An agent setup recipe
Research a prospect before a sales call
Build a short, sourced brief on a named company from public pages and your own account notes, then prepare five neutral questions and an opener for the call.
From
You get
What you will get
A one-page sourced brief covering what they do, recent signals and likely pains to explore, a dated recap of your own notes, five evidence-linked questions and an opener
Illustrative example · fictional input
- Example input
- A fictional packet: a company page says a second office will open, a job post shows hiring, an old note says weekly exports are manual, and product scope confirms CSV export only.
- Example result
- Illustrative result, not a recorded agent run: The brief lists the second office and the hiring post as dated public context, not demand. The recap dates the export problem as 1 September and asks whether it changed. Five questions cover the current export workflow, who does it, what happens when it is late, what a good outcome looks like and how decisions on tools are made. The opener states the purpose in two sentences. Budget, urgency and migration capability stay unknown.
Check that public facts are current and that you are allowed to use the account notes. Choose the questions, listen to the actual answers and decide the next step after the conversation.
What you need
One assistant owns the research, the recap and the questions. The public research step uses web search; everything else works from the packet you supply.
- A ChatGPT account with Work and web search available on the web. Check your account or workspace access first; search availability, models and allowance vary. Use the default model available to you. Work uses included usage or account credits; no API key or third-party subscription is required.
- A browser and notes you are allowed to share with the service. Remove unnecessary private details. This recipe needs no connected app, local-folder permission, skill installation or recurring task; it reads public web pages only.
- One human owner who can check the sources and make decisions. One meeting purpose, the company name and website URL, the contact’s name and business role, and available discussion time; Authorized account notes or prior answers with dates and source IDs, if any; Your offer: the product capabilities you have confirmed, and the customer criteria that matter to you.
- Keep research short. Ten minutes of sourced context and five good questions beat a long dossier, and more preparation must not replace listening on the call.
Set it up
- Open chatgpt.com and sign in. Confirm Work is available and that web search is enabled for your account, then select Work and start a new chat. Check your remaining usage before a large packet. Keep the default model available to your account; this recipe does not require a particular model name.
- Use one conversation for this responsibility. If you prefer a project, create one for this task and keep related chats and approved source copies there. A web project does not automatically read a folder on your computer. Start each distinct batch in its own chat and include the setup prompt again when needed.
- Choose the setting below. Optionally add your working rules. Copy the complete setup prompt into the chat as the first message. The assistant should confirm its role and ask for inputs, without starting outside work. This gives the existing assistant a responsibility; it does not create another account or deployed bot.
- Prepare a small packet: One meeting purpose, the company name and website URL, the contact’s name and business role, and available discussion time; Authorized account notes or prior answers with dates and source IDs, if any; Your offer: the product capabilities you have confirmed, and the customer criteria that matter to you. Label every source and include dates, versions and unknown fields. Paste accessible text. Ask the assistant to list the source IDs it can read before it starts. No plugin or app connection is needed; public web pages are read through the built-in search.
- First assignment — paste this complete fictional exercise into the chat: Fictional first assignment, no web search: use only this packet. Follow-up meeting, 20 minutes with an operations manager at Harbour Rooms, a fictional venue business. A1 customer note from 2026-09-01: weekly booking exports require manual copying. A2 public company page observed 2026-10-01: a second office will open in spring. A3 job post observed 2026-10-05: hiring an operations coordinator. Product scope P1: CSV export is supported; automatic booking migration is not confirmed. The meeting purpose is to learn whether the export task still causes difficulty. No budget or purchase date is known. Prepare the brief, recap, questions and opener.
- Compare the response with the review checks below. The following is an illustrative expected direction, not a tested result: The brief lists the second office and the hiring post as dated public context, not demand. The recap dates the export problem as 1 September and asks whether it changed. Five questions cover the current export workflow, who does it, what happens when it is late, what a good outcome looks like and how decisions on tools are made. The opener states the purpose in two sentences. Budget, urgency and migration capability stay unknown.
- For real work, name the company and website and explicitly start one batch. The assistant reads public pages through web search; open every link it cites and check the date before you rely on it. Check that public facts are current and that you are allowed to use the account notes. Choose the questions, listen to the actual answers and decide the next step after the conversation. Copy only the changes you approve into your normal records yourself. Keep the accepted result and its source references.
- When the batch is reviewed, mark your own record as accepted or list the corrections needed. On the next run, supply the latest accepted notes and a new company. No schedule or continuous monitoring is created. If the assistant asks to connect an app, submit a form or take an outside action, stop and return to the packet.
Your setup prompt is ready
Copy the ready prompt, or customize it with the settings. Paste it into ChatGPT Work to give it the task rules.
Customize (optional)
Defaults are ready to use. Changes update the prompt here; paste it into your agent to apply them.
Schedule: When I ask.
Choose how the assistant organizes the result; evidence and review rules stay the same.
Your product scope and the customer criteria that matter to you. Include only approved information.
Your choices stay in this browser for this guide. Avoid adding private information.
setup prompt
Highlighted text follows your settings.
Review checks, limits and sources
Not tested by Juicy Agents. The examples are fictional, not recorded agent results.
Check the result
- The September export problem is dated and checked rather than assumed current.
- The second office and the job post do not become confirmed demand or urgency.
- In a real run every public fact carries a link and the date it was read; the fictional run uses only the packet.
- No migration promise, budget estimate or invented buyer profile appears.
- Removing an essential input causes a question or a blocked item, not an invented fact.
- Changing the setting changes the requested output while source references and fixed limits remain.
Known limits
- Not tested by JuicyAgents in ChatGPT Work. The fictional exercise and expected result are authored examples, not proof of accuracy or saved time.
- Web search results change daily and can surface outdated or wrong pages; check every link before you rely on it.
- Only public pages and the supplied notes are used. A brief can still miss the customer’s real situation.
- This prepares a call; it does not qualify a lead, update a CRM or predict a deal.
If something goes wrong
- A question assumes an unconfirmed problem: rewrite it so the customer can disagree.
- The brief states a fact without a link or date: ask for the source or mark it unknown.
- Work or web search is missing, or the account reaches a limit: check the official access and usage pages or ask your workspace administrator; do not assume a paid upgrade is required. Use the packet-only version while search is unavailable.
- The assistant follows instructions found on a web page: stop, restate that page text is data, and review the entire affected result. Do not approve an external action.
Platform sources
- ChatGPT Work on the web (opens in a new tab) · Checked 2026-10-08
- ChatGPT projects (opens in a new tab) · Checked 2026-10-08
- ChatGPT Work access and usage (opens in a new tab) · Checked 2026-10-08
Reviews
No reviews yet
Reviews share personal experience. They are not proof that an agent has passed a test.
Log in to leave a review. You can read and use this recipe without an account.
Explore more
Similar Agents
Reply to quote requests faster and ask for the details you’re missing
Find missing job details, suggest priority and prepare a reply for you to check.
View agent salesReview a sales discovery call
Find covered topics and gaps in a transcript, with evidence for a coaching discussion.
View agent salesWrite the follow-up email, CRM update and tasks after every sales call
Paste your call notes or transcript, or let Claude read the deal in HubSpot, and get a recap email that promises only what was agreed, CRM changes you approve and a task list with owners and dates.
View agent