Set Up a Custom GPT for Carrier and Freight RFP Evaluation

This is the step most guides skip, and it matters here. As of September 2026, a personal ChatGPT account, even a paid Plus subscription at $20/month, can no longer create or publish a new Custom GPT. Only ChatGPT Business, Enterprise, and Edu workspace accounts can build one, and only when the workspace's admin settings allow it. OpenAI has also published a retirement date of December 11, 2026 for custom GPTs across the board, with a stated migration path to a plugin-based replacement.

1. Ask your IT contact or admin whether your company has a ChatGPT Business, Enterprise, or Edu workspace
2. If yes, confirm you personally have permission to create GPTs in it, since workspace admins can restrict this
3. If your company has no such workspace, or the retirement timeline makes a new build impractical this quarter, use the ChatGPT Project approach from the negotiation prep guide instead: a Project with saved rubric instructions gets you most of the same consistency without needing GPT-builder access

What you should see: A clear yes or no on workspace access and permission. Troubleshooting: If nobody at your company can answer, treat the answer as no for now and use a Project instead. Don't wait on this guide to make that call.

1. In a Business, Enterprise, or Edu workspace, click Explore GPTs in the sidebar, or go directly to chatgpt.com/gpts
2. Click Create
3. Choose either the conversational builder (describe what you want and let ChatGPT draft it) or Configure to fill in the fields yourself

What you should see: Either a chat asking what you want your GPT to do, or a form with Name, Description, Instructions, Conversation starters, and Knowledge upload. Troubleshooting: If Create is greyed out or missing, your workspace has GPT creation turned off for your account. Go back to Step 1.

1. Name it something specific: "Carrier RFP Evaluator," not "RFP Helper"
2. In the Instructions field, paste your scoring framework

You evaluate carrier and freight forwarder RFP responses for a mid-market [manufacturer / distributor]. When I paste in bid details, score each carrier on: base rate, fuel surcharge structure, transit time, historical on-time performance if provided, capacity/equipment availability, insurance and claims history, and any value-adds (tracking, dedicated capacity, etc). Weight rate and transit time most heavily unless I say otherwise. Always show your scoring math, not just a final ranking, so I can defend the recommendation to my team. Flag anything that looks like an apples-to-oranges comparison, such as one bid quoting all-in and another quoting base-plus-accessorials. Use generic labels (Carrier A, Carrier B) unless I've already made a carrier's identity non-confidential.

1. Under Knowledge, click Upload files
2. Add two or three anonymized past bid comparisons so the GPT can match your usual format
3. Save

What you should see: The uploaded files listed under Knowledge with no error message. Troubleshooting: If a file fails to upload, check that carrier names and exact pricing are stripped before you upload it, both for confidentiality and to keep the example format-focused rather than deal-specific.

1. Add two or three conversation starters, like "Compare these three carrier bids" or "Which bid wins on total landed cost?"
2. Under sharing settings, choose Only me or Only people in my workspace with the link, not a public listing, since this GPT's knowledge files may reference internal formats even after anonymization
3. Click Save or Publish, depending on the option shown

What you should see: The GPT appears in your "My GPTs" list, restricted to the visibility you chose. Troubleshooting: If a public option is the only one available, stop and check your workspace's sharing policy before publishing anything.

1. Open the GPT and paste in two or three real (or realistically simulated) bid summaries
2. Ask it to score and rank them
3. Check its math against your own quick mental estimate on at least one line item

What you should see: A structured comparison table or list with per-criterion scores and a final ranking, plus visible reasoning. Troubleshooting: If it collapses everything into a single vague ranking with no visible scoring, restate the instructions and ask it to show its work.