What working with us produces
Working ways of working in ChatGPT or Claude, people who use them, and an owner inside the business who keeps them going. Here's how it went for three businesses.
Thunderbird Press
Commercial printing, Richmond, BC. Family-run since 1967.
Thunderbird Press prints labels: the ones on paint cans, spice jars, and nut butter, for manufacturers, food brands, and marketing teams across North America. They print catalogues and packaging runs too, and they've been doing it as a family business since 1967.
Orders used to arrive by email. Each job was a thread: the specs in one message, the artwork in another, a change in a third. Someone had to read all of it, work out what was actually being ordered, and pass it down the line. Which jobs were pending, and how far along each one was, lived in inboxes and in people's heads.
We put Claude in the business, with the CEO and the production team. It started with an AI Opportunity Assessment to find where AI would genuinely help, then an AI Action Plan we built with them and delivered together, and we've stayed close through ongoing advisory support.
Built with it: an order intake on the website that captures each order as complete, structured information and routes it to the right person, and a job system the team uses to see what's pending, follow progress, and track inventory.
What changed: orders come in complete, so the back-and-forth emails are gone. The system processes each order and assigns it to the right person, and anyone on the team can see where any job is without asking.
Today the CEO and the production team use Claude every day: to get work done, keep the website current, track inventory, and follow jobs through production. They maintain all of it themselves, and call us when they need specific technical support.
Skin Precision Medical
Skin clinic, Blackburn.
Skin Precision Medical is a skin clinic in Blackburn. It opened as a new business, so the first thing it needed was a website, and the founder built it with us using an AI build tool rather than having it built for them.
In the early days, bookings came through Instagram and WhatsApp messages, handled one at a time. That was right for a new clinic getting to know its first clients, and wrong once it became the thing eating the founder's day. So the founder used the same tool to set up a booking system on the website.
Today the founder runs the website, the bookings, and the clinic's social media content plan alone, with the tool we set up together.
Dripping Hive
100% British honey.
Dripping Hive sells 100% British honey. Until this summer, the only place to buy it was a market or a fair, and nothing came in between them: no orders, no enquiries.
We set the team up with Codex and generated the website with it: a shop, an order and enquiry flow connected to WhatsApp so customers can order directly, and the privacy and consent pieces a shop needs. The site runs on hosting they own, with a catalogue they own.
What changed: people can find Dripping Hive, and buy from them, between markets. Orders and questions come in through the site and WhatsApp.
Today the team uses Codex for their marketing, their social media content, and the website itself. They wanted to be known. They now have a website and an Instagram page that do that.
How we measure
For every workflow we deliver, we agree the metric and the baseline before we build, then track assisted time, human checking, rework, adoption, interventions, and incidents alongside the business outcome over an agreed period. The workflow owner inside your team owns the measurement with us. That's why, when we do publish a number, it comes with how it was measured.
Results depend on the workflow, starting point, adoption, review requirements, and client-controlled conditions. Numerical estimates are not guarantees. Each published result is identified as measured, client-reported, or modeled, with its measurement period and material assumptions.