AI that inherits your permissions.
Most CRM AI is a chat panel wired to a separate pipeline, which is why the honest answer to “what can it see?” is usually “more than that user can”. UnifyLoop’s AI reads through the same object and query layer as every screen, so scoping and permissions aren’t re-implemented for it; they’re inherited.
- Permission model
- Inherited, not re-implemented
- Runs
- Only when you ask
- Sends messages itself
- No, drafts only
- Spend controls
- Monthly budget per workspace
It uses the product, not a copy of your database
The AI reads records and conversations through the same layer your screens use, so it sees exactly what the signed-in user can see, and no more. There is no separate index or export built for it, which is the usual reason bolted-on CRM AI leaks across teams or tenants.
Useful on the record in front of you
The AI is where the work already is: on the record, in the inbox, and inside automations. It runs when you ask rather than in the background, and every result shows which model produced it and what it cost, so it is something you can check, not a confident paragraph from nowhere.
- Record summaries, a health read on a deal, and a suggested next step
- Ask a question about any record from the command palette
- Summaries of long inbox threads, and drafted replies you edit before sending
- A “Draft with AI” step inside automations
- Build automations by describing them: typed, dictated, or by voice
- AI steps in automations that sort a reply into the right path, or pull details like budget and timeline out of a message
- Plain-language search that returns a real, reusable filter Coming soon
- Meeting summaries Coming soon
Spend has a ceiling, per workspace
AI usage runs through a budget ledger with a monthly limit for each workspace, so cost has a ceiling by construction rather than by good intentions. When a workspace reaches it, AI features stop with a clear message and everything else keeps working. Answers are cached, so asking the same thing twice doesn’t cost twice.
- An in-app breakdown of AI usage by feature Coming soon
Is my data used to train models?
No. Your workspace data is used to answer your prompts and nothing else. See our security page for the detail on subprocessors and retention.
Can the AI see records a user isn’t allowed to see?
No. It reads through the same query engine that enforces record visibility for the UI. There is no separate index with looser scoping, which is the usual reason bolted-on CRM AI leaks.
Can the AI message customers on its own?
No. In the inbox it drafts and a person sends; inside automations the “Draft with AI” step writes text rather than sending anything. Consent rules apply to every send regardless of who wrote it.
What happens when we hit our AI budget?
AI features pause with a clear message until the next month or until the limit is raised. The rest of the platform is unaffected.
Every step of the customer journey.
- 1Get foundAI writes the pages and the blog posts. To change anything, just say what you want.
- 2Lead comes inAI sums up who they are and what they asked for, right on their record.
- 3Reply in seconds“Draft with AI” writes the reply from the whole conversation. You read it and press send.
- 4Talk & bookEvery call is transcribed, so you can read it back instead of listening again.
- 5Close & get paid“Next best action” tells you what to do with each deal, and why.
- 6Bring them backAI writes each follow-up and sorts the replies into happy, neutral and unhappy.
Try it on your own data.
Free workspace, contacts and pipelines ready at signup, and every price published. No mandatory call.
Free plan · No credit card · Every price on the site