Workflow AI · 2025
02
Meridian
Automation
The brief
Automate the work.
Free the people.
Meridian's 200-person operations team was spending 40% of their working week on repetitive manual tasks — routing requests, chasing approvals, re-entering data between systems. The processes existed because the technology didn't support automation. We were brought in to change that.
Over 8 weeks we mapped, redesigned, and automated their core operational workflows — connecting legacy systems through a new API layer and embedding AI decision-making at the points where it could have the most impact.
Challenge
40% of time
lost to admin.
Every client request touched four systems in sequence. Approvals were email chains. Escalation paths were tribal knowledge. New starters took weeks to learn the routing rules that existed only in senior colleagues' heads. The business was scaling its headcount to manage process complexity rather than to create value.
Every inbound request required a human to classify, prioritise, and assign it — a bottleneck that grew with volume.
Core systems had no APIs — data had to be re-entered manually at each handoff, introducing errors and delays.
Managers had no real-time view of where requests were in the process, making SLA tracking impossible.
Solution
AI routes.
Systems talk.
We built an orchestration layer connecting all five core systems through a unified API gateway. An AI classification model routes incoming requests with 94% accuracy. Approval workflows are automated with smart escalation rules. A live dashboard gives operations leadership full real-time visibility for the first time.
Fine-tuned model classifying and routing 12 request types with 94% accuracy — replacing 3 FTEs of routing work.
Custom connectors bridging 5 legacy systems — eliminating all manual data re-entry across the end-to-end process.
Real-time workflow visibility with SLA tracking, bottleneck detection, and automated escalation alerts.
Less manual
processing
Throughput
increase
Annual
savings
From brief
to live