Skip to main content

How to build an AI assistant that reviews project briefs

Use Zapier and OpenAI to review every new request brief and post the feedback as a comment, public or internal.

Written by Gabriel Lecointere

Use case: a client submits "Design a landing page" but forgets the copy, brand assets and references. A Zap sends the brief to OpenAI as soon as the request is created, and the AI's review lands on the request as a comment, so your team (or the client) knows what is missing before work starts.

This automation uses three tools:

  • ManyRequests, your client portal

  • Zapier, to connect the tools (the ManyRequests Zapier app is in public beta)

  • OpenAI, through the ChatGPT (OpenAI) app in Zapier, to review the brief

The Zap has three steps: a ManyRequests trigger, an OpenAI action, and a ManyRequests action that posts the result.

Step 1: Connect ManyRequests to Zapier

  1. In ManyRequests, go to Settings > Integrations and click Zapier.

  2. If you haven't joined the beta yet, click the Zapier. link next to Go to on that page to accept the invite.

  3. Click Copy domain and Copy token. You paste these into Zapier when it asks you to connect your ManyRequests account.

    Zapier settings page with the Copy domain and Copy token buttons highlighted
  4. In Zapier, create a Zap and choose the ManyRequests trigger Request Created. Connect your account with the domain and API token, then test the trigger to pull in a sample request.

Step 2: Connect OpenAI

  1. Log in to platform.openai.com, add billing details and create a secret API key.

  2. In Zapier, add an action from the ChatGPT (OpenAI) app that sends a prompt and returns a reply. Connect it with your API key.

  3. Pick a current GPT model. Your Zapier and OpenAI accounts show which models you can use.

Step 3: Write the prompt

Give the AI clear instructions, for example:

"Act like a friendly project manager reviewing a creative brief. Point out missing files, links or vague goals. Reply in short sentences, without markdown, in under 200 words."

Then map the request fields from the trigger into the prompt. Useful fields are Title, Description, Service, Client name, Attachments and Form (the answers to your request form).

Step 4: Post the review on the request

  1. Add the ManyRequests action Create Comment.

  2. Set Request ID to the ID field from the trigger. Don't use Number: that's the request number your clients see (like #1001), not the ID Zapier needs.

  3. Set User Email to the email of a team member. The comment is posted under that person's name.

  4. Set Content to the reply from the OpenAI step.

  5. Choose how the review is shared with Internal Comment:

    • True: the review is an internal comment that only your team sees. Good when you want a checklist for your team before you reply to the client.

    • False (default): the client sees the comment and can answer with the missing details.

  6. Test the step, then turn the Zap on.

Good to know: internal comments need a team member's email; Zapier gets an error if you use a client's email. If you turned on blocking comments on completed requests, public comments on completed or closed requests are refused unless the email belongs to an admin.

Test and fine-tune

  • Submit a few test requests as the test client and read the AI's comments.

  • Adjust the tone and length in the prompt until it sounds like your team.

  • Add a filter step in Zapier if you only want reviews for some services.

  • Optional: add an email or Slack step at the end to tell your team a review was posted.

Related

Did this answer your question?