Every AI feature in Trazer stops one step short of acting. It reads the deal, writes the draft, extracts the notes, ranks your morning. Then it waits for you. That gate is the product, not a setting you can switch off.
The same four steps sit behind every AI feature on this page. There is no path around them.
Only the records your role can already reach.
Watermarked as AI output from the moment it exists.
Edit, rewrite or discard. The watermark clears only by your action.
The send is yours, and it is recorded as yours.
Every deal carries a score out of 100. It is not a prediction from a model nobody can inspect. It is five signals with fixed weights, and you can see every one of them.
That matters when a rep disagrees with the ranking. With a black box the conversation ends in a shrug. Here it ends in a look at which signal moved and why.
Ask a question in plain English and get an answer with clickable citations to the records behind it. You verify rather than trust, and it only ever reads what your role permits.
Dictate after a call. It pulls out the contacts, the commitments and the dated tasks, and proposes them against the deal for you to accept.
Incoming mail is classified, matched to the right deal, and given a suggested reply that sits in your queue until you clear it.
A prioritised digest each morning: what is overdue, what has gone quiet, and what is waiting on you. Computed from your pipeline, not guessed at.
Worst, likely and best case, weighted by stage probability, with risk flags on the deals that threaten the number. The likely figure ties back exactly to the sum of value times probability.
Every AI call is recorded against your own workspace. Your administrators can see how much is being used and what it costs, rather than discovering it on an invoice.
One honest note on record-keeping: the running log of accepted and rejected suggestions is currently stored per device in the browser, not as a durable server-side record. If you need a permanent AI decision trail for an audit, tell us, because that is a change we would need to make rather than a claim we will stretch.
Worth being explicit, because most of what makes AI risky in a business system is what it is permitted to do unsupervised.
AI features run on a key enabled per workspace. Where it is not switched on, scoring, stale flags and the digest continue to work, because those are rule based rather than AI generated.
Bring a real deal and a real call note. We will run them through live, and you can judge the output rather than the description of it.
Request access