Case studies Craster
Craster: Gave a fast-growing in-house AI programme a shared language and a clear plan to scale safely.
- Sector
- Hospitality: luxury products for hotels and restaurants
- Size
- 55 people, with a head office in London and a US operation
- Engagement
- AI readiness audit, then ongoing support project by project
- Status
- Ongoing since May 2026
craster.com (opens in a new tab)
The starting point
Craster makes beautifully designed front-of-house display products for many of the world’s top hospitality providers, and they weren’t starting from scratch with AI. In about a year, a small in-house team with no dedicated software developers had automated most of the company’s order processing. Customer service queues had dropped from 50–80 tickets per person to 20–30. Leadership had big plans to go further, with AI opportunities mapped across every department.
The problem was that progress had outrun the structure around it. In Craster’s own words, they were “moving, but without an architect”. Decisions were being made tool by tool rather than by design. There was no agreed approach to who could use which AI tools, the AI policy was still in draft, and it wasn’t clear who owned which decisions. Every new tool brought more risk and more hesitation, and the small team doing all the building was already at full stretch.
They wanted someone independent to pressure-test the plan and tell them plainly which pieces mattered most, and in what order. As their own brief put it: strategy before tools.
What we did
- Reviewed Craster’s own strategy, systems and documents before setting foot in the building.
- Spent two days on site in London: ten sessions, from leadership down to walkthroughs of live workflows.
- Gave the team a shared language for AI work, from simple automation up to agents, and mapped what was really running.
- Delivered a board-level findings report on technology, data, people and the commercial case, with a list of quick wins.
- Stayed on as an adviser, picking up new projects across the business as they come up.
What changed
| Measure | Before | After | How we know |
|---|---|---|---|
| A shared language for AI | No common way to describe different kinds of AI work | A five-rung ladder, from basic automation to agents, used by the team and leadership | It appeared in Craster’s own audit documents within a week of the visit, and leadership proposed it as the basis for training and governance |
| A clear picture of what’s running | An ambitious map of 50+ AI opportunities, with ideas and live systems side by side | A clear line between what’s running today and what’s still an idea | Walkthroughs of live workflows on site, checked against the team’s own documents |
| What’s holding growth back | Not yet pinned down | Team capacity and structure, not technology, budget or willingness | Ten sessions across the business, backed up by follow-up conversations with sales and HR |
| Actions ready to go | Good governance instincts, mostly still in draft | Nine quick wins the team can act on without outside help | Listed in the findings report, June 2026 |
The biggest change was clarity. Leadership got an independent view confirming the foundations were strong: the team’s builds were plain, reliable and cheap to run, and the job ahead was about spreading that good work across more people rather than fixing poor work.
The team also saw many of their own ideas come back as the recommendations, such as a visible roadmap, access matched to training, and a written structure for the team. That made the plan easy to get behind. The question moved from whether to keep going to what to do next, in what order, and with what structure. Craster has kept us involved since, bringing new pieces of work to us as they arise.
“We were aware that our processes were outdated and required a lot of administration to run. With AI as a tool we built a lot ourselves, but we were moving without an architect. AgentifAI took the time to understand what our team had actually done, gave us a shared language for talking about it, and was straight with us about what needed to be in place before we go further.”