MCA’s consultants help global brands understand how much their advertising campaigns should cost to produce.
They analyse production estimates from around the world to identify strategic savings for clients.
But those estimates arrive in widely varying PDF formats, so instead of analysing, consultants were spending most of their time locating and manually inputting data.
MCA believed AI could remove the bottleneck. But could it be done safely and reliably?
Challenge
At the heart of MCA’s service is Control Room, a proprietary pricing and budgeting tool that turns production data into strategic cost intelligence.
Consultants were spending way too much time working through estimate documents line by line, manually identifying and entering the information they needed.
Time pressure meant some source documents would often be left out, useful data missed or entered inconsistently, fundamentally weakening MCA’s strategic analysis.
CCQ was asked to build an AI process that could do the heavy lifting, while ensuring it could be trusted to handle a real client workflow.
Solution
Rather than trusting the AI to get everything right, we designed the process around the fact that sometimes it wouldn’t.
Machine validation: We taught the system to recognise different document types, formats and variations, and we built in validation so AI-generated data couldn’t simply flow unchecked into Control Room.
Human judgment: Where the AI is uncertain, the workflow flags it for human review and shows consultants where the information came from.
Failsafe rollback: And an audit system would provide a rollback option, recording any changes, who made them and their source – making it easy to safely roll changes back without disturbing the rest of the estimate.
Technologies
AWS Bedrock
Anthropic Claude
TypeScript
Fastify
React
PostgreSQL
Results
The system delivered a 70% reduction in the time consultants spent preparing production estimates for analysis.
That means they spend less time finding and transferring information and more time applying their expertise.
Richer analysis: Where consultants would typically only work from one to three source documents before, the new system processes up to 20 per estimate - without multiplying the workload.
Improved consistency: Automating the initial data capture also creates greater consistency. Standard information is captured systematically rather than relying on individual users to enter every field manually, creating a stronger data foundation for MCA’s benchmarking and analysis.
Control: And with human review, source tracking and rollback built in, those efficiency gains haven’t come at the expense of control.