--- updatedAt: 2026-07-21T20:49:03.000Z agentTools: projectIndex: https://docs.conveyor.com/llms.txt --- # Using the AI Connector **Time to complete:** 5–10 minutes **Prerequisites:** An AI client connected to the [Conveyor AI Connector](ai-connector) — see [Connecting Your AI Client](connecting-your-ai-client). *** ## Overview The AI Connector is at its best when your assistant combines several tools in one conversation. This page collects copy-paste prompts for the most common jobs — answering security questions, staying on top of questionnaires, keeping knowledge current, running your Trust Center, and reporting on your program. Prompts marked ⚠️ change data in Conveyor — review the assistant's confirmation before it applies. *** ## Writing Effective Prompts The AI Connector finds things by the words you give it. Five habits separate a clean result from a wrong guess: * **Name things specifically.** Your assistant looks up documents, customers, and questionnaires by name — there are no ID numbers to memorize. "Our SOC 2 Type II report" lands one clean match; "the report" makes it guess or ask. * **A clarifying question is a feature.** If two records share a name, the assistant asks which you mean — that's it keeping you from editing the wrong one. Answer it, don't fight it. * **Give analytics a time window.** "Last quarter," "since the start of the month," "the past 30 days." Leave it out and you get everything. * **One change per prompt.** For anything that grants access, changes a status, or deletes — keep it single and explicit so you can confirm before it happens. * **For read-only questions, go broad.** "Summarize Trust Center engagement this month and flag anything odd" is fine — let the assistant pull and synthesize. *** ## Answer Security Questions > *"Do you encrypt data at rest?"* Answers pull only from the Knowledge Base and documents you've approved, so anyone can self-serve without waiting on the trust team — and every answer traces back to a source you signed off on. *** ## Stay on Top of Questionnaires **See what's waiting on you** > *"Which questionnaires are waiting on my review? Sort them by due date."* **Run a weekly check-in** > *"Show me all questionnaires that are due in the next 7 days, grouped by status. Flag any that are still in 'started' and haven't moved to review yet."* **Spot emerging question trends** > *"Find every questionnaire question we've received about AI or data training, grouped by theme."* Use a search like this to build out a Knowledge Base section before a topic becomes a common blocker. *** ## Keep Your Knowledge Current **Find stale or unreviewed Q\&As** > *"Find the curated Q\&As marked unverified so I can review them."* **Audit a topic** > *"Show me every Knowledge Base entry related to encryption, grouped by relevance. Flag gaps and any unverified entries."* **Find Knowledge Base gaps** > *"Which recent questions couldn't ConveyorAI answer confidently? Group them by topic — include questions from the Trust Center Agent, questionnaires, and one-off asks."* **Add a curated Q\&A** ⚠️ > *"Add a curated Q\&A to the Knowledge Base. Question: 'Do you support SSO?' Answer: 'Yes — we support SAML 2.0 SSO on all plans.'"* **Push a fact change through the library** ⚠️ > *"We've migrated our infrastructure from AWS us-east-1 to us-east-2. Find every Q\&A that references us-east-1 and propose updated wording for each — I'll review before anything changes."* *** ## Run Your Trust Center **Approve access in bulk** ⚠️ > *"Find all pending access requests from @stripe.com email addresses, approve them, and add them to the Enterprise access group."* Routine Trust Center administration — approving access requests, managing invitations, updating documents — works through the same conversation. Keep changes like this single and explicit so you can confirm each one. *** ## Report on Your Program **Spot your most engaged customers** > *"Pull this month's Trust Center interactions and cross-reference with our connections list. Who are our top 5 most engaged customers, and what are they downloading?"* **Find customers going cold** > *"Which customer connections have had no Trust Center activity in the last 60 days?"* **Trace a commitment you made** > *"Search our past answers: which customers did we tell that we support on-premises deployment?"* When something changes, a search like this shows you the blast radius before customers find it for you. **See who's asking what** > *"Pull the access logs for the past 30 days. What has the sales team been asking, and when?"* *** ## When to Use the Conveyor App Instead A few jobs are better done in the Conveyor app — reach for these through your assistant and you'll fight the tool: * **Bulk imports** — hundreds of Q\&As or documents at once. The AI Connector writes one record at a time; use the in-app import built for batches. * **Large file uploads** — big reports or anything near the 100MB limit. Upload them in the app, not through chat. * **Anything you can't easily undo** — deletes and access revokes. Do them one at a time, and confirm before you commit. *** ## Common Questions **Q: Are answers ever made up?** A: Answers are grounded in the Knowledge Base and documents you've approved inside Conveyor — the same source your trust team already maintains. If the knowledge isn't there, the assistant can't invent an approved answer for it. **Q: What happens when a prompt changes data?** A: The assistant confirms what it's about to do before it applies. Keep write prompts single and explicit — one change per prompt — so the confirmation is easy to check. *** ## What's Next? * [AI Connector overview](ai-connector) — how authentication and permissions work. * [Knowledge Tools](knowledge-tools) — keep the knowledge your assistant draws from current. *** **Need help?** Visit the [Integration Issues guide](integration-issues) or contact .