, no market news and no sales e-mail.

By submitting you agree to the privacy policy — privacy