245 Hours Reclaimed Every Week. A Legal Operations Platform Built in Four Months.
In four months, we built an AI-accelerated legal operations platform designed to free seven paralegals from repetitive setup work and replaced developer-managed legacy rules with tools managers could configure themselves.
Building the foundations for AI-assisted legal pack creation
As lead volume increased, the legal pack department were overwhelmed by repetitive administration and installed a 'Set Up' team to manage weekly allocation. The team consisted of seven paralegals, equivalent to 6.5 FTE. Their legal and operational knowledge was being used to manually prepare work for other teams.
Every file required people to order searches, review title information, create tasks for the paralegal that they assigned the file to. That represented approximately 245 operational hours every week that could instead be used to create and progress auction packs.
Context
Legal packs are critical to the property sale journey and the business promise of selling at speed, but growing lead volumes turned the process into a major operational bottleneck.
Problem
Legal pack creation relied on fragmented manual workflows that averaged about 15-20 minutes per file to set up and allocate to the appropriate paralegal.
Goal
Design an operational platform that automates allocation, improves task management and frees up seven paralegals to assist with the growing volume of leads.
20+ process maps were drawn up over the course of the project, spanning the current process and 'To be' process, including continuous updates as our agile approach required flexibility for change.
A working group was set up to gather input from across the business, surfacing important actions and decisions for wider review.
A process held together by people, not systems
I combined stakeholder interviews, cross-functional workshops, process mapping and repeated design playback sessions. Discovery revealed that setup was not one consistent process. It relied on knowledge held by experienced paralegals that took months to train and was costly to scale to match the increase in workload.
It also exposed the fragility of the legacy system and limitations in the legal AI’s training data. These findings changed the direction of the initiative from reproducing the existing workflow to building a configurable platform with explicit failure routes, human review and controls that managers could adjust themselves.
To match the speed of the developers, I explained ideas and refined prompts to produce AI generated high fidelity designs. They weren't quite right though, so I used these to help shape ideas faster.
After a few tweaks, I would update the designs to match our design system so our front end developer could build even faster with ready-built components.
Four months to build what similar projects had taken more than a year
Previous legal technology projects involving AI had taken more than a year to deliver. Given a four month timeline for design and build, we trialed an agile approach that included use of AI to accelerate our process. AI helped us compress the journey from an idea to working software.
ChatGPT and Codex helped accelerate discovery, requirements, design exploration, development and iteration. An agile approach meant we could quickly turn operational problems into rules and build them into working software.
Early stage development of process maps, detailing different users' steps and system triggers. Multiple workshops went into shaping the final processes and requirements gathering.
The payment pathways were previously unmapped and confusing to explain. Distilling down to a simple flow helped break barriers in understanding.
Search ordering no longer needed someone to keep it moving
When a new file arrives from the legacy system, the new platform automatically contacts the Land Registry and places the required orders. It then monitors each order, processes returned documents in seconds and tags them appropriately.
Documents are kept 'Internal only' until a team member has reviewed and redacted them. Once approved, they are made public and added to the live legal pack.
We built tools the business could control
Allocation was designed to put control in the hands of operational managers, rather than relying on developers for every change. Managers can build visual decision flows, set conditions, review results and decide to which team each file should go.
Our legal AI reads title registers and populates the task list
Once the Title Register is returned, its uploaded to our internally developed legal AI to identify proprietors, restrictions and other relevant title information. Those findings generate the tasks required for that specific file, finishing the set up process so files are allocated already prepared. No longer taking 15-20 minutes to set up one file, multiple files can be set up at once in less than a minute.
Conclusion
The result is more than 245 hours of additional capacity. The longer term value is the platform’s ability to keep changing. It doesn't simply automate the process as it exists today. It gives the business a cleaner foundation and the tools to adapt allocation, teams and workflows without returning to the same technical bottleneck.
And in the meantime:
Capacity: seven people released into pack creation.
Development speed: a legal and AI-enabled platform built in four months rather than the previous internal benchmark of 12+ months.
Adaptability: operational managers can change workflows themselves instead of waiting weeks for a developer.