The new platform is live, but operators still use messages and spreadsheets to finish the work. That is evidence to inspect, not proof that the software is useless or that the team needs another application.
Find the decisions the current workflow cannot support and why people route around it.
Trace an actual request
Start at intake and follow the work to completion. Record the required facts, owner, decision boundary, tools used, exceptions, and evidence that the next team accepted the handoff.
Separate a missing feature from a configuration problem, unclear authority, insufficient training, or unavailable capacity. Each finding can require a different change.
Inspect existing capabilities
CRMs, project systems, and agency platforms often include workflow features. Test whether the supported rules, assignments, triggers, and integrations can handle the specific scenario.
A better field or approved configuration can be the smallest useful fix. A custom operating layer is justified when a necessary decision or handoff cannot be represented reliably with the current tools.
Define the missing connection
For a client follow-up, the gap may be that a recorded commitment has no owner or next date. For a cross-system request, the receiving team may lack a required input or durable record.
Specify the trigger, source authority, permitted action, receiving owner, failure behavior, and acceptance evidence. Test the exception path before calling the workflow complete.
Keep model output reviewable
A model can help draft or summarize permitted context. It does not supply the authority to act or resolve contradictory source facts. Put the draft where the responsible person can check it and record the authorized action.
RachelOS illustrates a practical operating layer around relationship work. The synthetic healthcare lab illustrates separate cross-system authorities. Neither implies that every business needs the same software.
A Constraint Diagnostic identifies the specific gap and first build. When the client-inquiry path is already clear, explore the Owned Demand & Governed AI Engine. For an approved payer obligation or platform transition, use the Readiness Sprint.