Skip to content
AI Metric

Case studies

Results, not slideware

We publish engagements only once real results are approved for release. We never invent client quotes or figures. Until the first named study clears approval, here is exactly how the work runs, which is the part most people want to know anyway.

Why this page has no logos on it

Most of what you would read here would be invented

It takes about ten minutes to write a convincing case study with a plausible percentage and an anonymous client. Plenty of sites in this market have done exactly that. We would rather show an empty shelf and be trusted on the day we do publish a number. Live engagements are being written up now and go here once the client has approved the wording and the figures. If you want the detail before then, ask on a call and we will walk you through the work in the open.

What an engagement actually looks like

Six weeks, one workflow, one measurement agreed before anything is built. Nothing about this shape depends on client approval, so it can be published in full.

  1. Week 0

    Discovery, paid and short

    A half day on your side of the desk: which documents get re-keyed, which reports get written twice, where the delivery team loses hours that nobody has ever counted. You get a written findings note naming the three workflows worth automating and the ones that are not worth touching. That note is yours whether or not you go further.

  2. Weeks 1 to 6

    One pilot, one measured workflow

    We build against a single workflow rather than a platform. Six weeks, a defined output, and a measurement agreed at the start: hours per week, error rate, turnaround time, whichever one your business actually feels. Small scope is deliberate. A pilot that touches everything proves nothing and never finishes.

  3. Week 6

    The number, honestly

    We measure the same thing we measured at the start and tell you what changed, including when it is less than we hoped. If the pilot did not earn its place, we say so and you stop. A consultancy that has never reported a disappointing pilot is not measuring.

  4. After

    You own it, running or not

    The workflow, the prompts, the configuration and the documentation are handed over and are yours. No lock-in to us and no licence you cannot leave. If you want us to keep running it, that is a separate, cancellable arrangement.

The rules we work to

We never invent a figure

No fabricated quotes, no invented percentages, no anonymous logos implying clients we do not have. Anything with a number attached is published only when the client has approved that number.

Small scope, measured

Every engagement targets one workflow with one agreed measurement. It is the only way to know whether AI helped or whether the team simply worked harder that month.

Construction first, AI second

The failure mode in this market is a general AI consultancy learning construction on your job. The sequence here is the other way round: twenty years of Tier 1 delivery first, the models second.

Tool neutral

We are not a reseller and take no vendor commission. The recommendation is whatever fits the workflow, including the answer that a spreadsheet and a better process beat any model.

Want to hear about live work before it is published?

Book a call and we will talk through current engagements in the detail client confidentiality allows, including what has not worked. If you would rather see finished output than hear about it, the product we built for ourselves is running publicly: Construction Metric turns a site WhatsApp group into an evidence-grade project record, from £19 a site a month.

Book a call