AI prompt to set up a custom assistant with agency instructions

Good results from AI depend on clear prompts, the right tool for the job and protecting client data. These guides help agents and staff use AI safely and effectively.

Below is a ready-to-copy prompt to set up a custom assistant with agency instructions, a detailed version that asks you clarifying questions first, and tips for using it. It works in any AI assistant: ChatGPT, Claude, Gemini or Copilot.

Try it in the search box ›

The prompt

Write custom instructions for an AI assistant used at an insurance agency: tone, compliance guardrails (no guarantees, no PII), formatting preferences, and when to ask clarifying questions.

Detailed version (asks clarifying questions first)

Act as an experienced assistant to a licensed insurance agent. Before answering, ask me up to 3 clarifying questions (your role, the AI tool you use and what you want to achieve). Then complete this task, and finish with a short checklist of statements I should verify with my compliance team.

Task: Write custom instructions for an AI assistant used at an insurance agency: tone, compliance guardrails (no guarantees, no PII), formatting preferences, and when to ask clarifying questions.

What a good result looks like

Good custom instructions are short and specific: who the assistant works for, the tone to use, firm guardrails (no guarantees, no client personal information, no coverage decisions), preferred formats such as short emails with placeholders, and when to ask a clarifying question first.

Common mistakes and watch-outs

Instructions don't make an AI tool safe for client data; check the tool's data settings and your agency's policy. Don't paste carrier manuals or confidential documents into a shared assistant without approval, and test the instructions with tricky requests before rolling them out.

How to adapt the prompt

Add your agency's lines of business, sign-off and words to avoid. Create separate versions for service, marketing and Medicare teams if their rules differ, and review them whenever your AI use policy changes.

EXAMPLE output (placeholder data, not a real client)

Custom instructions - [Agency Name] assistant

You help staff at an independent insurance agency.
Tone: friendly, clear, professional; short paragraphs.
Always: use placeholders like [First Name]; end client-facing drafts with a note to review before sending.
Never: promise coverage, savings or approval; ask for or repeat client personal information; give legal or tax advice.
Ask first when: the line of business, audience or channel isn't clear.
Format: emails under 150 words with a subject line.

How to use it

Copy the prompt. Copy the short prompt above, or the detailed version if you want the AI to ask clarifying questions first.
Add your details. Add the details for your situation (line of business, audience, tone). Keep it de-identified; never paste client personal information.
Run it in the right tool. Paste it into any AI assistant: ChatGPT, Claude, Gemini or Copilot.
Review before you use it. Have compliance or legal review the final policy before you roll it out. Check every fact, name and number yourself.
Take the next step. Next, see: Draft an AI use policy for the agency.

Which AI tool to use

Any AI assistant

ChatGPT Claude Google Gemini Microsoft Copilot

Paste the prompt above (short or detailed) into any of these.

Before you use it

Have compliance or legal review the final policy before you roll it out.
Check what each AI tool does with your data (training, retention) before staff use it.
Never paste client personal information (names, policy numbers, health or financial details) into AI tools your agency hasn't approved.
Review every AI output for accuracy and compliance before you use it.

Was this page helpful?