Data readiness: the universal blocker
Every platform promises super intelligence. Across all of them, the work stalls for the same reason, and it is never the model. It is the data underneath: fragmented, low in quality, ungoverned and not ready.
Data readiness is the universal blocker. Models have become a commodity; an organization’s data has not. A system is only as good as the data it reasons over, and industry research is blunt that most enterprises cannot move beyond pilots because they lack a unified data foundation.
The striking thing is how identical the story is across vendors who otherwise compete. Salesforce calls it data readiness, Databricks calls it governance, ServiceNow calls it CMDB health, Microsoft calls it oversharing. It is one truth in different words: ready the data, or the results disappoint.
- Sequence data first. Unify and govern before deploying, never the reverse.
- Unify and fix quality. Super intelligence amplifies bad data into confident errors.
- Govern. Which is also what unlocks production.
The same failure, everywhere
Look across enterprise programs and one pattern repeats. The pilot demonstrates well, then stalls on the way to production, and the examination afterwards almost always finds the data rather than the model.
This is structural. Foundation models have become a commodity and any organization can call a capable one. What differs between organizations, and what determines whether the work succeeds, is the data the model reasons over. A system inherits the quality, completeness and governance of its data, and most organizations hold data that is fragmented, duplicated and ungoverned.
Teams that believe the model is the hard part spend their energy choosing and tuning models, and stay in pilot. Teams that understand the data is the hard part fix the data and ship. Recognizing where the real blocker lives is the first and most important decision a program makes.
What data readiness means
Ready is not a vague condition. Four specific properties are needed, and a weakness in any one undermines the result.
- Unified. The relevant data is assembled into one coherent picture rather than scattered across systems holding partial and conflicting versions.
- Quality. The data is accurate, deduplicated and reconciled. This matters more here than in traditional reporting, because the system does not merely report bad data, it acts on it at scale.
- Governed. Access, lineage and controls ensure only the right data is touched and every use can be proven.
- Accessible. The data can be reached where and when the work runs, in real time, inside the boundary, without copying it somewhere it should not go.
The useful consequence is that readiness is measurable and fixable. An organization can assess where its data stands on each property and close the specific gaps, instead of treating a messy estate as an unsolvable condition.
The pattern across the five platforms
The most convincing evidence that this is universal is that five competing vendors, each selling a different story, all tell the same one underneath.
| Platform | The same truth, in its own words |
|---|---|
| Salesforce | Most enterprises cannot move beyond pilots for lack of unified data, and Agentforce depends on Data Cloud to supply it |
| Databricks | Governance is the true enterprise bottleneck, and organizations using Unity Catalog governance push about 12 times more work into production |
| ServiceNow | Operations intelligence and agents are only as good as the CMDB beneath them |
| Pega | Automated decisioning is only as good as the data it decides on, and needs clean, governed, consistent inputs to be trusted |
| Microsoft | Copilot reveals oversharing immediately, so Purview governs the data first and makes deployment safe |
Five platforms, five products, one identical lesson. When vendors who compete for the same budget independently arrive at the same root cause, the blocker is not platform specific, which means neither is the solution.
How to get ready
Because the blocker is the same everywhere, so is the fix. The discipline does not change with the vendor.
- Sequence data first. Unify and govern before deploying. Deploying agents on unready data is the most common and most expensive mistake, and it is the same mistake on all five platforms.
- Unify. Assemble the relevant data into one coherent view. The tool differs by platform, but the job is identical: one trustworthy picture instead of many conflicting fragments.
- Fix quality. Clean, deduplicate and reconcile as the work proceeds, because unification surfaces the quality problems that were always there.
- Govern. Apply access, lineage, masking and audit, which protects sensitive data and is what actually lets the work reach production.
- Scope to use cases. Ready the data the specific use cases need, which keeps the effort bounded and the cost under control.
The good news is that this is repeatable. An organization that builds the muscle on one platform can apply it to the next, because the hardest part turns out to be a single transferable discipline rather than five separate problems.
The Varmoda approach
Varmoda is data first on every platform, because the hardest and most decisive part of the work is the same everywhere.
Whatever the stack, Varmoda’s teams build the data foundation before deploying: Data Cloud on Salesforce, Unity Catalog and the Lakehouse on Databricks, a healthy CMDB on ServiceNow, clean and governed inputs for Pega decisioning, and Purview governance on Microsoft. Delivery is United States based and secure, with governance built in for the sensitive data these systems hold.
The conviction is earned across five platforms: fix the data first, and the rest works on any of them.