About the client
A professional services agency of around twenty people, billing a mix of retainers and fixed-scope projects. Delivery-led, with account managers owning client relationships end to end.
The challenge
The agency had solved every problem it ever had by buying another subscription. Sales lived in one tool, projects in a second, time tracking in a third, invoicing in a fourth, and the connective tissue between them was a set of spreadsheets maintained by whoever had time.
The cost showed up twice. First as licences: more than ten monthly subscriptions, most only partly used. Second, and far more expensively, as data silos — the same client existed in four systems under four slightly different names, so nobody could answer basic questions about account profitability without a manual reconciliation.
The practical effect was lost revenue. Hours went unbilled because tracking sat outside the tool where work happened, and scope crept because budget consumption wasn't visible until month end.
The solution
Discovery mapped the agency's real delivery flow end to end — how a lead becomes a proposal, a proposal becomes a project, a project becomes billable time, and billable time becomes an invoice. That map, not a product template, defined the schema.
What we built:
- One client record carrying pipeline, delivery history, logged time and invoices, replacing four partial copies across four systems.
- Deal-to-project continuity — projects inherit scope and budget from the deal that created them, so the number the client agreed to is the number the team delivers against.
- In-context time tracking, logged inside the task the person is already working on. That is what finally made timesheet data complete enough to bill from.
- Automated invoicing generated from approved time and milestones rather than reassembled by hand each month.
- Migrated history — records from the previous stack cleaned, deduplicated and loaded, so the system launched with the agency's past intact rather than empty.
What changed
The agency retired the bulk of its subscription stack and got a single system its team actually works inside rather than around. Account profitability became a screen instead of a reconciliation exercise, and unbilled hours stopped leaking because logging time no longer meant leaving the task.
Our AI workflows generated the architecture and the bulk of the module code, with senior engineers owning the business logic, the permission model and the data migration. That division of labour is what compressed roughly eight months of conventional development into four weeks.
The takeaway
Tool sprawl is rarely a tooling problem — it is the accumulated residue of solving one problem at a time. The fix isn't a better tool in the stack; it's a system that models the business, so the tenth subscription never becomes necessary.