How to use Responsive in ChatGPT, Microsoft Copilot, and Claude

Bring trusted knowledge and live project work into the AI tools your teams already use, then save work back to Responsive.

Your teams already use AI assistants to write, summarize, and prepare. What those assistants don't know is which of your company's answers are current, approved, or safe to send to a customer.

Connecting Responsive to ChatGPT, Microsoft Copilot, or Claude closes that gap. Your trusted knowledge and live response work show up in the tools people already have open, and the sources, permissions, and reviews behind every answer come along.

This guide gets you started fast. It follows one deal, a security review with Acme Corporation, through three sets of hands: a seller who needs answers by Thursday, an expert who writes the one answer nobody had, and a proposal manager who makes sure no one has to write it again.

The connection works both ways

Most AI integrations only fetch. This one also writes back, so the work you do in your assistant doesn't get stranded in a chat window.

• Content Library answers with sources and TRACE Scores™, your open projects, and the questions assigned to you. Let’s call this “Bring in.” 
• Draft, tailor, summarize, and turn answers into customer-ready work. This is “Act.”
• Responses to projects, new Q&A pairs, metadata, reviews, and moderation all land in Responsive, which stays the system of record. This is how you “Save back.”

Keep this loop in mind as you read our guide: Bring in, Act, Save Back. Each role below enters it at a different point, and each one leaves it in better shape for the next.

Find your starting point

You don't need to read this guide end to end. Find your role, see what you get, and jump to that section.

Each walkthrough is shown in one assistant. You can try similar workflows in ChatGPT, Microsoft Copilot and Claude, though available actions and results may vary.

Before you start

You need four things, and most people already have three of them:

• An active Responsive account, on a plan that includes MCP Server capabilities.
• A supported AI assistant. ChatGPT, Microsoft Copilot, or Claude.
• The right permissions. Your existing Responsive role decides what you can see and save.
• A configured connection. Your admin sets this up once for your organization.

Where the connection lives depends on your assistant:

One rule covers everything that follows: your assistant sees and does only what your Responsive role allows, and saving work back follows the same rules.

Your first request: ask one question you'd normally look up

The easiest way to begin is to ask your assistant something you'd otherwise go find in Responsive. Try this:

"What's our approved answer on whether customer data is used to train AI models? Include the source and TRACE Score™."

Your assistant searches your Content Library and returns three things:

• The answer. The approved response, ready to use or adapt.
• The source. Where the answer lives in Responsive, so you can check it yourself.
• The TRACE Score™. A rating from 0 to 100 of how complete and well-sourced the answer is. Higher is better, and a low score is your cue to verify before you send. 

Answer, source, score. Every workflow in this guide builds on that pattern, so it's worth seeing it a few more times before you move on:

"Do we have a current SOC 2 Type II report, and when was that answer last updated?"

"What's our approved answer on SSO and MFA enforcement?"

"Summarize our incident response process in three sentences for a non-technical buyer."

Notice what the last prompt does. It doesn't stop at finding the answer; it reshapes it for an audience. That move, from trusted answer to finished work, is where field teams spend most of their time.

Field teams: walk into the meeting with answers you can stand behind

Shown in ChatGPT. Also works in Copilot and Claude.

An account executive has a meeting with Acme Corporation's security team on Thursday. They need answers they can trust, fast, plus something to send ahead of the meeting. Normally that means a message to the proposal team and a day of waiting.

Step 1: Ask everything at once

You don't have to go one question at a time. Give your assistant the whole list and tell it what you want back for each:

"I'm meeting Acme Corp.'s security team Thursday. Using Responsive, answer these questions. For each one, give the approved answer, the source, and the TRACE Score™, and flag anything that needs an SME:

1. Do we have a current SOC 2 Type II report?
2. How do we enforce SSO and MFA?
3. What's our incident response and customer notification process?
4. Is customer data used to train AI models?
5. Which AI subprocessors do we use?"

Step 2: Read the scores before you read the answers

The answers will all sound confident. The sources and TRACE Scores™ tell you which ones have earned it.

• A high score with a source means the answer is well supported.
• A low score means the supporting content is thin or incomplete. Check the source before you repeat it to a customer.
• A gap, where your assistant finds no approved content, is not a prompt to improvise. It's a signal to bring in an expert.

In this scenario, four answers come back strong. Question 4, on AI model training, comes back weak: the Content Library has nothing current on it. Hold that thought.

Step 3: Turn answers into customer-ready work

Now put the strong answers to work:

"Turn these answers into a one-page executive overview of our security posture for Acme's CISO. Use only the Responsive content above, keep it non-technical, and end with three suggested discussion topics."

From the same answers, you can also ask for a pre-meeting email, three talking points, or a slide outline. The phrase that matters is "use only the Responsive content above." It keeps your assistant working from approved material instead of filling in from general knowledge.

SMEs: answer once, right where you already work

Shown in Microsoft Copilot. Also works in ChatGPT and Claude.

After Thursday's meeting, Acme Corp.'s security questionnaire arrives as a project in Responsive. The security lead is assigned the AI-governance questions, including the one the Content Library couldn't answer. They work from Copilot, where they already spend their day, and never open another tab.

Step 1: See what's waiting for you

Start with what's on your plate:

"Show me my open projects in Responsive."

Then narrow to the one that matters:

"Let's work on the Acme Corp. security questionnaire. Which questions are assigned to me, and which are still unanswered?"

Step 2: Draft from approved content, then add what only you know

Let your AI assistant do the first pass from what's already approved:

"Using the Responsive Content Library, draft answers to my assigned questions. Use only approved content, cite sources, and flag any claim that our current content doesn't support."

Most of the drafts will be close. Question 4, on AI model training, comes back flagged because the library has no current answer to draw on. This is the part no assistant can do for you. You know the policy, so you say it:

"Update question 4 to say that customer data is never used to train third-party models, and add our 30-day retention commitment."

That's the division of labor. The assistant handles retrieval and drafting, and the expert supplies the judgment and stands behind the claim.

Step 3: Save back to the project

When the answers are right, put them where the team can use them:

"Save these responses to the Acme Corp. project in Responsive."

Your assistant may ask you to confirm before it writes anything. Once saved, your answers sit in the project like any others, ready for review.

While you're there, you can also leave a note or close out a review:

"Add a comment for the proposal manager on question 4: this answer is new and reflects our current AI data policy."

"Mark my review of the AI-governance section complete."

The questionnaire is unblocked. But that answer to question 4 is worth more than one deal, and the proposal manager has noticed.

Bid and proposal teams: turn one good answer into the next ten

Shown in Claude. Also works in ChatGPT and Copilot.

The proposal manager sees that the SME's AI-governance answer is new and strong. They also know the next ten questionnaires will ask the same thing. Left in the Acme project, that answer helps one deal. In the Content Library, it helps all of them. They make the move without leaving Claude.

Step 1: Create a Q&A pair

Start by turning a deal-specific response into a general one:

"Take the SME's saved response to question 4 on the Acme Corp. project and create a new Q&A pair for the Content Library. Write a general question that will match how buyers phrase it, and remove anything specific to Acme."

Step 2: Add metadata

An answer nobody can find is no better than one that doesn't exist. Tag it so it surfaces:

"Tag it Security, AI Governance, and Data Privacy, set the language to English, and set the owner to the InfoSec team."

Step 3: Send it through moderation and review

New content goes through the same approval process as anything else in your library:

"Send it for review by the security SME, and set it to expire for re-review in six months."

Step 4: Save to the Content Library

Finish by saving and asking for a receipt:

"Save it to the Content Library and confirm what you created."

The loop closes

Remember the seller's question 4, the one that came back weak? The next time anyone asks whether customer data is used to train AI models, the answer comes back with its source and TRACE Score™. One expert answered once, and every seller after that gets it in seconds.

Keep active projects moving

Capturing knowledge is half the job. The other half is running the live project, and the same connection helps there too. Three prompts cover most of it:

"Which questions on the Acme Corp. project are unanswered or unreviewed, and who owns them?"

"Where is our Content Library coverage weak for this questionnaire? Group the gaps by SME type and draft the question I should send each one."

"Review the saved responses for unsupported claims and flag any that need another look. Don't rewrite them."

That last instruction, "don't rewrite them," is deliberate. You want your assistant to point at problems so the owner can fix them, not to quietly change an expert's answer.

What your assistant can read and write back

The walkthroughs show the loop in motion. This table is the same loop as a reference: every action available from your assistant and who typically uses it.

Reading never changes anything. Writing back always follows your Responsive permissions, so an action you can't take in Responsive is one your assistant can't take for you.

Build your own prompts

The prompts in this guide aren't magic words. They share a structure, and once you see it, you can write your own for any situation. A good Responsive prompt has five parts:

1. Context. Say what you're working on. "I'm preparing a security questionnaire response for Acme Corp."
2. Task. Say what you want done. "Find our strongest approved answers to these questions."
3. Constraints. Say where the boundaries are. "Use only current approved Responsive content. Don't invent information."
4. Output. Say what shape the result should take. "Return the results in a table with the question, recommended answer, source, TRACE Score™, and SME review needed."
5. Save back. Say where the work should land. "Save the final answers to the Acme Corp. project in Responsive and tell me what you changed."

The first four parts get you a good answer. The fifth is what makes the work last, and it's the one people forget.

Templates to copy and adapt

Replace the bracketed parts with your own details.

Security, permissions, and control: Responsive stays the system of record

Giving an AI assistant the ability to write raises a fair question: what stops it from writing the wrong thing in the wrong place? The answer is that nothing about your governance moves. Permissions, sources, metadata, and review processes all stay in Responsive. The assistant is a new way in, not a way around.

Write-back follows your role. Your assistant can only do what you could do yourself in Responsive. An SME can save a response to a project they're assigned to, but can't publish to the Content Library unless their role allows it.

Review still happens. Content created from an assistant enters the same moderation and approval process as content created anywhere else. Saving a Q&A pair is not the same as approving it.

Getting better results

Good results start with specific prompts. Four habits make the biggest difference:

• Say "using Responsive" or "from the Content Library" so your assistant draws on approved content, not general knowledge.
• "Use only approved content" and "flag anything our content doesn't support" turn silent guesses into visible gaps.
• "Save this to question 4 on the Acme Corp. project" works. "Save this" leaves your assistant guessing.
• End with "List what you wrote to Responsive and where." You get a record, and you catch mistakes while they're easy to fix.