The software almost never breaks. The build does. After enough rescue projects, the failure pattern becomes predictable — and so does the fix.
Industry research has put CRM disappointment as high as 70% for years. Leaders hear that number and assume the platform is the risk. It isn't. Salesforce will do almost anything you ask of it. The risk is in what you ask, and whether the foundation can carry it.
Check 1 — Is the data clean before it moves?
Most stalled rollouts inherit a mess: duplicate accounts, half-filled fields, files that never made the jump. Automation and AI then amplify the mess at scale. Run a data audit and a de-duplication pass before migration, not after the complaints start.
- Profile every object for completeness and duplicates
- Define the source of truth per field before you map
- Migrate files with an audit trail, not a best-effort export
This is exactly why we built Relay — a managed package that stages every file through your own S3 bucket with a full audit log, so you can prove nothing was lost.
Check 2 — Was it built for the rep at 4pm?
An org designed around a process diagram looks great in a demo and dies in the field. If the person actually entering data needs six clicks and a training doc, adoption collapses — and with it, every downstream report.
If usage isn't designed in, ROI is designed out. Adoption is an architecture decision, not a training afterthought.
Role-based layout design, clean onboarding flows, and minimising required fields aren't polish — they're the core deliverable. Every sprint review should include someone from the team who'll actually use it daily.
Check 3 — Is the foundation AI-ready?
Agentforce and autonomous workflows are only as good as the model beneath them. Over-customised, KPI-less orgs can't carry agents safely. Build clean, governed, and measurable from day one and the AI layer becomes an upgrade — not a rebuild.
- Every automation has a named KPI it moves
- Data model is documented and ownership is clear
- Einstein Trust Layer settings are configured before agents go live
None of these checks need a single license to run. That's the point: the cheapest time to prevent a stalled rollout is before it starts. A one-day pre-mortem before you sign the SOW is worth three months of a rescue project.
Frequently Asked Questions
Why do most Salesforce implementations fail?
Industry research puts CRM disappointment as high as 70%. The failure is rarely the platform — it's the build. The three root causes are dirty data inherited at migration, an org designed for diagrams rather than the people using it daily, and a foundation that can't support AI and automation. All three are preventable with a pre-mortem before the project starts.
How do I prepare data before a Salesforce migration?
Before migrating, profile every object for completeness and duplicates, define the source of truth per field, and use a tool with a verified audit trail rather than a best-effort export. Relay — QuickBild's file migration package — stages every file through your own AWS S3 bucket so you can prove nothing was lost in transit.
What does AI-ready mean for a Salesforce org?
An AI-ready org has a clean, documented data model with clear ownership, every automation tied to a named KPI, and the Einstein Trust Layer configured before agents go live. Over-customised orgs with no governance structure can't run Agentforce safely — the agents inherit whatever mess is already there.
Running a Salesforce project right now?
We'll run this pre-mortem with you in one call — and tell you exactly what's at risk before you spend another dollar.
Book a strategy call →