How much customer information our agents can see, how it's restricted, and how sure we are
Only the fields agreed for the job, through a connection your administrator grants and can revoke, and no copy of your CRM sitting somewhere else. That's how we set up our own work, and it's the standard we'd hold any provider to – us included.

Access is three different things
When you ask how much a provider's agents can see, you're really asking three questions at once, and a trustworthy answer keeps them separate:
- Read: which records and fields the agent can look at.
- Write: which fields it can change, and whether it can create or delete records.
- Retain: what, if anything, the provider keeps after the job, and for how long.
An answer that only covers the first one is incomplete. An agent that reads one field and can delete records is more dangerous than one that reads everything and can write nothing.
How we answer the three for our own work
The scope gets agreed in writing before any connection is made: which objects (companies, contacts), which fields it reads, which fields it may write, and that it never deletes. Your administrator grants the connection as a scoped integration or a named user with those permissions, in your CRM, where you can see it and switch it off. The research itself is done against public and licensed sources and the results get written into your records. The records stay yours, and we don't keep a mirror of your CRM. Outreach, where a customer switches it on, runs from your reps' own mailboxes with templates you've approved, and every reply gets filed against the account in your CRM rather than anywhere of ours.
How confident we are
Three levels, from most to least certain:
- The list: we can list the fields read and written, and you can read that list. High confidence, because it's written down and checkable.
- The connection: the scope is set in your CRM's own permission system, and your administrator can open the screen and see it. High confidence, provided the CRM's permission model is as fine-grained as the list. Some aren't, and then the straight answer is "the connection can see more than the agent uses" – which you should hear said out loud.
- The setup: people configure connections, and people make mistakes. This is where our confidence is lowest, and where the control is a routine rather than a promise: access reviewed at agreed intervals, and a log of what was written that your team can check against the scope.
A provider who tells you they're completely confident hasn't thought about the third level.
The five questions to put to any provider
| Question | What a good answer looks like |
|---|---|
| Which fields do you read, and which do you write? | A list, in writing, that matches what your administrator sees. |
| Can you delete records? | No, and the permission doesn't allow it. |
| Do you keep a copy of our data? | No, or a precise statement of what, where and for how long. |
| Can we revoke access ourselves, now? | Yes, in your own system, without asking us. |
| Show us the permission screen. | They do, and it matches the list. |
The last one is the real test. Anyone can describe their restrictions. Seeing the scope on your own screen, next to the written list, is the only time you actually know.
What to check after setup
Once a month for the first quarter, then quarterly: open the permission screen, compare it with the list, and read a sample of the log of what was written. It takes about 20 minutes, and it's the difference between restricted and believed to be restricted.