AI experiments on real work

Your AI edge, tested not borrowed.

Build your unfair advantage at work. Experiment your way to AI that actually works for you.

For individuals Make AI work for you Test AI on your real work and build an approach that fits your context, standards and judgement. See the experiments → For businesses Find where AI can create value Identify the opportunity, prove what works and build only what pays back. Explore AI consulting →
A decade of experiments for NatWest Vitality RNLI
Portrait of Steve Quinlan
10+ years
of experiments, keeping only what works.
Built, not copied
Steve feeding one real task into a testing machine. A gauge measures it, and it comes out as a green-lit checkmark and an upward trend.
Tested, not borrowed
The belief

Stop following. Start testing.

The best way to get more from AI is to start with your own work. Take one real task, test an approach, and shape it around your context, standards and judgement. Keep what helps, adapt what does not and build on each proven gain. You can learn from other people’s prompts and tools without being limited by them. The result is a way of working with AI that fits you, and an advantage nobody else can simply copy.

The experiments

Real AI experiments on real work.

Each experiment compares an AI approach with how the work gets done now. I measure the difference, show the evidence and publish the honest verdict. That published work lives in the Experiments hub.

Browse all experiments →
Latest experiment · six runs One task · one model · one day

A 733-word prompt scored 52.5. An 87-word brief scored 100.

Prescribed route Binned
52.5 out of 100
733 words, 20 prescribed steps. The prompt told the model which analysis to run, and that analysis excluded the problem in the evidence. All three runs recommended the wrong fix.
vs
Open route Kept
100 out of 100
87 words, no prescribed steps. The brief defined the outcome, evidence, constraints and checks, and left the route open. All three runs reached the supported decision.
VerdictBrief the destination, evidence, constraints and checks. Do not prescribe the route before the model has seen what the evidence requires.
See all six runs and the scoring →
The method

The Edge Loop: five steps, one proven gain at a time.

It turns scattered AI use into proven improvements in how you work. Start with one real task. Prove a better way. Then stack the next win.

Use the Edge Loop →
  1. 01
    Pick
    One real task from your work.
  2. 02
    Test
    An AI approach against how you do it today.
  3. 03
    Measure
    The difference in time, quality or output.
  4. 04
    Keep
    The approach only if it works for you.
  5. 05
    Stack
    The win into your workflow, then run the loop again.
The Experiment Log · fortnightly

Get the next experiment, not another list of tools.

Every fortnight I send one real AI experiment, plus one focused test you can run on your own work.

01
One real test
A recognisable task, worth doing better.
02
The evidence
The number, and enough detail to check it.
03
An honest verdict
Kept or binned, with the limits stated.
04
Something to try
One focused test you can run on Monday.
The evidence

I run the experiments. You see the numbers.

At a UK retail bank I designed an AI-supported system for GEO and code reviews. Five hours across three roles became 20 minutes with one engineer.

93%
less time on every review
486 hrs
saved a year, two reviews a week
£33,150
saved a year on the same work
A decade of this before AI

+28% donation journey conversion at RNLI. +14% health quote leads at Vitality. +162% year-on-year colour consultant bookings at Farrow & Ball. Building production AI inside a UK bank since 2023.

Who is behind it

Your AI advantage has to be yours.

I’m Steve Quinlan. I help you build an AI advantage around your work, not borrowed from a guru or copied from a prompt pack. A decade of experiments in product and conversion taught me the rule that still holds. Test it on your own work, keep what wins.

More about me →
Product and conversion experiments since 2012. Building with AI since 2022.
For businesses · AI Consulting

AI consulting that turns scattered tools into systems that pay back.

Your team is already using AI. The same discipline works at firm level: find the work worth changing, test the smallest useful approach, measure the result and build only what earns its place. No AI initiative earns the next pound until it has proved a result.

See how I help firms →
Start here · free to try We test one journey on your website The AI Mystery Shopper For firms with a customer journey that matters: a quote, a booking, a purchase. This is mystery shopping your website with an AI agent. The test shows whether it reached the correct outcome, where it stalled and what to fix first. Free snapshot one agent attempt on your own site, back within two business days Then a verdict from paid testing I have reviewed and signed Start with the free snapshot →
Primary engagement 3 to 4 weeks The AI Opportunity Roadmap Interviews, workflow mapping and costed bottlenecks, ranked into a sequenced 90-day plan built for a board decision. Explore the AI Opportunity Roadmap → A bounded session Practical AI Workshops Your real work, not prompt demos. You leave with two or three opportunities and an owner for the first test. A specific decision Focused advisory Pressure-test a proposal or value case. If there is a better route than bringing me in, I will say so.
Direct reply from me within two business days. No sales sequence.
Where to next

Choose where to go next.

01 Experiments The published work: tests, tools and build notes. 02 The Edge Loop Run the method on your own work, one gain at a time. 03 About Why experimentation became the thread through my work. 04 AI Consulting Bring the same discipline into your business.