Skip to content

I don't think your business is using AI correctly. Want to know why?

I train your team on the work they should own, build the parts where being wrong is expensive, and tell you what to leave alone.

Photo of Oliver
Drop a real photo at /public/oliver.jpg and set founder.photo.enabled to true.

What I've noticed

Three mistakes I see again and again and again

I help lead AI adoption for over 2,000 people at Macquarie Group and have seen what it costs.

  • You bought the tool before you sorted the work

    There are no perfect tools, only less-bad ones. You focused on a tool instead of the outcomes you want to achieve.

  • Your team is using AI like a chatbot

    They hand off work manually and repeat the same instructions to an agent that forgets everything each chat.

  • You are trying to fix it all at once

    AI makes it really easy to build just about anything, and that becomes a license to do just that.

The problem

Most AI spending produces nothing, and the tools are not what let you down

89%

of firms report no measurable productivity impact from AI, three years in.

NBER, 2026

85%

of organisations underestimate what their AI costs. Nearly a quarter are out by more than half.

Benchmarkit & Mavvrik, 2025

$186

per employee, per month: the cost of rework on AI-generated work a colleague has to unpick.

BetterUp Labs & Stanford, HBR 2025

What I sell

Building software is close to free now. Judgement about what to build is the scarce part.

Anyone can build you a chatbot. I will also tell you when not to build one.

I have watched more than $10,000 go in a few days on routine daily analysis: a model far bigger than the job needed, running work nobody had scripted. Nothing about it malfunctioned.

The triage

I sort your work into three buckets

The axis is consequence: how expensive it is to be wrong, not how tedious the work feels.

  1. 01

    Contained, low consequence

    Your team owns it.

    That is exactly why I train them first.

  2. 02

    High consequence

    I own it.

    Outcome focused to ensure maximum value.

  3. 03

    Not worth automating

    Left alone, on purpose.

    This should be the largest bucket to begin with.

How the triage works in practice

The training

Bucket one is yours. I make your team fluent enough to hold it.

One full day, up to 20 people, on your tools and your own work. $10,000 per workshop.

  • How the models behave

    Why context decides what you get, and why prompt tricks age out while the reasoning underneath keeps working.

  • Your own tools, your own work

    Claude, ChatGPT or Gemini, used on the work your team already has in front of them.

  • The anti-patterns

    The answer that sounds right and isn't, and the polished draft someone else has to unpick.

  • Every tier in the room

    Executives and analysts in the same room on the same day, because adoption that only reaches one tier stalls.

What the day covers

The automation

Bucket two is mine. I own the outcome, not just the build.

The triage decides which workflows land here. What I am judged on afterwards is what they produce. Priced on that outcome, not on hours, and quoted once the scope is set.

I helped build a governance model for non-deterministic tools to the bar APRA set for the entities it regulates in April 2026. If you are not regulated, it loosens to fit.

  • You can point at what changed

    Measured against how the work ran before, so what you got out of it is a number rather than a feeling.

  • Proven before you depend on it

    You watch it run against your own cases and sign off on the evidence, not on my word for it.

  • You pay for the result

    Priced on the outcome I am responsible for, so finding a faster way to reach it is on me.

  • It still works in six months

    Your processes change and the models change under them. Holding the outcome steady through both is the job.

How I prove it before you rely on it

Who this is for

Big enough that getting AI wrong is expensive. Too small to hire someone whose whole job is getting it right.

Worth a call if

  • You pay for AI tools and can't point to what changed.
  • A confident wrong answer would cost you real money.
  • A pilot started well, then quietly stopped.

Probably not, if

  • You want the cheapest possible chatbot.
  • You've already decided what to automate and want it validated.
  • Every process still changes weekly.

Who you would be working with

What it costs

One price published, one quoted once I have seen the work

Foundations

$10,000per workshop

One day, up to 20 people.

For your team to get immediate value.

Automation

Scoped to the work

Outcome based pricing.

For the work you can't manage and maintain.

What’s included in each

Next step: one 30-minute call

Book a free 30-minute call

There's no deck and no obligation. You leave with my read on where AI is worth your money, including "you don't need this yet" if that's the truth.