CASE STUDIES
See how it works in practice.
Five recent projects. Different businesses, different problems, same approach.
How we helped a UK gym operator get the foundations right before scaling AI
A large UK gym operator, marketing around 50 new gym openings and 30-plus refurbishments a year across a growing UK estate.
What started as a scrappy challenger is now a maturing business – private equity backed and pushing hard for growth. The tools and ways of working that suited a smaller, simpler company no longer fit that scale or pace.
AI literacy across the team was low, so nobody had a clear view of where automation could actually help.
We started with the work. We spoke to people right across the marketing team – local marketing, CRM, planning and studio – and followed how information actually moved through the business.
We found a fragmented operation. Lots of trackers, several tools, and constant manual copying between them. One tracker was so complex that a specialist data agency called it among the most complicated they’d ever seen. No single trusted record for each gym, so everything downstream inherited the mess.
To make the fix concrete, we mapped 29 data requirements across 11 themes.
The team came away with a clear picture of what to fix and in what order, and a shared language for raising the gym database dependency with senior stakeholders.
The work laid the groundwork for a broader engagement – from standalone training and tooling through to an embedded model spanning marketing and commercial.
The team had been treating most of this as an AI problem. The real fix was automation and cleaner data.
Underneath almost everything sat one thing: a single, trusted record for each gym. Without it, the high-value automation stays blocked. And fixing it needs senior sponsorship from outside marketing, because the data sits beyond the team’s control.
How we trained 300 marketers on AI using a brand that doesn’t exist
A global FMCG business operating in a heavily regulated category. Around 300 marketers spread across every region, from people using AI every day to people who had never used it.
Copilot was available to everyone. Earlier AI training had been pitched at either the basics of how to use it (IT focused) or around what users couldn’t do (compliance focused). Our training was marketing oriented – helping them to see directly how AI tools can be applied to their everyday work.
We built the whole programme around Halo, a made-up 0.5% ABV craft lager. Behind it we created a full set of files: sales data, brand health tracking, a consumer research debrief, a category read, a focus group transcript. Realistic, messy, and safe to get wrong.
Marketers worked Halo through the same stages they face on a real brand – diagnosis, positioning, campaign, innovation pipeline – using the company’s own brand strategy framework. The AI work sat inside the process they already use.
Four 90-minute live sessions, each run twice a day to cover time zones.
We based the teaching on two simple frameworks:
- CRAFT – how to brief an AI chatbot so it gives you a decent output
- FACT – how to check what comes back before you put your name to it
We ran every prompt in advance. We knew exactly where it would invent a number, merge two unrelated tables, or report a growth rate as a share point change. Those failures were part of the session on purpose. Marketers need to see the tool get it wrong on data they understand.
Everyone left with a prompt pack and a guide to which tool does which job.
Going in, around one in eight marketers rated their confidence at 1 or 2 out of 5. Coming out, nobody did.
That satisfaction score is important. Nothing kills effective AI training like people leaving sessions unconvinced.
Participants were eager to get the prompt pack, which is a better signal than any score – it means they had a job waiting for it the next day.
The measurement also told us what to build next. We are now planning for the following year’s programme.
Most of the audience had a basic Copilot licence, which is far more restrictive than most AI chatbots, even the free consumer versions. We were careful not to teach above that ceiling in these sessions – we train for the licence people have.
But, since the two most requested capabilities by the audience were undeliverable inside the existing toolset, we are now working with the client to build the business case internally to upgrade access and licences.
“Absolutely fantastic course. Hyper relevant, clearly communicated by the presenter and using a great brand campaign example. I look forward to practising using the tools that will be provided. Thank you!”
Session participant, rated the session 5 out of 5
How we helped a UK Plc unlock AI across a 240,000-asset creative operation
A large UK gaming company producing more than 240,000 marketing assets a year across multiple brands and channels. The business was shifting to a more automated and personalised marketing model. The in-house studio needed to increase output by 2–4x without sacrificing quality or burning out the team.
Previous advisors had spent months exploring tools and platforms without getting any closer to a workable solution.
We started with the work. We mapped how briefs, files, feedback, and approvals moved through the studio by talking directly to the people doing it.