Mining and Adding Your Highest Value Prompts

Most teams treat prompt sourcing as a one-time setup task, missing the high-value prompts already hiding in their own data. In this lesson, Dillon walks you through a repeatable process for uncovering, prioritizing, and adding the prompts that move the needle for your brand.

What You'll Learn

TL;DR

  • Your highest-value prompts come from first-party data. Sales calls, support tickets, and Slack threads capture the exact questions buyers ask at the moment of decision.
  • Four source types carry different signal strength. Voice of the customer sits at the top; SERP mining sits at the bottom. Prioritize accordingly.
  • MCP connects your data sources to AirOps through AI. Connect VoC tools, extract questions with Claude, and route prompts directly into tracking.
  • Prompt mining is continuous, not a one-time setup. Keep VoC sources connected so new questions flow into AirOps as they surface in conversations.
  • First-party prompts unlock strategic use cases beyond visibility. Revenue-focused tracking, early risk detection, and market consensus monitoring all depend on prompts your competitors cannot access.

Your best prompts are already hiding in your organization. They live in sales calls, support tickets, Slack threads, and discovery conversations. In this lesson, Dillon walks through how to find those prompts, extract them with AI, and route them into AirOps for continuous tracking using answer engine optimization (AEO) tools and MCP (Model Context Protocol) connections.

Most teams start AEO tracking by manually building prompt lists or pulling from third-party keyword tools. That approach gives you the same universe of questions every competitor already has. The real advantage comes from first-party data: the exact questions your buyers ask, in their own words, at the moment they are making a decision.

Where your highest-value prompts come from

Not all prompt sources are equal. Each one carries different signal strength, scope, and proximity to your buyer. Dillon breaks them into four categories.

  • Voice of the customer (VoC): Gong calls, support tickets, sales conversations, and chat logs. These capture exact buyer language tied to deals and buying stages. The scope is narrow, but the signal is precise because it comes from people actively evaluating or using your product.
  • Social discovery: Reddit, forums, and community threads. These surface unfiltered, real-world language that no keyword tool captures. The data is category-dependent and noisy, but it reveals how people actually talk about problems in your space.
  • Answer engine panels: Market-level signal from platforms like ChatGPT, Perplexity, and Gemini. Useful for understanding what the broader market asks, but not specific to your buyer or tied to any deal context.
  • SERP mining: Broad category coverage that transforms traditional keywords into natural-language prompts. This reflects how people search on Google, not how they phrase questions in AI search.

Why first-party data is the priority

Your buyers' words, captured at the moment of decision, are the most valuable input for AEO tracking. Here is why:

  • 60% of ChatGPT prompts are 10 or more words. These long-tail questions dominate AI search, and they are exactly the kind of phrasing your customers use in real conversations.
  • Third-party prompt lists give every competitor the same starting point. First-party data gives you prompts competitors cannot access without your customer relationships.
  • When you build tracking around first-party questions, you turn a generic AEO setup into a proprietary asset. Your prompt library reflects your market, not a shared industry average.

How to add prompts through MCP

Dillon demonstrates a three-step process for connecting first-party data sources to AirOps through MCP, the protocol that lets AI assistants like Claude and ChatGPT interact with external tools.

  • Step 1: Connect your VoC sources. Set up MCP servers for tools like Gong, Zendesk, Intercom, and Slack so your AI assistant can access conversation data directly.
  • Step 2: Extract questions with AI. Use Claude or ChatGPT to process transcripts, support tickets, and chat logs. The AI surfaces the natural-language questions your buyers actually type into AI search.
  • Step 3: Add prompts to AirOps automatically. Extracted questions flow directly into AirOps tracking, assigned to topics and ready to monitor. You skip the spreadsheets and manual entry entirely.

Live demo: from transcripts to tracked prompts

In the lesson demo, Dillon connects Claude to AirOps MCP and several first-party data sources. He asks Claude to pull top questions from a set of transcripts. Claude uses brand context from AirOps MCP to recommend topic assignments for each prompt. The prompts are added directly to AirOps and appear in the tracking interface within seconds.

Building a continuous prompt pipeline

This process is not a one-time migration. New customer conversations generate new questions every day. When your VoC sources stay connected through MCP, new prompts trickle into AirOps tracking as they surface in calls, tickets, and chat threads. Your prompt library stays current with what buyers ask this week, not what they asked last quarter.

Strategic use cases for first-party prompts

Once first-party prompts are flowing into AirOps, you can use them for more than basic tracking.

  • Revenue-focused tracking: Prioritize prompts that correlate with revenue instead of volume. Questions pulled from close-won deal conversations carry more strategic weight than high-traffic generic queries.
  • Early risk detection: New objections from discovery calls and support tickets surface automatically. You can spot emerging competitive threats before they appear in market-level data.
  • Market consensus and pricing: Track whether new messaging from product launches, pricing changes, or positioning shifts establishes consensus across AI search. Pull prompts from Slack conversations and launch documents, then monitor whether the market echoes your message or ignores it.

What comes next

With AirOps and Claude MCP, prompt mining becomes a continuous process. New questions arrive from customer conversations as they happen, giving you an edge competitors cannot replicate. In the next lesson, Dillon covers how to read the metrics those prompts generate so you can act on what you are tracking.

Key takeaways

  1. Not all prompt sources carry equal signalDillon ranks four source types by proximity to your buyer. Voice-of-customer data from Gong and support tickets sits at the top because it captures exact language tied to deals and buying stages. SERP mining sits at the bottom because it reflects how people search Google, not how they phrase questions in AI search.
  2. Close-won deals outrank search volume for prompt prioritizationWhen you mine prompts from voice-of-customer data, weight them by revenue signal, not just search volume. Prompts that surface in close-won deals represent the exact language your highest-value buyers use when evaluating solutions.
  3. Connect Claude to AirOps MCP and first-party data at the same timeIn Dillon's demo, Claude reads brand context from AirOps MCP while simultaneously pulling from first-party sources like Slack conversations and launch documents. This dual connection lets Claude recommend topic assignments for new prompts automatically, based on your existing Brand Kit categories.
  4. Schedule prompt mining to run continuouslyPrompt discovery is not a one-time audit. You can set prompt mining on a recurring schedule so new high-value questions flow into your AirOps tracking pipeline without manual effort. New prompts appear in the tracking interface within seconds of being added.
  5. Support tickets surface competitive threats before market data doesNew objections from discovery calls and support conversations flow into your prompt library automatically through MCP. You can spot emerging competitive questions before they show up in third-party keyword tools or answer engine panels.

FAQs

Third-party keyword tools give you the same prompt universe your competitors already have. First-party data from sales calls, support tickets, and Slack threads gives you the exact questions your buyers ask in their own words, tied to real deal context. When you build AEO tracking around those questions, your prompt library becomes a proprietary asset that competitors cannot replicate without your customer relationships.