An operations platform for a creative agency: from scattered requests to a six-week pilot

The situation

A regional creative and marketing agency working across Iraq, Egypt and Dubai had grown faster than its operating model. Client requests arrived by WhatsApp, email and phone. Schedules lived in spreadsheets that only one or two people understood. Nobody could see who was overloaded until a deadline had already slipped.

The agency had looked at off-the-shelf project tools, but every one of them assumed a workflow the agency did not actually follow. The real question was not which tool to buy. It was how work really moved from a client request to a delivered piece, and where it was getting stuck.

What I did

I started by mapping the real operating flow, interviewing the people who did the work rather than only the managers, and comparing what I found against established international agency practice. That produced a future-state model the agency could agree on before any software was built.

Then I designed and built the platform around that model.

• Intake: every request becomes one record, however it arrives
• AI-assisted briefs: the brief is drafted from the request, and the AI proposes deliverables from the agency’s deliverables catalogue; a person completes and approves both
• Automated workflow end to end: once an item is approved it is forwarded automatically to the people who need it for their own tasks, with no manual hand-off
• Resource-first scheduling: capacity is checked before work is committed, not after, with SLAs and time tracking against each job
• Document control: automatic versioning, filing and archiving of every deliverable, so the current version is never in doubt
• Role-based access control and a full audit trail across the platform
• Approvals, production workflow, resource tracking and finance export in one place
• An exceptions dashboard: alerts focus attention on the issues that threaten delivery or a client relationship
• A non-blocking model: warnings for impossible deadlines, weekends and holidays let people proceed, and every override is logged

The result

A working pilot was running with the Client Services and Creative teams about six weeks after mapping began. Teams create and schedule projects through the new workflow, with AI assisting the routine parts and people keeping the decisions.

The platform was built multi-tenant and role-based from the start, so it can serve other agencies without being rebuilt.

The visualisation on this page shows one selected workflow, a request moving through intake, brief, capacity check and approval, and does not represent the complete platform.

Built with

React and TypeScript; Supabase (PostgreSQL with row-level security, Edge Functions, Realtime, Storage); database tests, unit and end-to-end tests with CI; Vercel. English and Arabic.