Salesforce Does Not Trust What Your Reps Typed Either

The short answer. In a Dreamforce 2026 session, a Salesforce data governance lead described the data hygiene behind one of the company's oldest production orgs: 2,600 objects, tens of terabytes, and third-party vendors used to certify account data rather than relying on what a sales rep entered years earlier.

The most reassuring session at Dreamforce 2026 was the one where Salesforce described its own org. Not the demo org. The real one, twenty-five years old, that the company runs its business on.

What does it mean to be a 25-year-old org at Salesforce? Well, that's like a lot of scary numbers that you see here. … Over 31 terabytes of data.

Source: Build Trusted Data for Agentforce with Data 360, Dreamforce 2026

Later in the same session, the scale gets worse, and funnier:

What do you do when you have 2,600 objects in your org, in one org?

Source: Build Trusted Data for Agentforce with Data 360, Dreamforce 2026

Two thousand six hundred objects. If you have ever felt bad about the state of your org, read that sentence again and feel slightly better.

The admission that stopped us

Then the governance lead described how they establish whether account data is correct. This is the part worth sitting with:

The third thing that we're doing to understand our customer data is leveraging external vendors, Moody's, DNB, Leadspace, so that we certify our customers and our accounts, so that we don't rely on a sales rep entering that customer or that account data years ago.

Source: Build Trusted Data for Agentforce with Data 360, Dreamforce 2026

Salesforce pays three outside vendors to certify account data rather than relying on what its own sellers typed into Salesforce.

We want to be careful with that sentence, because it would be easy and cheap to turn it into a joke. It is not a gotcha. It is good practice, and any company at that scale that claimed otherwise would be lying.

But it settles an argument that many admins have been losing internally for years. When you say the account data cannot be trusted because it was typed by someone in a hurry in 2019, you are not being cynical about your colleagues. You are describing the normal condition of CRM data, and the company that makes the CRM handles it the same way you would if you had the budget.

Data hygiene you can copy without a budget

Three things from that session transfer directly, and none of them need a vendor contract.

First, assess before you argue. The session claims a timescale worth holding them to:

When you assess your data in your org, you can get these insights in one to five days. So you can make informed decisions, you can go to your business stakeholders and say, hey, we apparently have one out of five times incomplete data about our leads.

Source: Build Trusted Data for Agentforce with Data 360, Dreamforce 2026

One in five leads incomplete is a sentence a business stakeholder can act on. "Our data is bad" is not. The difference between those two sentences is a week of profiling, and it is the highest-leverage week an admin can spend.

Second, make cleanup a standing operation rather than a project. The same speaker was blunt that this is a discipline problem inside Salesforce too, and that it rarely gets the priority it needs.

Third, and this one is free: stop deprecating fields by consensus. Their approach is to hide unused fields from end users immediately and mark them as deprecation candidates, rather than waiting for every dependency to be resolved. That unblocks a cleanup that otherwise never starts.

The uncomfortable implication for AI

Put the two halves together. The session's premise is that agents are only as good as the records they read. And Salesforce buys external certification because it will not rely on rep-entered account data.

If you are pointing an agent at account data in your org and you have done neither of those things, you now know what the risk actually is, from the people best placed to know.

And the fix has two halves: knowing which records are wrong, which is what profiling and enrichment do, and correcting them, which is still somebody typing, at whatever speed your tooling allows.

Where we come into this

We sell the second half, so discount accordingly.

Valorx Wave takes the list a profiling pass produces and makes fixing it fast. The records open in an editable grid inside Salesforce, bulk changes are typed like a spreadsheet, and every change is shown old value beside new, so you can revert what you disagree with before saving them together.

You will probably never buy three data vendors. You can still get the week back.

Frequently asked questions

How does Salesforce manage its own CRM data hygiene?

In a Dreamforce 2026 session, a governance lead described profiling the org, using external vendors to certify account data rather than relying on rep entry, enforcing matching and merging rules with validation rules, hiding unused fields immediately, and archiving old data.

How many objects does Salesforce have in its own org?

The session described one 25-year-old org with 2,600 objects and more than 31 terabytes of data.

How long does a Salesforce data assessment take?

The session put it at one to five days to get insights you can take to business stakeholders, such as the proportion of leads with incomplete data.

Sources

Quotes are verbatim from the session captions, with punctuation and capitalization added for readability, and link to the Salesforce+ recording. The speaker describing Salesforce's own org identifies as a Salesforce employee; other speakers in the session do not work for Salesforce. Figures given on stage are as stated and were not independently verified.

1. Build Trusted Data for Agentforce with Data 360, Dreamforce 2026, Salesforce+