Turning booking gaps into revenue
Appointment slots were going unfilled, not for lack of customers but for lack of insight into which customer to offer them to.
The problem
The business needed to analyse client history, service preferences, recency and booking patterns close to real time, and turn that into a short list of who to contact when a slot opened. The existing logic was rules based, which meant it was easy to reason about and poor at ranking.
The second requirement mattered as much as the first: whatever got built had to be a platform the company could keep building AI on, not a single-purpose bolt-on. This client is not named at their request.
What we did
- Replaced rules with behavioural ranking, scoring customers on real booking history and preference signals rather than fixed conditions.
- Designed a Databricks and Kafka architecture over the existing SQL Server estate, so nothing had to be ripped out to get started.
- Set an hourly refresh cycle, which is frequent enough to be useful for same-day gaps and cheap enough to run continuously.
- Framed it as phase one of a modernisation roadmap, with the later phases named and sequenced.
Where it landed
- A first use case tied directly to revenue, not to efficiency
- Databricks and Kafka architecture over the existing SQL Server estate
- Recommendation logic grounded in real customer behaviour
- A platform that carries the phases that follow
Our view
The real challenge here was not the AI. It was picking a first use case that pays for itself quickly and still leaves you with a platform worth building on. Most first projects manage one or the other.
Sitting on customer data you are not using?
Revenue use cases are usually the easiest first projects to get funded.