Optimizing Salesforce CRM Workflows for Enterprise
Lessons from auditing 870+ work orders at government scale: where CRM data actually breaks, and the automation that keeps it audit-ready.
by Yassine Emhamed // field notes
I run Salesforce data operations for DC's public buildings — 900+ facilities, every work order documented, approved, and audit-ready, because the output feeds municipal funding decisions. In one audit cycle I cleaned 870+ work orders end to end. Before that, I kept enterprise CRM data accurate for high-volume hospitality clients at a SaaS company in a zero-downtime environment. Different worlds, same conclusion: CRM systems don't fail loudly. They rot quietly.
02 // Data doesn't break at entry — it breaks at handoff
Almost none of the 870+ records I remediated were wrong because someone typed badly. They were wrong because a work order crossed a boundary — field tech to supervisor, department to department, system to system — and the workflow logic at that boundary was broken or ambiguous. An invoice attached but never linked. A field-service photo uploaded to the wrong stage. A multi-tier approval that skipped a tier because the routing rule predated a re-org.
So the audit method that actually works is boundary-first: map every handoff in the record's lifecycle, then check the records that crossed each one. Auditing by record is archaeology; auditing by boundary is engineering.
03 // Make the report the byproduct, not the project
In both government and enterprise SaaS, the most expensive ritual I've seen is the hand-built status report. At Harri I produced operational reports leadership actually used — platform health, adoption, failure points — and the only reason they stayed sustainable was that the data collection was automated underneath; the report was just a view. At DGS, compliance reporting runs through Salesforce Lightning, SharePoint, and Power BI on the same principle.
If a human assembles the numbers, the numbers are already stale and the human is already tired of it. Instrument the workflow once; let every report become a query.
04 // 'Mostly accurate' is not a standard
When CRM data feeds funding allocation, a bad record isn't a typo — it's a policy incident. That stake changes how you build: validation at the moment of capture instead of cleanup campaigns later, documentation attached to the record instead of living in inboxes, and workflow fixes engineered with the stakeholders who own the boundary, not patched around them.
The discipline transfers directly to the private sector. The dental clinic, the law firm, the real-estate team I've automated — none of them needed 'more CRM.' They needed their records to survive an audit by their own future selves.
Audit by boundary, not by record. Automate the collection so reports are queries. And hold business data to the standard government taught me: if it can't survive an audit, it isn't data — it's a guess.