Drafts now wait for real research on a prospect before they're written, and any draft that makes a claim its research doesn't support gets held, with plain guidance on what to fix.
What changed
Before a draft goes out, FirstSales checks that it has current research on the prospect and their company to draw from, the kind of detail that makes a cold email feel like it was written by someone who actually looked into the business rather than someone guessing. If that research isn't ready yet, the draft now waits instead of going out based on stale or missing information. A draft written without solid research to back it up is more likely to get a detail wrong, and getting a detail wrong about someone's own company is one of the fastest ways to lose their attention.
If a draft does get written but contains a claim that goes beyond what its research actually supports, for example a specific detail about the company's size, product, or recent activity that the source material doesn't actually back up, that draft is held instead of being approved for sending. Instead of a generic rejection, you get plain guidance describing what the issue is and what to check, so you can see exactly which claim is the problem and correct it, rather than trying to guess what tripped the check.
This isn't limited to AI-written drafts. Manually edited drafts are re-checked before approval too. If you edit a draft yourself and introduce a claim or detail that isn't supported by the research on file, for instance adding something you heard secondhand or assumed was true, the same check catches it before the edited version goes out. Previously, an edit you made by hand was treated as automatically safe once you saved it; now it goes through the same verification as an AI-generated draft.
How to use it
You don't need to change anything to benefit from this; the checks run automatically. If a draft is waiting on research, you'll see it flagged as pending research within the campaign, rather than sitting unexplained in a queue with no indication of why it hasn't gone out. Give it time to complete, or check back once research for that contact finishes.
If a draft is held for an unsupported claim, open it to see the specific guidance describing what's flagged. Review the claim against the research attached to that contact, correct or remove it if it isn't actually backed up, and resubmit the draft once it's fixed.
When you edit a draft by hand, save it the way you normally would. It goes through the same claim check before it's marked ready to send, so if you introduce something unsupported, you'll see a hold on your own edit rather than having it slip through because it came from you instead of the AI.
Why it matters
An email that states something false or unverifiable about a prospect's company damages trust fast, and it's the kind of mistake that's hard to walk back once it's already landed in someone's inbox. A prospect who catches you getting a basic fact about their own business wrong has no reason to trust anything else in the email, and little reason to reply. Waiting for research instead of drafting on guesswork addresses that at the source, by making sure a draft only claims what its research can actually support in the first place.
Extending the same check to claims that outrun their source, and to manually edited drafts, closes the two most likely places for a bad claim to slip through: an AI draft reaching too far beyond thin research, or a human edit adding something that felt true but wasn't actually verified. Catching both means what goes out is grounded in something real about the person you're emailing, whether the words came from the AI or from you.
This is also a change in how much you personally have to remember to check. Before, verifying that a draft's claims were actually supported by real research was something you had to do yourself, if you did it at all, by cross-checking the email against whatever you knew about the prospect. Now that check happens automatically, as a required step before any draft, AI-written or hand-edited, is marked ready to send. That doesn't remove the value of reading your drafts before they go out, but it does mean the most damaging category of mistake, an outright false or unverifiable claim, gets caught even on the drafts you don't have time to review closely.
Handling a hold on your own edit
If you're editing drafts by hand regularly, it's worth treating a hold on your own edit the same way you'd treat a hold on an AI-generated one: as useful information about a specific claim, not as a false positive to work around. Check the research attached to the contact before assuming the check is wrong. If the research genuinely is missing something you know to be true from another source, that's worth adding to the contact's information directly so future drafts can draw on it too, rather than repeatedly re-adding the same claim by hand.
Why "wait" is better than "guess"
There's a real trade-off behind waiting for research instead of drafting immediately: it can mean a draft isn't ready the moment you'd otherwise expect it. That trade-off is deliberate. A draft that goes out a little later but is actually grounded in real information about the prospect is worth more than one that goes out immediately but is built on a guess, since the cost of a wrong claim, a damaged first impression, a prospect who stops reading, is higher than the cost of a short delay. The same logic applies to holding a draft with an unsupported claim rather than sending it and hoping the detail happens to be right.
Availability
This applies to all AI-generated and manually edited drafts before they're approved for sending, across every active campaign.