Custom GPT: A Proposal and Pricing Assistant Built on Your Agency's Own Templates

What This Builds

A proposal builder in ChatGPT that knows your agency's own templates, service descriptions, and pricing rules, so a first-pass proposal comes back on-brand and priced consistently, whether you write it or an account manager does. Where a plain ChatGPT conversation forgets everything between chats, this Custom GPT starts every conversation already knowing how your agency scopes and prices work.

Prerequisites

• A ChatGPT Plus subscription ($20/month); Custom GPTs require a paid plan to build and publish
• Five to ten of your agency's best past proposals or SOWs, saved as PDF, Word, or plain text files
• A written summary of your pricing structure and service tiers, even a rough one
• 1-2 hours to build, configure, and test before handing it to your team

This build has one ongoing cost: the Plus plan at $20/month. Anyone you share the finished GPT with can use it on a free account, since sharing a built GPT does not require the recipient to pay.

The Concept

A Custom GPT is like training a new account manager who has already read every proposal your agency has ever won, and who never forgets your pricing rules between conversations. You do the setup once. From then on, anyone who opens it starts from the same shared understanding instead of guessing at how your agency prices things.

Build It Step by Step

Part 1: Gather your source material

Pull together the documents that define how your agency actually works, not how you wish it worked:

• Five to ten proposals or SOWs that won deals you were happy with, spanning a few different service types if your agency offers more than one (SEO, paid ads, social, web design)
• Your pricing structure: tier names, what's included at each level, and how you decide which tier fits a given deal
• A short style note if your proposals have a consistent voice or format you want preserved

Part 2: Create the Custom GPT

1. Go to chatgpt.com and open your profile menu, then select "My GPTs" and choose "Create a GPT." Alternatively, visit the GPT creation page directly from the ChatGPT sidebar.
2. Skip the conversational builder and click through to "Configure" for direct control over the setup.

What you should see: A form with fields for Name, Description, Instructions, Conversation Starters, and a Knowledge upload section.

Part 3: Write the configuration

Name: "[Your Agency Name] Proposal Assistant"

Description: "Drafts client proposals and SOWs using our agency's templates, service descriptions, and pricing rules. Built for anyone on the team scoping a new deal."

Instructions (copy and adapt):

You are a proposal-writing assistant for a marketing agency. You have been given our past winning proposals and our pricing structure as reference material.

When someone gives you discovery call notes or a description of a prospect's needs:
1. Identify which of our standard service types the deal falls under (or flag if it is a custom scope).
2. Draft a proposal following the structure and tone of our past proposals: objectives, deliverables with specific quantities, timeline, and what is explicitly out of scope.
3. Recommend a pricing tier based on our pricing structure, and explain briefly why that tier fits.
4. If the deal does not clearly match our existing pricing tiers, say so and suggest the closest comparable instead of inventing a number.
5. Flag anything in the notes that sounds like scope creep risk (vague deliverables, undefined revision counts, open-ended timelines) so it can be addressed before the proposal goes out.

Match the tone of our past proposals: direct, confident, no filler. Do not use pricing figures that are not grounded in the uploaded pricing structure.

Conversation Starters:

• "Draft a proposal from these discovery call notes: [paste notes]"
• "What pricing tier fits a client who wants social management plus one paid ad platform?"
• "Review this scope for anything that looks like it will lead to scope creep."
• "Turn this into a proposal in our standard format."

Part 4: Upload your knowledge files

1. In the Knowledge section, click "Upload files."
2. Upload your five to ten past proposals.
3. Upload your pricing structure document.
4. Wait for each file to finish processing before moving on.

What you should see: Each file listed under Knowledge with no error state next to it.

Part 5: Test before sharing

Open the Preview panel and try it with a real (or realistic) scenario:

• "Draft a proposal for a local dental practice that wants SEO and a monthly blog post, budget around our mid tier."
• "What's included in our top pricing tier?"

Check that the draft matches your actual pricing and reads like something your agency would send. If it invents deliverables or pricing that are not in your source material, the knowledge files may need clearer labeling of tiers and prices, or the instructions need a stronger line about not inventing numbers.

Part 6: Share with your team

1. Return to My GPTs, open your new GPT, and click "Share."
2. Choose "Anyone with the link" if you want your whole team to use it without individual setup, or restrict it to specific people if your agency's plan supports that.
3. Share the link in your team's usual channel, along with a short note on how to use it (paste discovery notes, ask for a pricing recommendation, request a full draft).

Real Example: A New Inbound Lead

Setup: You've built the GPT with eight past proposals and your three-tier pricing structure (baseline, growth, full-service).

Input: An account manager pastes discovery call notes for a regional HVAC company wanting local SEO and Google Business Profile management, with a stated budget of "a few thousand a month."

Output: The GPT drafts a proposal recommending the growth tier, with deliverables scoped to local SEO plus profile management, a 90-day initial timeline, and pricing pulled from the uploaded structure. It flags that the client's stated budget range fits growth better than full-service and notes the discovery call did not mention a revision limit for content deliverables, worth clarifying before sending.

Result: The account manager reviews and adjusts the draft in fifteen minutes instead of building a proposal from a blank page, and the pricing lines up with what the agency actually charges rather than a guess.

Time saved: Proposal drafting commonly runs three to six hours when built from scratch each time; a first-pass draft from this GPT turns that into a review-and-edit pass.

What to Do When It Breaks

• It recommends the wrong pricing tier → Check whether your pricing document clearly states what separates each tier. Vague tier descriptions lead to vague recommendations; add explicit criteria (deliverable count, channel count, budget range) to the source file.
• It invents deliverables that were not in your uploaded proposals → This usually means the knowledge files did not parse cleanly. Re-upload as text-searchable PDFs rather than scanned images, and remove any proposal with unusual formatting that might confuse the parser.
• A new team member cannot find the GPT → Pin the share link somewhere permanent, like a shared drive folder or your onboarding document, rather than relying on a single chat message.
• Your pricing changes and the GPT keeps citing the old numbers → Delete the outdated pricing file from Knowledge, upload the new one, and run a couple of test prompts to confirm the update took.

Variations

• Simpler version: Use a single ongoing ChatGPT conversation with your proposals and pricing pasted in at the start of each session, no Custom GPT build required, though you will re-paste every time.
• Extended version: Build a second Custom GPT trained on your creative briefs and campaign recap format, so the same knowledge-base pattern covers delivery work, not just new business.

What to Do Next

• This week: Test the GPT against two or three real deals in your pipeline and compare its draft to what you would have written unassisted.
• This month: Add your team's actual usage patterns back into the instructions. If everyone keeps asking it to check for scope creep risk, make that check more prominent.
• Advanced: Connect this GPT's proposal structure to the automated weekly reporting pipeline's report template, so new-client proposals and ongoing reports share the same visual and pricing language.

Advanced guide for marketing agency owner professionals. Requires a ChatGPT Plus subscription to build and publish. Anyone using the shared GPT link can do so on a free account.