The Challenge
In Hire Space’s scaleup, they faced a bottleneck. For every received inquiry, a venue expert would have to read it, understand and extract the needs of the client and then manually identify the best venues for the prospect client. This process was slow, repetitive and inefficient.
The speed problem was critical: in sales, the faster you answer, the higher the probability of closing the deal. Venue bookings are no different.
Current Stats: In 2025, Hire Space receives over 40,000 enquiries per year, with an enquiry value per month of £20M. That’s an average of £6k per enquiry.
If Hire Space took hours to respond while competitors replied in minutes, they’d lose the booking - regardless of having better venues. Every hour of delay means lost revenue.
We were approached by Will Swannell, co-founder and CEO of the company to identify ways data and AI could be used to facilitate this process.
Will didn’t want to replace their venue experts. Instead he wanted to empower them with tools that would make their job easier, and faster, and that would allow Hire Space to expand beyond London.
The Technical Problem
We needed to go from text inquiries to venue recommendations, in a way that was quick, consistent and scalable.
These are obviously two different spaces. While venues can have text descriptions, it’s not as simple as checking which venue description is the closest to the client’s inquiry. This was before the current generation of LLMs, which limited our technical options.
As we said before, this also needed to run as fast as possible. We couldn’t have a complex model going through every single venue and checking for how well it fits against every single inquiry.
Our Solution
We began by working with Will and his team to understand the complete business challenge. What exactly made a venue “right” for a specific inquiry? What factors did experts consider? We formalized these implicit business rules into technical requirements - a critical translation that many overlook.
The data existed but was fragmented across SQL databases, MongoDB collections, and other systems. Our first major contribution was creating a unified view of this data, giving Hire Space clarity on what they actually had.
We knew we had to build something custom. This was before modern LLMs, and existing NLP tools weren’t sophisticated enough for this specific matching problem. And it needed to be fast.
We designed a two-stage architecture:
- A custom pre-filtering model that quickly eliminated 90% of unsuitable venues based on key criteria
- A neural network that scored the remaining venues by matching extracted features from both the inquiry text and venue descriptions
This wasn’t just a technical choice - it was a business decision. Running neural networks against every venue would have been prohibitively slow and expensive. The pre-filter ensured the system could scale.
We then entered an iterative refinement phase. In weekly strategy sessions with Will and his data scientists, we analyzed performance, identified bottlenecks, and continuously improved the system.
Our Results
Together with Hire Space’s leadership, we defined a performance metric: “Can the model predict which venue will be selected for a given inquiry?” In other words, given an inquiry, how often is the model able to pick the venue that was actually selected by the prospect?
Not only that, we tested this against human venue experts to see how well the model performed.
AI achieved 254% better accuracy than human experts alone.
With over 40,000 enquiries per year worth £6k each, even a 1% improvement in conversion rate represents £2.4M in additional annual revenue. Our system delivered far more than that.
Strategic Impact
This wasn’t just about automation - it was about building a proprietary advantage. While competitors were still manually matching venues, Hire Space could now:
- Respond to inquiries in minutes, not hours
- Maintain expert-level quality at scale
- Expand to new markets without proportional hiring
The custom architecture mattered. Off-the-shelf solutions couldn’t handle their unique matching requirements. By building something tailored to their business model, we created a system that competitors couldn’t replicate by simply buying software.